#qi-hardware IRC log for Saturday, 2011-08-20

qi-bot[commit] kyak: killall gmenu2x.bin (not gmenu2x) for faster Off button reaction (master) http://qi-hw.com/p/openwrt-packages/72b876700:02
Action: wpwrak blushes00:17
wpwrakzedstar: i can't quite grasp what it does, but it looks cool :) and yes, the music is good, too00:18
kristianpaulzedstar: actually i wonder how trusty is to measure vibration in your future aplication00:24
jow_laptopkyak: do you know whether gemenu2x writes a pid file?01:10
wolfspraulgood morning everybody01:36
wolfspraulDocScrutinizer: you there? remember you suggested supercaps once?02:27
wolfspraulinstead of batteries? I forgot the details... I'm looking a bit into rtc's upon larsc feedback that his m1 forgets the time whenever he turns it off02:28
wolfspraulso far I've found that there are thousands of discrete rtc chips at digikey, that we could probably implement an rtc in the Milkymist SoC itself, that we would probably want the cheapest and simplest chip that only does what the fpga itself cannot do (to track time when power is off)02:29
wolfsprauli2c and spi seem to be common interfaces, but I also found 1-serial and other serial interfaces02:30
wolfspraulthen I thought of your supercaps :-)02:30
wolfspraulbecause in any way power has to come from somewhere...02:30
wolfspraulwould rtc be a possible application? maybe I look for a tiny rtc chip that has a battery/supercap included? or is my thinking all wrong?02:31
kristianpaul(rtc chip) great idea !02:33
wolfspraulyes02:36
wolfspraulwe can implement the part with a lot of features in the SoC, somehow I think the only thing we need outside is a new tiny and cheap chip with some ability to hold power and track the time02:37
wolfspraulno interrupts, no high-precision stuff, nothing. all that could come from the SoC (theoretically)02:37
kristianpaulwell, keep the time is just enought :)02:38
kristianpaul(high-precision) well, werner already show  us that ntp can help a lot on this02:40
wolfspraulonce you assume network connectivity you don't need an rtc at all02:40
wolfspraulI'm looking for a very very simple fix. an unconnected m1 looses the time on every shutdown.02:41
kristianpaulyeap, i like that :)02:41
wolfspraulbecause we have an fpga, we don't need a sophisticated rtc chip with lots of bells and whistles (interrupts, alarms, high-precision, etc)02:41
wolfspraulso I want to understand this supercap thing more02:42
wolfspraulthat + a super simple time tracker (poll only) is all we need02:42
wolfspraulkristianpaul: you build milkymist software a lot, how computing intensive is it?02:45
wolfspraulI am thinking about upgrading the buildhost02:45
wolfspraulfor one, the nanonote builds keep getting slower (31 hours now)02:45
wolfsprauland I can only see us building more software and on a faster/more regular schedule in the future02:45
wolfsprauland then I want to start adding milkymist builds too02:45
wolfspraulincluding bitstream builds02:45
wolfspraulany thoughts?02:46
wolfspraulright now we have an AMD Athlon single-core machine. optimized well but fairly basic server power. I chose a cheap one for 29 EUR / month.02:46
kristianpaulxst, cant do multithread02:46
wolfspraulis there anything a buildhost could help you with? would you use it for something if it had certain power?02:46
kristianpaulas default..02:47
wolfspraulthat's ok, it's only one part02:47
wolfsprauleven if we go to a quad-core, it will constantly build enough stuff I think02:47
wolfspraulI also need to get this 31 hour thing down02:47
wolfspraulI'm curious about your Milkymist perspective02:47
wolfspraulanything a buildhost could do for you?02:47
wolfspraulhow long do the builds take?02:48
kristianpaul15minutes for my GPS porpuses02:48
kristianpaulas i use a minimac soc02:48
kristianpaulbuildhost not for me at least for bitstream.. unless it speedup sinthesis to 5minutes, that will be cool :)02:49
kristianpauland after i send a mail to automate build..02:49
kristianpaulhe,  i lile dream ;)02:49
kristianpauls/lile/like02:49
kristianpauls/minimac/minimal02:51
wolfspraulhmm02:53
wolfspraulok02:53
wolfspraulthat helps02:53
wolfspraulthere could be a web interface to kick off a build, or email interface for Werner, of course02:53
wolfspraulurl to source repo, hit the build button, done02:54
wolfspraula few minutes later the binary is available for download02:54
kristianpaulabout that 31minute building i wonder if you need alwats build all from scratch?02:54
wolfspraulit's not so much about need02:55
kristianpauli mean you change kernel and toolchain version in every build?02:55
wolfspraulit's about minimizing human working hours and maximizing machine power02:55
kristianpaulsure02:55
kristianpauland speed up02:55
wolfspraulyou can always compensate for less machine power with more manual hours spent, but who wants that02:55
wolfspraulI think we want it the other way round ;-)02:55
kristianpaul15minutes still... arggh :)02:56
wolfspraulwhen we started the buildhost, it wasn't clear whether it would be used at all02:56
