#qi-hardware IRC log for Thursday, 2011-07-07

qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-07052011-2120/01:02
wpwrakwolfspraul: did you see the discussion here ? http://www.heise.de/meldung/Bericht-Microsoft-bittet-Samsung-fuer-Android-zur-Kasse-1274692.html01:34
wpwrakwolfspraul: nobody thinks free software is dead ;-) they all see pretty clearly where the problem is01:34
wolfspraulheise is not read by newcomers01:35
wpwrakduh. Bild then ? ;-)01:35
wolfspraulhmm, read it01:43
wolfspraulI'm surprised that you can still read the heise comments. I just click on a few, see all the screaming, trolling, bad jokes etc. and stop01:49
wolfspraulnot 1 original thought in 100 of them, or so01:49
wpwrakoh, i usually just pick a few. to see if there's anything interesting01:58
wpwrakwolfspraul: by the way, i relived a bit of an openmoko experience today. tried to build the cross-toolchain of openwrt, to pick up some new features.03:01
wpwrakwolfspraul: the experience reminded me very much of OE at openmoko ...03:01
wolfspraulwhat feature were you missing?03:03
wolfspraulwe are trying to release prebuilt toolchains, but I guess it didn't help (I saw the chat)03:04
wpwrakwolfspraul: so i'm afraid the core of the problem is still the complexity of a "desktop" distribution. you can try to escape it by starting with something much simpler (like you did with openwrt), but once you bring it up to speed, it'll be difficult, too. it's not quite as bad as OE, where your only choice was to ask the priests of the inner circle for some magic modifications/commands, but still not user-friendly03:04
wpwrak(feature) SDL_gfx03:04
wpwrak(chat) ah, good. then i don't need to repeat it all :)03:04
kristianpaul how ironic trying to fix my camera shuter when not having another camera to take pics of the disasembling process~ :-/03:05
wpwrakthe best approach would be to have an environment with a minimum toolchain to get started plus the ability to install packages into that environment03:06
wpwrakkristianpaul: ;-)))03:06
wpwrakkristianpaul: you could remember the tradition of biologists in ancient times and make drawings :)03:06
wpwrakwolfspraul: even the "default" minus "modules" (not quite sure what they are) is fairly heavy. a minimum core with gcc and libc should be much smaller.03:07
kristianpauli think i will later this month, for now no camera to carry :(03:08
wpwrakif you fail at pygames for the "base", you know something is a little off :)03:08
kristianpaultalking about games i think zear_ is in the same path of wpwrak, finding a non problematic way of porting games to the nn03:27
wpwrakatrf-rssi isn't exactly a game, but ... ;-)03:28
wpwrakyou could use atrf-path as a game, though :)03:28
kristianpaulwell, no that wpwrak will port games but youget the idea03:29
wpwrakif i had brought a usb hub (or a laptop with more than 1 host port) to fisl, we could have had some fun with atrf-path :)03:29
wpwrakkristianpaul: truth be told, what i find most worrying about today's experience is that i realized that people either have to use some fairly old toolchain, which was easy to build and works, or they should either have had some questions or be very independent.03:31
kristianpaulyeah, i agree, it was easy to build..03:33
kristianpaulshame i deleted that folder.. when trying to get latest stuff03:34
wpwrak;-)03:34
wpwrakfirst lesson: rename, don't delete :)03:34
kristianpaultoo much renames, more when the thing is about 10GB big !03:34
kristianpaulbut anyway, i think wiki could be improved easilly, just with the well experience from xiangfu and kyak for example :)03:35
wpwraki'm now back tp the configuration of 9 months ago. which - oddly enough - seems to be more modern than the one from ~1.5 months ago03:35
wpwraki think the wiki is up to date now03:35
kristianpaulgood !03:35
wpwraknot sure about the .config03:35
wpwrakbut there's also the quirk that uClibc regresses03:36
kristianpauljust use minimal, and grown from that03:36
kristianpaulwell, may be you can find something more stable from openwrt upstream for backfire03:37
kristianpaulbut of course no the same feeds will be there..03:37
wpwrakit seems that this is what that ABI breakage is about03:37
wpwraki didn't quite realize there was an ABI barrier until today03:37
kristianpaulhum..03:38
wpwrakABI incompatibilities are very very bad. means that you can't easily try and then decide whether to proceed or revert03:39
wpwrakat least not when changing one thing at a time03:39
wpwrakso this sort of thing should be headline news, along with a migration plan03:39
wpwrakcould be that i just overlooked it, though03:40