wolfspraulbut now we have multiple people using it regularly02:56
wolfsprauland with Milkymist the need will only go up02:56
kristianpaulbut compared to 40minutos for a full soc build, well :-)02:56
wolfsprauland it's used for schhist and maybe brdhist one day02:57
rohwolfspraul: if you upgrade.. check for ram. that also speeds things up considerably (disk cache)02:57
roheq4 is a good deal02:57
kristianpaulon my computer.. a 2.6Ghz amd sempron 1gb. dunno numbers from lekernel ?02:57
wolfspraulyes I was thinking about eq402:57
Action: kristianpaul need to upgrade to eq4 some day...02:57
wolfspraulit's an x2 now, the cheapest physical hw I could get02:57
rohwolfspraul: i know ;)02:58
wolfspraulI would probably also setup the new one with Fedora02:58
kristianpaul:-D02:58
wolfspraulI tested Fedora a little and it looks good. I like their more speedy release schedule.02:59
wolfspraulselinux is causing hickups sometimes, but so far I've been able to solve them quickly02:59
kristianpaulkill it !02:59
kristianpaul:)02:59
wolfspraulhopefully that wouldn't cause trouble to buildhost users (selinux)02:59
wolfspraulwell, I like security02:59
wolfspraulI have no big pro/con about selinux in particular, just a user.02:59
kristianpaulinsecurity is among us ;-)02:59
kristianpaulsure02:59
wolfspraulroh: I was hoping for an eq4 price reduction or improved hw specs, but didn't come :-)03:00
wolfspraulI want more power, more ram, ssd, etc.03:01
wolfsprauleq is old now and hasn't been upgraded in a while. maybe they come out with something new soon...03:02
kristianpaulso for now i see no diference if build host speed and my computer to get sinthesize stuff03:03
kristianpauls/if/in03:03
rohwolfspraul: well.. that usually comes with 'new series' and the old ones get dumped out cheap.. as happened with the ds series03:03
wolfspraulye03:03
wolfspraulyes03:03
wolfspraulI know. So I was hoping for something new. eq is old.03:04
kristianpaulcan we test driver a build in that new server?03:04
rohwhat new should there be after i7 cpus?03:04
wolfsprauldon't understand03:04
kristianpaulanyway i guess you upgrade to quadr core03:04
kristianpauli mean before make the move03:04
wolfspraulno not really.03:05
wolfspraulbut you could get a server and cancel if after a month.03:05
wolfspraulthe amount of working hours you would put into this test would weigh far more than the 100 EUR or so you would still need to pay to hetzner in that case.03:05
kristianpaul100EUR?? nah... i still with my own build server :)03:06
wolfspraulno you asked about a test03:06
kristianpaulhelps have a laptop and mostly comand line tools for the rest of the tasks :)03:06
kristianpaulyeah, but better not ;)03:06
wolfspraulthe easiest way to test at hetzner is to just get the machine, do what you want, and if it's not good cancel after 1 month03:06
kristianpaul100eour is not a nice number to me :)03:07
wolfspraulsure, it costs money03:07
kristianpaulfot this test03:07
wolfspraulyes of course. but if we need a bigger and better buildhost, we'd just do it. upgrade from the current x2 (29 EUR/month) to something better...03:08
wolfspraulno test03:08
rohstill cheap.. i couldnt provide the hw including support and power as well as ethernet for that money. not even if i got the space and ip for free.03:08
wolfsprauljust upgrade and we know it'll be more powerful, from the specs03:08
wolfspraulsure you cannot, but hetzner has scale and is managed quite professionally03:08
rohack. just ressurected chandra for openmoko03:09
wolfspraulwow, Linux seems to have support for ca. 100 rtc chips :-)03:12
wolfsprauljust looking at drivers/rtc03:12
lemayavast03:22
wolfspraullemay: hey there03:22
kristianpaulwb lemay :)03:24
lemayI brought home 'Understanding GPS Principles and Applications' Edited by Elliot Kaplan03:25
lemayand GNSS Applications and Methods, by Gleason and Gebre-Egziabher03:26
kristianpaulthose books have topics related to software aprouch?03:32
lemaythey're textbooks with all the standard approaches03:33
DocScrutinizerwolfspraul: yes, I suggested supercaps for a number of good tested purposes and for some nonsense purposes (like battery swap buffering for phones)05:15
DocScrutinizerfor both GTA02 and RX51/N900 the supercap alternative for powering RTC has been implemented and I got reports it works flawlessly05:17
DocScrutinizerI strongly discourage chips with battery built in, afaik there's only been dallas chip and I frown at it05:18
wolfspraulok great, can you point to a supercap that could work to power an rtc?05:20
wolfspraulthe very first rtc I found is a Seiko S-35390A http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=728-1006-2-ND05:21
wolfspraul60 cents, existing driver in kernel.org05:21
wolfsprauldatasheet http://datasheet.sii-ic.com/en/real_time_clock/S35390A_E.pdf05:21
wolfspraulI had no idea that RTC chips go from 60 cents to 40 USD (!)05:22
wolfspraulamazing05:22
wolfspraulI was hoping to find something super simple that essentially just keeps track of one number on a tiny power budget05:22