kyakwpwrak: yeah, it make sense - combination of toolchain-mipsel_gcc-4.3.3+cs_uClibc-0.9.30.1 is the correct one06:24
kyakwpwrak: so how is sdl-gfx doing? You should have it in your toolchain now06:27
kyakwpwrak: perhaps at some point earlier you switched over to uClibc 0.9.32, but this is not right. Backfire uses 0.9.30.1, trunk uses 0.9.32 and it breaks binary compatibility06:28
kyakwpwrak: i suggest that you also flash the latest release image to your Ben, it could be that you have one of the images when we switched over to 0.9.32 shortly, but then went back to 0.9.30.106:30
wpwrakkyak: (sdl-gfx) surprisingly, it doesn't seem to be there10:57
wpwrakkyak: (backfire vs. trunk) what's the plan for the future anyway ? do you intend to switch one day from backfire ?10:58
wpwrakkyak: i quite dislike the strtof regression in backfire. not being able to compile perfectly posix-compliant code without a warning isn't nice.11:03
kyakwpwrak: the builds in buildhost are already based on trunk. 05-28 was the last backfire-based release11:07
kyakand sdl-gfx has to be there :)11:08
kyakgrep libsdl-gfx .config11:08
kyakshould be =y11:08
kyakand sdl-config and sdl libraries and headers should all be installed11:09
wpwrakCONFIG_PACKAGE_libsdl-gfx=y11:09
kyakls staging_dir/target-mipsel_uClibc-0.9.32/usr/lib/libSDL_gfx.*11:10
kyakshould be your target-mipsel_uClibc..11:10
wpwrakNo such file or directory11:10
wpwraktarget is target-mipsel_uClibc-0.9.30.1 anyway11:11
kyakyeah11:11
kyakthis is strange11:11
wpwrakah yes, the library is there ...11:11
wpwrakah, includes too :)11:11
wpwraki looked under toolchain-mipsel_gcc-4.3.3+cs_uClibc-0.9.30.111:11
kyakit's target11:12
wpwraklet's see what this changes ....11:12
kyakanyway, locate the sdl-config11:12
kyakit will give you the right path11:12
kyaksdl-config is somewhere here: ./staging_dir/target-mipsel_uClibc-0.9.32/host/bin/sdl-config11:13
kyak$ ./staging_dir/target-mipsel_uClibc-0.9.32/host/bin/sdl-config --libs11:13
kyak-L/home/bas/build/openwrt-xburst/staging_dir/target-mipsel_uClibc-0.9.32/usr//lib -lSDL -lpthread11:13
wpwrakgrmbl. no cross-toolchain under target. that's where the problem comes from11:14
kyakthe toolchain is under the toolchain*11:14
wpwrakdoes gcc have target in its include search list ?11:14
wpwrakif you split toolchain and target (and don't add target to gcc in some other way) all the automatic searches fail11:15
wpwrakthat's why you need hacks like STAGING_DIR11:15
wpwraknow it all begins to make sense ;-)11:15
kyakit works fine from within openwrt build system11:15
kyakbecause all the necessary variables are passed to configure/make11:16
wpwrakyes, because you override all the search paths :)11:16
wpwrakthat's an exquisitely inconvenient way to do cross-compilations. instead of just having to rename your compiler, you need to pass the directory structure into the makefiles11:17
kyaki'm sure there is a perfect explanation for that. I'm just not an expert regarding openwrt build system11:18
wpwrakwhere's xMff when we need him ? :)11:18
kyakand again, you don't need to pass anything to makefile in most cases11:19
kyakwell designed applications a ported like that11:19
wpwrakbut i need to pass -L$(TARGET)/usr/lib and -I$(TARGET)/usr/include, right ?11:20
xMff$(STAGING_DIR)/usr/...11:20
wpwrakat least if i'm compiling outside the openwrt build system11:20
wpwrakso STAGING_DIR = ..../target..., okay11:20
xMffyes11:21
wpwrakxMff: why are you making this so complex ? :)11:21
wpwrakxMff: if you had everything in the same hierarchy, gcc would automatically pick the right includes and libraries11:21
xMffthats wichful thinking11:22
xMff*wishful11:22
wpwrakxMff: works great on jlime ;-)11:22
xMffits already hard enough to get autofail to not act up11:22
xMffwell whenever I hear from Oe its people having troubles compiling :)11:22
wpwrak(autocrap) well, that's certainly true :)11:22
wpwrak(oe) i think the secret is to let others do the compiling for you :)11:23
xMffbbl work11:23
wpwrake.g., i've never ever compiled an entire redhat, debian, or ubuntu from sources. maybe a source package every few years. why should we have to work so much harder in the embedded world ? :)11:24
Action: wpwrak wonders if one couldn't just symlink the target tree into the toolchain tree. there are a few duplicate files that need looking at11:30
jow_laptopbecause most software simply sucks at cross compiling11:50
jow_laptoplast time I checked not even gcc could compile itself without patches11:50
jow_laptop*cross-compile itself11:51
jow_laptopperl does not cross compile11:54
jow_laptopruby does not cross compile11:54
jow_laptopmost libtool deployments do not cross compile either11:54