wolfspraulbut I guess anybody who adds a discrete RTC to an embedded product will more likely choose one with a lot of features05:23
wolfspraulDocScrutinizer: do you have a name or URL for a supercap? or the solution that was done on N900?05:26
DocScrutinizersure05:26
DocScrutinizerhttp://www.mouser.com/catalog/632/2009.pdf  http://il.farnell.com/taiyo-yuden/pas414hr-va5r/cap-0-06f-3-3v-80ohm-4-8mm-coin/dp/1853000?Ntt=PAS414HR-VA5R05:27
DocScrutinizerI think the guys that modded N900 followed my suggestion for 2nd URL, the GTA02 reworks I dunno exactly which supercap-buttoncell got used for that, maybe the first URL bottom-of-page type05:29
wolfspraulgreat, thanks05:32
wolfspraulrtc chips seem to have all sorts of features like square wave output, watchdog timer, trickle charger. interesting.05:33
wolfspraulno idea what these things are good for :-) well. the watchdog timer makes sense.05:33
DocScrutinizernota bene the rationale for using this supercap buttoncels instead of proper LiIon cells comprised the fact that on a phone you usually _swap_ battery and thus need bupbat power for some couple of minutes only. Those caps can not power the RTC for months or years as the LiIon cells do05:33
wolfspraulneed to check how much power they can hold vs. lowest power rtc05:35
wolfspraulbut of course, understand05:36
wolfspraulDocScrutinizer: what is a square wave output or trickle charger in an rtc good for?05:36
wolfspraulreading about trickle charger now...05:36
DocScrutinizersquare wave output usually to clock other low power components like e.g LP5521 indicator LED flasher05:36
wolfspraulthe rtcs I find so far seem to want about 250-300 nA at 2V05:37
DocScrutinizerGSM modems also frequently have a 32kHz xtal  to keep the shedule during deepsleep (see #1024 where that clock generator was flawed by our layout)05:38
DocScrutinizer0.2..0.2uA - seens correct05:39
DocScrutinizerseems*05:39
wolfsprauloh I see, this rtc datasheet even mentions supercaps and charging them http://media.digikey.com/pdf/Data%20Sheets/Pericom%20PDFs/PT7C4302.pdf05:39
DocScrutinizer0.2 .. 0.305:39
wolfspraulthat's what they mean with 'trickle charging'05:39
DocScrutinizer\o/05:39
wolfsprauldoes that mean an rtc that does not support trickle charging is harder to connect to a supercap? I guess they added a little bit of circuitry inside this one...05:42
DocScrutinizerwell, afaiu the super cap is a working replacement for LiIon cells, and should handle all chips' chargers designed for LiIon. Chips that have no charger at all are designed for primary cell power source and I strongly suggest this variant whenever you can afford to have a holder for a CR2016 (or similar) LiIon primary cell on your PCB and housing allows access to that cell to swap it05:47
DocScrutinizerno option for neither freerunner nor N900 obviously05:48
DocScrutinizertime for 12648430    05:50
wolfspraulDocScrutinizer: ok, the taiyo yuden supercap has a capacitance of 60,000 uF06:13
wolfspraulhow long can that power an rtc needing 200nA@2V... calculating :-)06:13
DocScrutinizeryep, aka 60mAs/V06:13
DocScrutinizerVtop 2.8V, Vend 1.5V Vdelta 1.3V ergo 60*1.3 mAs capacity06:14
DocScrutinizer60 * 1.3 / 0.003(mA) = seconds such a cap can power your RTC06:16
wolfspraullooks like a year?06:16
wolfspraulok, I got 1000 wrong somewhere06:16
DocScrutinizer433 minutes?06:17
wolfspraulyes, about 7h06:18
wolfspraulok, interesting06:18
DocScrutinizerhonestly, if your housing ME and the PCB size (and audience/usecase) allows, better go for primary cells06:20
DocScrutinizeras virtually all PC mainboards do today06:20
wolfspraulprimary cells? what do you mean?06:21
DocScrutinizerbut, use, dispose06:21
DocScrutinizerbuy06:21
DocScrutinizernon-rechargable swappable cells06:21
wolfspraulyou mean like the round watch batteries CR2016 and such?06:27
wolfspraulhow does a desktop PC keep the rtc powered? or a notebook - how long can you take out the battery without loosing the rtc?06:28
rohwolfspraul: usually also lithium cells06:29
roheither solder-on or as connected pack with a short cable06:30
DocScrutinizerwolfspraul: (CR2016) yes, exactly06:30
rohthey are good for 5-10 years. and one can replace them after that. the one on my tp600e went empty at some point and i just replaced them by some new ones and used new shrinking tube after resoldering the cable06:31
DocScrutinizermoin roh! how's chandra business? OM community is suffering06:31
wolfspraulwhich cell was it? CR20xx ?06:31
rohsometimes its also 2 cells, but that was on older models06:31
rohDocScrutinizer: should be working again06:31
DocScrutinizerOOOH06:31
rohplease check/test06:31
wolfspraulso basically the supercaps are good for what? only a couple hours it seems06:31
DocScrutinizer\o/06:32
DocScrutinizerworks06:32
wolfsprauleven on a mobile device you probably better have 2 batteries, one to power the whole device, and a cr2016 or so for the rtc?06:32
DocScrutinizermail.om at least06:32