wpwraklibtool is a losing proposition anyway :)11:54
jow_laptopmany autotools based packages fail in configure due to AC_TRY_RUN checks11:54
jow_laptopautoreconf is not possible because automake broke its own api11:55
wpwrakit's amazing how creative people are inventing solutions that just break things in more obscure ways11:55
wpwrakautocrap is another set of problems, yes11:55
wpwraki was more thinking of sane developers who stay away from all such mess11:55
jow_laptopand in openwrt we split toolchain and target because we support multiple targets with the same toolchain11:56
jow_laptopmunching that together breaks that completely11:56
wpwrakoh. that makes merging hard :-(11:56
kyaki knew there was a reason for splitting it! :)11:57
jow_laptopuntil recently we had to support gcc 3.x etc.11:58
jow_laptopnow that this is gone we can maybe thing about --sysroot11:58
jow_laptop*think11:58
wpwrakthat sounds promising12:00
wpwrakkyak: regarding backfire vs. trunk, what's the migration plan ? and what bnary compatibility does exactly break ? the kernel ABI ? the libc ABI ? a set of things scattered all over the place ?12:02
wpwrak(multiple "yes" answers are possible ;-)12:02
stefan_schmidthmm, yes, no, yes, yes?12:03
Action: stefan_schmidt play blind answering12:03
wpwrakstefan_schmidt: the good new: i had dirtpan send pings around. the bad news: RTT is horrible. dirtpan times out all the time and retries something like 10-20 times for each packet.12:03
stefan_schmidtwpwrak: ups12:04
wpwrakstefan_schmidt: (1011) not a bad choice :)12:04
wpwrakstefan_schmidt: i think if i relaxed the timeouts in dirtpan it would be happier. but the root cause is the 10 ms interrupt synchronization delay12:04
stefan_schmidtwpwrak: int missing or something else12:04
wpwrakstefan_schmidt: naw, i think everything works as planned12:05
wpwrakstefan_schmidt: just the plan needs some improvements :)12:05
stefan_schmidtwpwrak: ok, so it is the delay12:05
wpwraki'm now trying a different approach for synchronizing that shouldn't need the delay12:05
stefan_schmidtwpwrak: I was just going to prepare my little lowpan-perf tool for a run12:06
stefan_schmidtwpwrak: (w/o delay) cool12:06
wpwrak(perf) good. then we can compare before and after :)12:07
wolfspraulhas anybody found out about atben/atusb sales numbers now?12:15
wolfspraulmaybe we have to do without them...12:16
wpwrakwolfspraul: i haven't made it that far in my telepathy course yet ...12:16
jow_laptopwpwrak: the libc abi will break12:16
kyakwpwrak: the uClibc binary compatibility is broken between backfire and trunk.. For the migration plan, i think xiangfu wanted to make a release couple of week after the next openwrt release (not sure when it would happen now though). Anyway, the trunk is already stable enough and all development is carried out in trunk only12:16
jow_laptopalso the threading system changes12:17
jow_laptopfrom linuxthreads.old to ntpl12:17
kyakyeah, the nptl thing12:17
wpwrakokay, so host toolchain, libs, root image, extra packages, and local developments are affected12:18
jow_laptopyep12:19
wpwrakwill the names of the tools change as well ? (away from mipsel-openwrt-linux-uclibc-*)12:19
kyaknah, these names remain the same12:20
wpwrakgood. so just a re-make will do for posix things12:23
wpwrakpity openwrt doesn't use eglibc. otherwise, we could even have binary compatibility with jlime12:25
jow_laptopit allows to use eglibc12:26
wpwrakoh, great ! what's the experience with it ? good ?12:27
jow_laptopno12:28
jow_laptopyou trade one bag of annying bugs against another one :)12:28
wpwrakwhat are the problems ?12:28
wpwrak;-))12:28
jow_laptopwell dunno, one has to try it again12:29
jow_laptopits been a while since it was added and only recently some work was invested in it12:29
jow_laptop... again12:29
wpwrakfunny. atusb is making a little bit of noise when transmitting :)12:40
wpwrakstefan_schmidt: things are still slowish. grmbl.12:41
stefan_schmidtwpwrak: hmm12:45
wpwrak362 packets transmitted, 361 received, +1055 duplicates, 0% packet loss12:53
wpwrakit's a feature: enhanced redundancy :)12:54
wpwrakheh. and i found another bug in at86rf230.c ;-)12:58
wpwrakhmm, a good marketing slogan for advertizing 64 vs. 32 bits just occurred to me: "pointer enlargement"13:20
stefan_schmidtlol13:20
stefan_schmidtwpwrak: have the same feeling you had with your ubuntu today with my debian13:22
wpwrakbeen a while since the last update ? :)13:22
stefan_schmidtwpwrak: the kvm install I have had not a signle devel package installed. no pck-config, bison or whatever13:22