rohwolfspraul: depends on the power use. i guess they wont survive a month-long period without recharge while driving a rtc06:32
wolfspraulof course that's big. if space is a concern the supercap may be enough if people switch batteries immediately.06:32
wolfspraulroh: we just calculated it to be ca. 7h06:32
DocScrutinizerwolfspraul: not exactly a CR2016, but maybe a CR151006:32
wolfspraulok even smaller, nice06:33
rohwolfspraul: most simple mobiles (featurephones) dont have any batteries i know of. so either they survive on a cap or just dont have time after swapping battery06:33
rohi think my motorola survives a few minutes and then i have to reenter time manually06:33
DocScrutinizerusually the latter06:33
wolfspraulor they get the time from the network, I would think06:33
wolfspraulfor a connected device, if that's the only purpose of the rtc06:33
DocScrutinizeralso common06:33
rohmainboards sometimes even accept large coin style cells like cr203206:34
rohin a holder for easy replacement06:34
rohatx stuff06:34
DocScrutinizerusually CR2016 on mainboards of PC06:34
DocScrutinizererr, maybe I'm wrong06:34
rohwolfspraul: in germany afaik only one network transmits a time out of 4. and the quality (precision) is also not that good everywhere06:34
DocScrutinizercould also the 3.2mm 203206:35
rohsome networks get the utc vs local time wrong etc.06:35
rohDocScrutinizer: thats what i would use if i have the space06:35
wolfspraul2032 for what?06:35
wolfspraulfor the rtc sounds a little crazy, or you mean for the whole device?06:36
DocScrutinizeryeah, 2016 and 2032 most common and easily available nowadays06:36
rohcommon, available everywhere from many sources, enduser compatible replacement process, cheap, easy to use schematic wise06:36
rohwolfspraul: for the rtc/whatever needs to survive power off or battery change (depending on device)06:36
wolfspraulwow 2032 looks overkill for that06:36
DocScrutinizerfor 'small' devices there are other formfactors like CR1512 (15mm dia, 1.2mm thick) and others06:37
DocScrutinizerCR1208?06:37
rohwolfspraul: dont underestimate the factor of handling and sourcing on that. if its too small or to weird, endusers cant get it and have mechanical problems in handling when needed06:37
wolfspraulagreed but 2016 is very common06:37
wolfspraulhow long could that power a 200nA rtc?06:38
DocScrutinizeraeons06:38
DocScrutinizer;-)06:38
roh80-90mAh seems common on quality cells (checked sanyo and energizer specsheets)06:39
DocScrutinizerI'd guess a CR2032 has several 100s of mAh06:39
DocScrutinizerso 10k to 100k hours06:40
roh2032 is 240mAh06:41
DocScrutinizer240 / 0.00206:41
DocScrutinizererr 0.000206:41
roh13years06:42
wolfspraul13 years for 2032, probably a little less than half for 2016, let's say 5-6 years06:42
roh136 years at 0.000206:42
roh13 at 0.00206:42
rohfor a cr203206:42
DocScrutinizeraeons, toldya06:43
rohi bet self-discharge is a bigger influence on it running flat then06:43
DocScrutinizersure06:43
rohmeh. its late/early (8:45 am) .. need to head home catch some sleep06:45
DocScrutinizerwolfspraul: if you by any chance can avoid soldered battery and allow user to access it to swap, go for whatever size of CRxxyy primary cell. Otherwise go for supercap, as the soldered LiIon bupbat seem a broken technology06:45
DocScrutinizerroh: :-D have a pleasant rest, and thanks for fixing chandra06:45
wolfsprauloh sure, no worries06:50
DocScrutinizerwolfspraul: my Sony scoopman NT-1 had a CR1220 for RTC, and a very "expensive" door mechanism to access it to swap. Alas there must've been a flaw in hw, as that friggin cell been empty every 6 months06:50
wolfspraulI am not planning anything in particular anyway, just trying to understand the ballpark numbers for different technologies06:50
wolfspraulbut of course, soldered li-ion sounds wrong by design06:50
wolfspraulsupercap is fine, but has use-case limitations06:50
DocScrutinizeryup06:51
wolfspraulonce you are in the 20xx or 15xx area it's probably mostly about size and how common / easy to get something is, even after 5, 10 years06:51
wolfspraulall understood06:51
DocScrutinizersupercap fine for devices that most usually have a main battery, to survive swapping of that main battery only06:51
wolfspraulso a typical 2011 CR2032 cell can theoretically power a 200nA rtc for 136 years?06:51
wolfspraulnot 1306:52
wolfspraulof course self-discharge will be the much bigger factor06:52
wolfspraulis it 1% / day? must be less for those I would think06:52
DocScrutinizernah, more like 10% per year06:52
wolfspraulI thought I had read something about 1% / day somewhere06:52
wolfspraulah ok06:52
DocScrutinizer1% / day is for LiIon rechargable, not for primary06:53
wolfspraulprimary is which type (chemical)?06:53
DocScrutinizerprimary is use-once-and-discard06:54
wolfspraulalso lithium?06:54
DocScrutinizeryep06:54
wolfspraulwhat's the difference between 'primary' and li-ion?06:54
wolfspraulor are both li-ion and one is rechargeable and the other not?06:54
DocScrutinizerdifferent semantic domain06:54
DocScrutinizeryes06:54
DocScrutinizerCR2032 is a Li(Ion)-cell, but not rechargable (by design, though in fact they are ;-D)06:55
DocScrutinizerso it's a primary cell06:56
DocScrutinizera BP-5L Nokia battery is a rechargable LiIon (or LiPo) so it's not a primary06:56
DocScrutinizerprimary is the opposite of rechargable06:57
wolfspraulif it's the same chemical, why does a primary cell have a self-discharge of 10%/year, but a rechargeable 1%/day ?06:58
DocScrutinizerhttp://en.wikipedia.org/wiki/Primary_cell >>A primary cell is any kind of battery in which the electrochemical reaction is not reversible, rendering the cell non-rechargeable.<<06:58
DocScrutinizerbecause it's not exactly the same cell chemistry06:59
DocScrutinizerdifferent electrolyte, different separator06:59
DocScrutinizerdifferent mechanical (and also somewhat chemical) electrode design07:00
DocScrutinizerin a primary cell you for example don't need much auxiliary material to keep electrode mechanical shape, the whole electrode can be built by virtually massive Lithium07:01
DocScrutinizercharging such a cell won't result in an electrode of same shape as the original07:02
DocScrutinizerthis plus a lot of subtle diffs make a cell primary, which usually also results in better capacity on same formfactor than a rechargeable cell would offer07:04
ignatius-Anyone know how to access/read a broken (CRC, UBI errors) Debian Nanonote system? It was working, but something happened, not sure what, and now it won't boot.07:04
ignatius-Something that doesn't let the system boot the kernel, without a kernel panic. I've tried the "root=" append line, no luck.07:06
wolfspraulDocScrutinizer: understand, thanks!07:06
DocScrutinizer>>Primary batteries are useful where long periods of storage are required; a primary battery can be constructed to have a lower self-discharge rate than a rechargeable battery, so all its capacity is available for useful purposes. Applications that require a small current for a long time, for example a smoke detector, also use primary batteries since the self-discharge current of a rechargeable battery would exceed the load current ...07:07
DocScrutinizer<<07:07
ignatius-Is it possible to access the Nano through the USB connection?07:09
wolfspraulwe had some plans but I don't think they are fully implemented yet07:09
ignatius-Ok.07:09
wolfspraulsuch as loading a rescue system via usbboot directly into memory07:09
wolfspraulyou can try to boot from the memory card though07:09
wolfspraulhow about that?07:10
ignatius-Not sure on how to do that.07:10
wolfspraulas long as u-boot still loads that should work07:10
ignatius-I tried to "HOWTO" on the official website, it needed me to be on the machine while creating the bootable card.07:11
ignatius-http://en.qi-hardware.com/wiki/Updating_Ben_with_an_SD07:11
DocScrutinizerI'm sure there's an easy way to create the card on arbitrary PCs07:12
DocScrutinizercreating the card on Ben just described as a convenience thing07:13
ignatius-Nod.07:13
DocScrutinizer(no Ben here, so I talk outa my a...)07:13
wolfspraulignatius-: that page is about updating the Ben from a memory card, there's another one about booting, but I cannot find it either :-)07:15
ignatius-:/07:17
kyakjow_laptop: nope, gmenu2x doesn't write a pid file. Probably it used to, when we started it via start-stop-daemon. But now it just spawns from inittab08:31
wolfspraulkyak: how do you feel about upgrading the buildhost to a more powerful machine? is it worth it?08:37
kyakwolfspraul: yeah, why not? Upgrading is always worth it. For now, the build for Ben takes more than 24 hours. It would be great if we could "fit" it into 24 hours08:45
kyakfrom the other hand side, if we look at the rate of commits to openwrt-xburst.git and openwrt-packages.git, it is longer than 24 hours08:46
kyakso basically consequitive images are very often the same08:47
kyakin this thinking, updating to a more powerful machine doesn't make sense08:47
kyakat the end, it also depends on the price you are going to pay for upgrade :)08:52
wolfspraulkyak: he. interesting point about commit intervals :-)09:14
wolfspraulbut I also want to make the build experience snappier and more fun for people who use the buildhost actively09:14
wolfsprauland if you commit something specifically, you may have to wait 2 days until you can try: not good09:14
wolfspraulI'll think about it. the monthly price would go from 29 to 49 EUR. Maybe I wait for a new server line to be introduced by the hoster.09:15
kyakwolfspraul: good :)09:33
kyakwolfspraul: the full boot is speed up :)09:41
kyaki press the button once, and the thing just powers on at the exact same moment09:41
kyakgoes through the whole boot sequence09:42
kyakyou could try it yourself, since poke is avaivable for quite some time now09:42
qi-bot[commit] kyak: mc: override the upstream version to include wide char support (master) http://qi-hw.com/p/openwrt-packages/7f56d7c09:51
DocScrutinizerwolfspraul: would it be feasible to cache some pieces, so a build wouldn't need to run 24h+ to build same cruft over and over again? Sounds kinda silly10:34