stefan_schmidtinteresting errors to figure out what is missing13:23
wpwrakyuck :)13:23
stefan_schmidtfinally got lowpan-tools with my measurement tool compiled13:23
qi-bot[commit] Werner Almesberger: at86rf230: we may be in BUSY_RX after commanding RX_ON (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/e21c6ed13:26
stefan_wpwrak: Received 10 from 10 packets13:28
stefan_Arithmetic mean rountrip time: 0.000000 seconds and 193495.203125 usecs13:28
stefan_wpwrak: thats roundtrip on lowpan level13:28
stefan_no re-transmit and other fancy stuff build in yet :)13:29
stefan_one meter distance directly next to an wifi AP :)13:29
stefan_wpwrak: 100 also without problem, going 1000 now13:30
wpwrakinteresting. you really have picosecond resolution ? ;-)13:30
wpwrak~200 ms. damn slow. lemme finish debugging my improved delay system13:31
stefan_wpwrak: nope, standard timeval, need to clean it up13:32
stefan_wpwrak: what wonders me is how reliable it is13:32
wpwrakprintf("%lu.%06lu", (unsigned long) tv_sec, (unsigned long) tv_usec); ? :)13:33
wpwrakyou can try a few times and see if the value is stable13:33
stefan_Received 1000 from 1000 packets13:34
stefan_Arithmetic mean rountrip time: 0.000000 seconds and 177617.531250 usecs13:34
stefan_Minimal time 0.000000 seconds and 161797.000000 usecs13:34
stefan_Maximal time 0.000000 seconds and 226437.000000 usecs13:34
wpwrakpity you don't have any ben+atben. there, things are faster.13:34
wpwraklooks consistent13:34
stefan_wpwrak: getting the 4 atusb from the university was all I asked for :) (I bought 2 myself)13:34
stefan_no lost package13:35
stefan_thats good. Now we can make it faster :)13:35
wpwrakyou said your work was on delay-tolerant networks. seems like the ideal playground ;-)13:37
stefan_heh13:37
stefan_point taken13:37
stefan_wpwrak: so your dirtpan retransmissions come due to lower backoff timer values?13:38
wpwrakyes13:40
wpwrakwell, too low ACK timeouts to be precise13:40
wpwrakdirtpan takes the klingon approach and never backs off :)13:40
stefan_heh13:40
stefan_never, listen, never I say I will backoff :)13:41
stefan_wpwrak: wrt driver name and location. It feels like it should move to drivers/ieee802154/13:42
stefan_name I don't care much13:42
stefan_But its not that easy to share code with atben anymore it seems13:42
stefan_not sure it is worth the sharing13:43
wpwrakyeah, they've diverged a bit. just a few infrastructure bits are common.13:43
wolfsprauloh delay tolerant networks!13:44
stefan_would be confusing just for the generic spi_master bits imho13:44
wpwrakalso, a lot of the code in atben was there just for demonstration purposes, e.g., the classifier or the interrupt forwarding. atben doens't need any of this13:44
wolfspraulif there are pointers to apps/libs that we can port to the Ben, please post13:44
stefan_wolfspraul: yup, doing it as my diploma thesis with 802154 :)13:44
wolfspraulI just want to make sure we create openwrt packages and get it to the Ben...13:45
wpwrak(confusing) agreed. not really worth the trouble for maybe 50 lines13:45
wolfspraulI assume there will be some middleware/libs or what not13:45
stefan_wolfspraul: using ibr-dtn here (implementation for the institute I'm working at). They already use openwrt for other systems and its all free software13:45
stefan_wolfspraul: its the daemon for the protocol and some tools that work with it. Let me search the URI13:46
stefan_wolfspraul: http://www.ibr.cs.tu-bs.de/projects/ibr-dtn/getstarted.html13:47
stefan_wolfspraul: overview: http://www.ibr.cs.tu-bs.de/projects/ibr-dtn/13:47
wpwrakwolfspraul: btw,july news on sunday ? you once mentioned 10 days :)13:47
stefan_wolfspraul: should be easy to grab the openwrt recipes(?) and integrate it with your build13:48
stefan_wpwrak: so we leave both drivers on its own13:48
stefan_wpwrak: but both should be in drivers/ieee802154 as they rely on at86rf230 anyway.13:48
wolfspraulwpwrak: no way, I'm behind some serious todos, need to do those first13:53
wpwrakhmm, maybe you could do the news for relaxation? :)13:54
stefan_wpwrak: Received 1998 from 2000 packets13:55
wpwrakstefan_: good. so you could test the pakcet loss handling as well :)13:56
stefan_wpwrak: thats with 7byte payload btw, lets see what happens with a full payload13:57
wpwrakyes, the delay difference would be quite interesting13:59
stefan_wpwrak: going to get my other parts running now. Measurements and DTN daemon. Once this is in place and I can have real numbers I'll come back to the driver level and work on auto ack and re-submit14:01
stefan_might take some time so14:01
wpwrakgreat14:01
stefan_wpwrak: will keep you in the loop with measurement numbers14:01