DocScrutinizerI mean even a complete kernel build for maemo takes <3h on my 1.8GHz dual core pentium laptop10:35
kyakDocScrutinizer: the point of build is to build from scratch10:35
kyakincremental builds are not interesting10:36
DocScrutinizerI'd not expect a buildhost to build a "new" (in fact same old) kernel, when all I did was correcting a typo in ls --help10:36
DocScrutinizernevermind, I got not the faintest clue what you guys are doing there on your buildhosts10:37
kyakit had been proved several times that change in one place can lead to strange results in other places. And using "cached" results leads to unpredictable things10:38
DocScrutinizerhmmm o.O10:38
kyakyeah, just search for "build from scratch" in irc logs10:39
DocScrutinizerI'm out, not intending to bug you with noobish musing10:39
DocScrutinizerI easily can agree on building core system in one run is a clean easy to understand thing10:41
DocScrutinizerstill 24+ hours? WTF!10:41
DocScrutinizerwhatever, I'm out again10:42
wolfspraulkyak: poke? you mean a tool I can use to write that rtc register manually?10:43
wolfspraulare we setting it automatically in the latest builds?10:43
kyakwolfspraul: yeah, that tool. THis is not yet done automatically..10:55
kyaki suggested xiangfu to include it in start script10:55
wolfspraulhmm, ok :-)11:06
wolfspraulamazing that we haven't included something so nice right away11:06
wolfspraulthe image release cycles are too long11:06
dvdkdo we need poke?  I guess dd of=/dev/mem bs=1 count=1 seek=0xblabla would work, too :)11:10
dvdkwait, no, with I/O mem it won't.  But in gforth it can be scripted.11:11
dvdkwell gotta go11:11
ignatius-Anyone know how to make a bootable SD card?11:51
ignatius-After installing the boot loader and the userland on the SD card, upon booting, I get a garbled screen, that says somehting like "In: serial / Out: lcd / Err: lcd / Hit any key t  xc"12:04
wolfspraulwhere did you get the binaries you put on the memory card from?12:07
ignatius-http://jlime.com/downloads/releases/muffinman/12:08
ignatius-(Couldn't find instructions on the main qi site)12:08
wolfspraulkyak: do you know where the instructions for booting from the memory card are?12:09
ignatius-Thanks for answering. I appreciate it.12:10
ignatius-I've been having problem after problem... all I wanted at first was a custom kernel, with certain loadable modules. And this crap happens.12:10
wolfspraulwell custom kernel is quite something to start with12:11
ignatius-Compiling kernel after kernel.12:11
ignatius-Yeah. True.12:11
ignatius-I've found that the new "qi kernel" (vanilla?) supports the 2GiB NAND. I'm guessing the OpenWRT team or whoever patches it to support less.12:12
ignatius-So, i've been going with that kernel. Still have to get the KS7010 driver working, too.12:13
wolfsprauloh, wow12:13
wolfspraulthat's an ambitious goal12:14
ignatius-I had it all working at once. Then it got screwed up somehow.12:14
ignatius-I (stupidly) compiled the kernel, and it did everything I wanted it to do, then, I think I must've deleted that source tree, and have been looking for it for a few weeks now.12:14
wolfspraulhmm. yeah. painful.12:15
ignatius-My goal was to have a smaller startup font. :/12:15
wolfspraulunfortunately there is so much activity in the kernel that I cannot really follow. I just hope the images coming out of the automated builds work :-)12:15
ignatius-Nod.12:16
ignatius-I think the Ben Nanonote should be a lot more popular. Even amongst the Dingoo A320 crowd.12:16
ignatius-It's really an awesome little gadget.12:17
wolfsprauland you say this after weeks of struggling :-)12:17
ignatius-Heh. Yeah.12:17
ignatius-I can't wait for the successor.12:17
ignatius-(If there will be one)12:17
ignatius-But, yeah. My goal is to get to where I was before. An ultra small WIFI device that can do more than a smart phone.12:19
kyakwolfspraul: not sure where the instructions are. But this is pretty simple12:34
kyakhttp://en.qi-hardware.com/wiki/OpenWrt_Software_Image#Image_2011-02-2312:34
kyakthis is described in this changelog how to boot from sdcard partitions12:35
kyak"fw_printenv" on Ben itself is also pretty explanatory on what are the available options12:36
kyakhttp://dpaste.com/598987/12:36
ignatius-Thanks.12:36
wolfspraulkyak: wow, thanks. believe it or not I couldn't find it earlier :-)12:38
kyaknp :)12:38
kyakignatius-: regarding the smaller font. Do you really need it at boot time? Wouldn't it be better to use setfont later?12:41
ignatius-I just like to be able to see how things are while booting.12:43
ignatius-Init scripts and whatnot.12:43
ignatius-Upon F4-PowerON "Wrong Image Format for bootm command"12:44
ignatius-F1-PowerON = same error as before. Garbled screen.12:45
kyakwhat is your sd card configuration?12:45
kyaki mean, partitions layout12:45
ignatius- /dev/sdc1 = ext2, /dev/sdc2 = swap12:46
kyakand what rootfs have you put of sd card? and where have you put the uImage?12:46
ignatius- /dev/sdc is a 1GiB MicroSD card.12:46
ignatius-The MuffinMan rootfs, and the uImage is in /boot12:46
kyakdo you have the latest bootloader?12:47
ignatius-AFAIK.12:47
kyakthis is vital important12:47
ignatius-Where can I get it?12:47
kyakhm.12:47
kyakhttp://downloads.qi-hardware.com/software/images/NanoNote/Ben/latest/openwrt-xburst-qi_lb60-u-boot.bin12:47
kyak(obviously?)12:47
ignatius-Flashing...12:48
kyakfor your layout, it seems fine12:48
kyakF1 should work for you12:48
kyakwhat is your version of xburst-tools btw?12:49
kyakthere was an intermediate problem with one of the versions12:49
kyakpreventing from flashing bootloader as expected12:49
ignatius-Same exact errors as before.12:54
kyakcan you paste the output of you flashing the bootloader?12:55
ignatius-usbboot 20110412:55
ignatius-Ok. Hold on a sec.12:56
kyakheh12:56
kyakthat is probably the "broken" version12:56
ignatius-:/12:56
kyaktry updating xburst-tools to the latest one12:56
ignatius-Ok.12:56
kyakhttp://lists.en.qi-hardware.com/pipermail/discussion/2011-March/007456.html12:57
kristianpaulhum, dd defeating poke? :)13:06
kristianpaulkyak: was is the porpuse of OpenWrt Image Builder?  i touhgt it once have a basis systems compiled allowed to crosscompile other packages with no need to recompile WHOLE system again..13:07
kyakthis is the commit that fixed the problem with usbboot: http://projects.qi-hardware.com/index.php/p/xburst-tools/source/commit/1d7a2f3f5610518f67bf789401764a2628acf02f/13:07
kyakkristianpaul: huh? you can do incremental builds if this is what you are asking. THe purpose of buildhost, however, is to run nice and clean build from scratch13:08
kristianpaulyes,  incremental builds, i like how it sounds :)13:09
kristianpaulyes i undertand that buidlhost is for that.. jsut wondering if its needed all time, i already guess it is13:10
kristianpaulas i tought makefiles allowed to recompiled what just changed to no more..13:10
kyaksure. This is what you do at home :)13:11
kyakbut not on buildhost, from my understanding13:11
wolfspraulbut anybody can use their user account on buildhost for whatever compilation they want13:13
kyakyeah, i was just talking about the automatic builds running via cron13:16
kristianpaulwpwrak: can you http/ping 216.239.32.2 ?13:22
kristianpaulhum nv13:23
kristianpaulwerner will need a openvpnv to other country soon ;) http://googleamericalatinablog.blogspot.com/2011/08/blogs-bloqueados-en-argentina.html13:24
ignatius-Dangit. :/ config.status: error: cannot find input file: `Makefile.in'13:24
ignatius-Is there an up-to-date XTools RPM somewhere?13:25
kyakignatius-: try ./autogen.sh13:28
ignatius-I did. It created "configure" then configure says "error: cannot find input file: `Makefile.in'"13:29
kyakthere is a deb here http://projects.qi-hardware.com/index.php/p/xburst-tools/downloads/13:30
kyakyou could try autoreconf, too13:32
ignatius-Apparently, that is the same exact version I had installed beforehand.13:34
kyakcan't be13:34
kyakyou told you had 20110413:34
kyakbut you need a version later than 27th of May13:35
ignatius-Well, that's the version it says when I do a "xbboot -v" 13:35
ignatius-I even tried booting the Nanonote with F1... same exact errors as before.13:36
kyakhave you read the messages on mailing list i pointed to? Do you have the same errors?13:37
ignatius-Yes.13:37
kyakand how do you know it is exactly the same version that you had before?13:38
ignatius-Because I did a "xbboot -v" then, too.13:38
kyakthis is nothing :)13:38
kyakit shows 201104 to me, too13:38
ignatius-Hmm.13:38
ignatius-Why the errors, than?13:39
kyakanyway, now that you have the version from 2011-05-30, try flashing the latest bootloader13:39
ignatius-I did.13:39
ignatius-I'll try again, though.13:39
kyakbtw. Where is the uImage you are using coming from?13:39
kyakyou could use the latest uImage form openwrt and see how it boots13:40
ignatius-The MuffinMan site.13:40
kyakeven with jlime rootfs, this should work13:40
kristianpaulweee !13:41
ignatius-Same errors as before.13:41
ignatius-Ok. I'll try an OpenWRT image.13:41
kyakignatius-: could you paste the output when flashing the bootloader?13:41
ignatius-Sure.13:42
kristianpaul2 second less of booting :)13:42
kyakignatius-: also, are you sure you have this partition on sd card as the first partition? Not the whole sd card formatted as ext2? In the latter case, you should hold the "M" while powering on13:44
ignatius-usbboot -c "nprog 0 openwrt-xburst-qi_lb60-u-boot.bin 0 0 -n"13:48
ignatius-usbboot version 201104 - Ingenic XBurst USB Boot Utility (c) 2009 Ingenic Semiconductor Inc., Qi Hardware Inc., Xiangfu Liu, Marek Lindner This program is Free Software and comes with ABSOLUTELY NO WARRANTY.  Now checking whether all configure args valid: YES Current device setup information: Crystal work at 12MHz, the CCLK up to 252MHz and PMH_CLK up to 84MHz SDRAM Total size is 32 MB, work in 4 bank and 16 bit mode Nand page pe13:49
ignatius-Ack.13:49
ignatius-Hopefully that looks better to you.13:49