stefan_will track the driver closely for all improvements :)14:02
wpwrakheh :)14:02
stefan_more serious, I will come back but time constraints push me to do some other things now.14:03
stefan_Received 2000 from 2000 packets14:04
stefan_Arithmetic mean rountrip time: 0.000000 seconds and 199789.437500 usecs14:04
stefan_thats with full payload14:04
wpwrakso that's a difference of how many bytes ?14:05
wpwrakand the full payload gets sent both ways ? or is it large foward packed and a small ack back ?14:07
stefan_wpwrak: 110 bytes difference and the full data is send both ways14:14
stefan_wpwrak: at86rf230 spi32766.0: invalid PHR 0xdd14:15
stefan_wpwrak: so it was not only a problem for you alone :)14:15
wpwrakcorruption does happen :)14:18
wpwraklet's do the math ... the difference is about 22 ms. of this, 2*110*8/250 = 7 ms are spent actually sending data over the air. leaving 15 ms for copying to and from atusb14:20
wpwrakthe copying happens 4 times. so 3.75 ms per copy. that's about 30 kB/s. hmm. pretty slow.14:21
stefan_wpwrak: what worries me more is the overall high timing value we start with for a small package14:22
wpwrakof course :)14:22
stefan_its factor 10 higher then it should be :)14:22
wpwrakthe driver does a lot of register accesses. maybe they're slow. need to check.14:23
stefan_but atben would do the same14:23
stefan_maybe the usb overhead is really that high?14:23
wpwrakyes, that's what i mean14:24
stefan_ah, now we understand each other ;)14:24
wpwrakthe spi is about as fast as possible in atben14:24
[g2]Hey stefan_ !  Long time14:24
stefan_wpwrak: I'm going to work on my little lowpan-perf pet. Will push it out later and let you know so you can test it on yours ben's if you like14:25
stefan_[g2]: indeed, quite some time. Still embedded linux is a small village ;)14:25
stefan_[g2]: how are you?14:25
[g2]well thx, hope you and yours are the same.14:25
stefan_[g2]: little bit stress to finally get my studies done and doing freelance work in parallel, but nothing dramatic :)14:26
[g2]I'm looking at the 6lopan stuff as of a week or two ago and found my new bff wpwrak that did the atben/atusb14:26
[g2]I'm just about done with the layout for mcu, I'm hoping to start the RF layout later this week14:28
stefan_[g2]: atusb starts to work since last night. :)14:29
[g2]sweet, you just built the hw ?14:29
stefan_[g2]: I just ordered it :)14:29
Action: [g2] got my test atben boards yesterday14:29
stefan_[g2]: needed some 802154 hardware for my diploma thesis and added some on top for myself14:30
[g2]stefan_, can you say what you are making ?14:34
stefan_[g2]: sure, delay tolerant networking over ieee802154 networks14:35
stefan_[g2]: university website is only in german, sorry.14:36
[g2]are you going to add QoS for the long delays :)14:36
stefan_[g2]: but I worked on this topic earlier: http://datenfreihafen.org/~stefan/papers/study-thesis-final.pdf14:37
stefan_[g2]: we don't care of the delays, thats the trick :)14:37
stefan_[g2]: My job is to write/extend a convergence laer between the ieee802154 stack under linux for our dtn implementation14:37
[g2]well if you don't care about delay just add a MicroSD card :) @ 250Kbps you could store years of traffic if one waits.14:40
[g2]stefan_, it's like getting a message from the Moto E780 ??  iirc that was the model of our Linux phone before openmoko.14:41
stefan_[g2]: A780, yes :)14:42
Action: [g2] still has his, but I really don't know why14:42
[g2]yes A78014:42
stefan_[g2]: still have one around, ... somewhere14:42
fossroxhttp://press.web.cern.ch/press/PressReleases/Releases2011/PR08.11E.html17:15
wpwrakfossrox: kewl. that article even uses almost the same words we do ;-)17:29
fossrox:)17:29
wpwrakit someone has contacts to CERN, perhaps they should ask if we could get an invited talk at that workshop. after all, we've been there, done that :) (and if the HEP community should get interested in sponsoring some of our work, that wouldn't be amiss either)17:31
lekernelwpwrak, ask terpstra on #milkymist, he's one of the GSI guys working on the "white rabbit" project talked about in the linked CERN courrier issue17:34
lekernelI think we definitely can take part in the workshop17:34
lekernel(and should)17:34
lekernelI wonder if they still have space in the speaker schedule though ...17:37
lekerneloh, and that's before "13th International Conference on Accelerator and Large Experimental Physics Control Systems"17:38
lekernelsince registration is 550E, i'll probably try gate-crashing :)17:39
lekernelalso, regarding "invited" talks. usually, it's very difficult to get them at academic conferences.17:41
lekernelmost of the times, speakers (or their employers) even pay for the expensive conference registration, paper publishing fees, etc.17:41
lekernelplus travel, of course17:41
wpwraklekernel: (terpstra) perfect. according to http://www.ohwr.org/projects/ohr-meta/wiki/OHWorkshop , they have room for more17:44
wpwraklekernel: more EUR 50 than 550: http://icalepcs2011.insight-outside.fr/Visitors/index.php?onglet=4&idUser=&emailUser=17:45
lekernelthat's for the open hardware workshop17:45
lekernelbut if you want to go to International Conference on Accelerator and Large Experimental Physics Control Systems, it's 55017:45
lekernelor gatecrash for free, there's usually little (if any) security in those events17:46
lekernelthat's how you learn physics17:46
[g2]LMAO17:47
wpwraklekernel: okay, of you want to attend the HEP part, too :) sure, then the usual rules apply17:48
lekernelok so what do we submit?17:51
lekernelI can propose a lecture on the intrinsics of FPGA synthesis and P&R for session 3. there are potentially good engineers there, so it might be worth putting them on free synthesizer projects :)17:52
wpwrakperhaps someone could give a presentation of qi-hardware, maybe based on my FISL slides17:57
wpwrakdescribe what we're doing and especially what tools we have17:57
rjeffriesthe Dangerous prototype people have what seems to be a sustainable business model18:04
rjeffrieshttp://dangerousprototypes.com/2011/07/07/dangerous-prototypes-about-us/18:04
wpwrakrjeffries: sounds like ours, just higher frequency and apparently more profitable :)18:12
rohrjeffries: yeah18:17
rjeffrieswell they are doing lesser projects, but yes, similar goals indeed18:21
rjeffriesAdafruit and Sparkfun likewise, but they ar esomehwat more derivative18:22
lekernelaren't they all doing devices for hobbyist?18:28
lekernelhobbyist = low price + low volume = low technology18:29
lekernel= boring (for me at least)18:29
rjeffrieslekernel they are not doing anything close to as complex as Milkymist. but some of their projects are well above hobbyist18:30
rjeffriesand in some cases they have decent volume.18:30
rjeffriesthey are not attempting any consumer electronics, which is what Nanonote wants to be18:31
lekernelcall me again when they're doing chip bonding, soc design, millimeter wave circuits, etc.18:32
rohrjeffries: i am full with you.18:35
lekernellast week, I learnt about IMPATT diodes. it's amazing how much stuff the average electronics enthusiast totally ignores. to their credit, they won't let me order some easily.18:35
rohrjeffries: less complex means also more market. i like that18:35
lekernelno, less complex doesn't mean more market, lol18:36
lekernellook at the iphone18:36
rohlekernel: arduino has more market than mm for sure. simply because there are more people who understand what they can do with it.18:36
rohalso because there are more people with enough money to even dare to try.18:37
lekernelmaybe, but this has nothing to do with complexity18:37
rohalso. people are afraid of complexity.18:37
lekernelalso, i'd never have made an arduino because I find it's plain boring18:37
lekernelthey're not afraid of the complexity in the iphone, are they?18:37
rohbecause its abstracted away. to be fair.. I dont have interrest in fpgas right now. why? because the sw toolchain is _SHIT_18:38
lekernelalso, the m1 is not a geek toy or a computer.18:38
lekernelso just ignore it, there are tons of ways you could use the m1 without touching any of the fpga software18:38
rohcompared to what we have with broken sw like gcc in the compile world for computers, the xilix/altera/lattice tools all suck18:39
lekernelalso if it's shit, you are welcome to contribute to llhdl18:39
rohlekernel: you dont understand that there are people who have more interresting things to do in life that compilerdesign?18:39
lekernelcompiler design is interesting18:39
wpwrakroh: what could that possibly be ?18:39
rohlekernel: from my pov: fpgas will stay playthings to a very small group of people till that changes.18:39
lekernela lot more interesting than many aspects of arduinos, in my opinion18:40
wpwraklekernel: i bet roh hasn't designed a compiler yet ;-)18:40
lekernelalso, the M1 is NOT a FPGA platform for most people.18:40
rohwpwrak: true. but ive had enough close friends do so in university to know that its nothing 'entertaining' or completely groundbreaking which can be learnt. its boring.18:41
lekernelit has been a bit of that so far, but this is definitely going to change18:41
rohwpwrak: something for people who find graph-theory fun. i dont.18:41
lekernelmost things in the M1 are abstracted billions of kilometers away from the FPGA, you know18:42