ignatius-Execute command: nprog 0 openwrt-xburst-qi_lb60-u-boot.bin 0 0 -n   Programing No.0 device, flen 623096, start page 0...  CPU data: Boot4740  Erasing No.0 device No.0 flash (start_blk 0 blk_num 2)......  Finish! Return: 00 01 00 00 00 00 00 00 (position 2)  Force erase, no bad block infomation!  Size to send 623096, transfer_size 524288  Image type : without oob  It will cause 2 times buffer transfer.  Writing NAND page 0 len 52413:51
kyakignatius-: your messages are truncated.. better paste somewhere13:51
ignatius-Nod.13:51
ignatius-Hold on.13:51
ignatius-Pfft. Of course. My shell account is down.13:53
ignatius-Damnit.13:53
kyakignatius-: dpaste.com :)13:53
ignatius-Ah. Ok. Cool.13:53
ignatius-http://dpaste.com/599039/13:55
ignatius-Damn. It's already 7AM.13:56
kyakignatius-: yeah, the log looks good13:57
kyakand 6PM here, heh13:59
wolfspraulI'm calling it a day, ignatius- wish you good luck! eventually something must work!14:01
ignatius-Thanks, man.14:02
wolfspraulwell, thanks to kyak for having him and helping14:02
ignatius-Is there a CVS, GIT, SVN version of the Xtools?14:02
ignatius-Yeah, thanks go to kyak, too.14:03
kyakhttp://projects.qi-hardware.com/index.php/p/xburst-tools/14:03
wolfspraulhe was faster :-)14:03
kyaki thought you tried the latest git when you said about missing Makefile.in :)14:03
ignatius-Yeah. True. I'm tired.14:04
ignatius-Is the "master" branch the latest?14:04
kyakyes14:05
ignatius-Ok... if I can figure out how to get it to compile..14:05
ignatius-#define PACKAGE_NAME "xburst-tools" #define PACKAGE_TARNAME "xburst-tools" #define PACKAGE_VERSION "201104" #define PACKAGE_STRING "xburst-tools 201104"14:10
ignatius-:/14:10
ignatius-I was able to "touch" Makefile.in... seems to be working.14:13
kyakstrange, perhaps xiangfu can be of a more help regarding xburst-tools compilation issues14:14
ignatius-Same errors. Again. :(14:19
kristianpaulkyak: can you confirm this poke 0x10003024 0 is persistant after at least 5 minutes?14:23
kristianpaul5 minutes after power-off14:25
ignatius-Well, time to call it a day. Thank you for all of your help. Hopefully i'll get this problem solved.14:28
kyakkristianpaul: at least 20 minutes have passed with Ben being switched off. The On button is still fast15:16
kristianpaulhum...15:40
kristianpaulwell, lets see more people to try..15:41
kristianpaulkyak: you added the poke to the system init?15:41
kristianpaulcause i asumed it should work until battery is removed15:41
mthkyak: you wrote something about the on button being fast...22:32
mthdid you check commit 81a776e43a349a9695f7bc9d51964276934fe099 in qi-kernel?22:32
mthit sets the timing for waking up from hibernation22:32
mthit's included in the jz-3.0 kernel branch22:33
rohre22:38
rohwolfspraul: you there? any ideas which hw to buy to get hands on other ingenic stuff?22:39
rohi found this http://www.androidtablets.net/forum/android-tablet-news-depot/12629-ronzi-a3-299rmb-android-2-2-4-3-pmp-wifi-cheap-they-come.html via the ml, but it seems twice as expensive over here as in china22:39
roh300rmb seems half of the 50Euro it costs me22:39
wolfspraulroh: no not really. quality will be hit and miss, both in the product/model and also the individual one you end up having.23:31
wolfspraulif something costs 300 RMB in China then 50 EUR in germany sounds quite fair to me, considering all the work necessary.23:32
rohtrue23:33
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08192011-1943/23:34
rohwolfspraul: what do you think about the future if ingenic for us?23:38
rohi think if we ever want to continue the way we should get the newer soc working on existing hw first before building our own. makes testing a bit easier maybe23:38
wolfspraulroh: just a hardware supplier to me. I like how successful we are with the Linux kernel, mostly Lars and mth and a few others. So that's a real treasure I think that we need to maintain well.23:50
wolfspraulif I would consider using any other Ingenic chip, I would look at the state of support in that kernel first23:50
wolfspraulwhen you say "get the newer soc working on existing hw first" that sounds like an anti-vendor port23:51
wolfspraulI don't believe in anti-vendor ports. I will continue to work on opening up the hardware we know best now, Ben NanoNote, make it easier to produce, simplify and open mechanical, etc.23:52
rohwell.. there seems to be some work happening already.. i think its more 'learning to know and use a soc' than 'doing a anti-vendor port' (without their support)23:53
wolfspraulanti-vendor ports are lacking a foundation for the future, I rather work on the foundation23:53
rohisnt the ingenic documentation open for the new chips?23:53
wolfspraulsure that's good. lars has been doing 4760 related stuff in the kernel for over a year I think.23:53
wolfspraulnot open23:53
wolfspraulor maybe Chinese way - under the table open23:54
rohwell.. then it should be time to open it ;)23:55
rohwhen we now can get hw on the open market with that chip23:55
--- Sun Aug 21 201100:00

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