lekernelthe FPGA design is just one more thing you can play with, if you wish. but you don't have to.18:42
wpwrakroh: hmm, i'd count cooking up new compilers among the more fun moments of my life. i'm admittedly a bit weak on the theoretical side.18:43
rohi know. but i also have no usecase for a that fast non-mmu cpu or something with vga to do visuals.. so i need to find some serious usecase.18:43
rohi really like the idea of a foss soc tho. thats why i even invest so much time in this project18:45
rohbbl18:46
[g2]writing compilers is over-rated :)18:52
lekerneldepends on the compiler, but I'd generally agree18:53
rjeffrieslekernel would havingt linux on milkymist make it easier to attache standard mass market (low cost) USB periferams, e.g. any USB keyboard, not just a hand selecte dto work keyboard?18:58
rjeffriesand to have a righeous linux milkymist SOC needs at least a simple MMU18:59
lekernelnot all USB keyboards work simply because of bugs in the softusb firmware which are a pain and time sink to fix, and no one including me feels motivated to do that, so we just ship a known good keyboard.18:59
lekernelok the MMU topic is coming all the time, won't you just please FDI?18:59
rjeffriesFDI ??19:10
wpwrakrjeffries: Foreign Direct Investment. he wants your money :)19:12
rjeffrieswhat a sweetheart. ;)19:12
wpwrak(any other interpretation, like the one offered by urbandictionary, would be rude ;-)19:13
rjeffriesi'll c heck that out. I can not imagine lekernel ever being rude, even accidentally19:13
rjeffriesah yes. good point, except that is not my skill set.19:14
wpwrakim-pens-able19:14
wpwrakrjeffries: so you're a creationalist ? created with a fixed skill set ? :)19:15
Action: [g2] looks at urban dictionary on the android19:15
wpwrakrjeffries: meanwhile us atheists happily learn new things ;-)19:15
rjeffriesI have other fish to fry. pretty interesting fish (to me)19:15
wpwrakrjeffries: well, i hope she's pretty ;-)19:17
rjeffrieslekernel /sebastian needs to attract a couple of more hardcore people like himself19:17
lekernelno, I need to sell the current stuff to VJs, musicians, etc.19:17
rjeffriesyou assume a she. well, so does my wife19:18
wpwrakcherchez la femme ;-)19:18
rjeffrieslekernel understood.19:18
lekerneltbh from a technical point of view I find the MMU a bit boring and unneeded anyway for my purposes19:19
lekernelplus, why should I do it for the sake of a free SoC? I get so few contributions to the FPGA design on this project.19:20
wpwraklekernel: mmu -> linux -> a chance to escape driver hell. so there should be value for you. but i understand that you have other things to do at the moment19:22
lekernelthere's no driver hell. USB gadgets are not supported, period.19:23
rjeffriesone could conclude that open hardware (great concept) is a non-starter for SOC19:25
wpwraklekernel: you'll write your fingers fuzzy with that ;-)19:25
lekernelrjeffries, oh yeah, maybe you are right, heh.19:25
rjeffriesdid you just accuse me of being RIGHT? omg19:26
rjeffries;)19:26
lekernelbut also keep in mind Mako's statistics which say that the vast majority of open source projects have ONE developer19:27
lekernelstuff built by "the community" rarely happens19:27
lekernelhttp://mako.cc/writing/hill-when_free_software_isnt_better.html the corresponding talk is also interesting to watch19:28
[g2]wpwrak, I've got a kicad question on the atmega board I'm working on.19:30
wpwraklekernel:  i often see a pattern of one main developer and people contributing on the side. that's what makes most sense anyway, each doing what "scratches their itch"19:31
wpwrak[g2]: fire away :)19:31
[g2]something is happening with the libraries/modules where the footprint of the resonator doesn't seem to get picked up.19:31
wpwrak[g2]: is it listed in *.pro ?19:31
[g2]yes, I'm pretty sure19:31
wpwrak[g2]: you can edit *.pro from within kicad, but it's very inconvenient. easier to just exit kicad and edit it with a text editor19:32
[g2]It's seen in the schematic, but not on the board file19:32
wpwrak[g2]: library (.lib) and footprint (.mod) are different things19:32
[g2]right.19:32
wpwrak[g2]: how did you make the footprint ? with kicad's footprint editor or fped ?19:33
[g2]I tried grapping other modules from the links on kicad.org19:33
wpwrakthat can work, too :)19:34
[g2]later today I was going to post the whole project on github or something similar.19:34
[g2]then, you can point and laugh at my errors :)19:35
wpwrakin *.pro, you should have a section [pcbnew/libraries] and there you should have an entry LibName#=path/to/your/file19:35
[g2]correct. I do.19:35
wpwrak# is a number. numbers must be consecutive. so if you have a LibName1 and LibName2, the next must be LibName3.19:36
wpwrakthe path can be relative, which i'd recommend to use19:36
wpwrakfurthermore, the file name probably must be without the .mod19:36
wpwrakif you have all this, then you should be at least able to place the footprint in pcbnew with "Add modules" (the chip symbol, 4th from the top)19:37
wpwraknow, that's not what you want to do, but it's a test that everything else works as it shuold19:38
wpwrak(i.e., try to place the footprint/module and see if it looks right. if all is well, exit without saving. or if you've saved, delete it the next time.)19:39
Action: [g2] checking on laptop19:39
[g2]wpwrak, the footprint is there in pcbnew and I do the add module.19:45
wpwrak[g2]: excellent. now, quit pcbnew and fire up eeschema. then invoke cvpcb. there, you should see the component in question. is a footprint associated with it yet ?19:46
[g2]wpwrak,  Ok, will do, but I didn't change anything in the .pro19:47
wpwrakthat's okay. the .pro seems to be fine19:47
[g2]ahh.. It's case-sensitive right ?19:49
wpwrakprobably, yes19:49
qi-bot[commit] Werner Almesberger: atusb/fw/: added improved support for interrupt synchronization (master) http://qi-hw.com/p/ben-wpan/42483d619:51
wpwrakstefan_: the commit above explains how it all works19:51
kyak"I think we're also seeing some shift from the mailing lists to IRC." - yeah, it's 2011! :)19:59
qi-bot[commit] Werner Almesberger: atusb: changed interrupt synchronization from fixed delay to SPI_WRITE2_SYNC (ben-wpan-stefan) http://qi-hw.com/p/qi-kernel/b2d71a919:59
wpwrakkyak: 2011 would be "from mailing lists to facebook" ;-) 2010 it would have been twitter. 2009 ... myspace ? fads come fads go ;-)20:00
kyakyou are right, but what's going to be next?20:01
wpwrakgoogle+ ?20:02
kyaki'm glad i'm stuck in IRC :)20:03
[g2]wpwrak, THX !!  So basically what I think happened is the part didn't have a footprint associated with it until I added with cvpcb (per you great directions :) 20:07
[g2]not it's associated, so the layout has a footprint.20:08
[g2]that's was the missing link for me.20:08
wpwrak[g2]: yup, sounds right20:09
wpwrak[g2]: so how's the overall kicad experience for you so far ?20:09
[g2]wpwrak, pretty good actually.20:09
[g2]I think for small boards, it's all very workable20:10
[g2]I like eschema and the text based nature of things20:10
[g2]meaning it could all be scriptable and one can do diffs20:10
wpwrak[g2]: people also have hand-routed multilayer boards. lemme check how many layers ...20:11
wpwraksix layers (project xue-rnc)20:11
[g2]wpwrak, um 10 years ago I did the frimware for a network processor that was on a 19x22 inch board 24 layers deep. :)20:12
[g2]there were several 1M+ FPGAs on the board too.20:13
wpwrakwhoopie20:13
[g2]SER-DES links on the backplane.  We had a 1 Million+ subsciber GGSN running in 200120:13
wpwraksounds quite massive20:14
[g2]GGSN was a 3G data switch, we were ahead of our time20:14
wpwraki haven't done multilayer yet. should be fun :) i hope we can find some money for making a Ya NN. that would be the ideal opportunity to dive into such things20:15
wpwrak[g2]: (3G in 2001) pretty much indeed20:15
[g2]Ya NN ?20:16
wpwrak[g2]: the hypothetical successor of the ben nanonote20:17
[g2]Ah...20:18
[g2]wpwrak, also is there a way in pcbnew to have it drag the traces around that are connected to the IC ?20:21
wpwrak[g2]: unfortunately, not really. all that's there is the block move. but not a proper drag (you'd need an interactive router for this)20:22
Action: [g2] suck :(20:24
[g2]kicad -120:24
[g2]pcbnew -120:24
[g2]kicad +020:24
wpwraknaw, after a while, you get good at quickly ripping up and manually re-routing things ;-)20:26
Action: [g2] practices20:32
stefan_wpwrak: will test the commit tomorrow, just back from bbq, going to sleep now20:42
wpwrakstefan_: (bbq) hehe :)20:43
[g2]sweet dreams.20:45
rjeffrieswpwrak do you remember how many layers Ben NN pcb is?20:53
stefan_night all20:54
wpwrakstefan_: hmm, eac register read or write from user space (via USB) takes about 1 ms. so far, so good.20:55
wpwrakrjeffries: i think it's 6 layers20:55
wpwraksame for buffer read/write20:56
rjeffrieswhen Carlos did his SAKC I think he may have only done 4 layers? not sure20:57
wpwrakcould be. i think he's at 2 layers now.20:58
wpwrakbut a nanonote with a similar keyboard layout would need at least one additional layer20:59
wpwrakand you almost certainly want a good ground plane20:59
wpwrakfor 4's the minimum20:59
wpwraks/for/so/21:00
--- Fri Jul 8 201100:00

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