#milkymist IRC log for Sunday, 2011-11-27

stekernwpwrak: it's a mouse and keyboard. It's the keyboard that is trigging the error messages, but the mouse isn't really working either anymore.03:33
stekernthe error messages are not always the same neither03:33
stekernthat is a typical boot-up03:34
xiangfustekern, thanks for report the bug.(about reflash_m1.sh). I will try to fix it.03:35
stekernxiangfu: nice one03:35
wpwrakwow. an even a low-speed device !03:36
wpwrakstekern: what does lsusb -v on a linux pc have to say about it ?03:36
wpwrakxiangfu: btw, i think it would be good to have a file with md5sums or sha1sums, and have reflash_m1.sh check them. that way, if someone starts a reflash and hits ^C in download, you don't end up using a bad cache03:38
xiangfuwpwrak, yes. that is what --snapshot do. :)03:38
xiangfuwpwrak, oh. --snapshot only check once before the download. yes. it should be better check again before release.03:39
stekernwpwrak: http://pastebin.com/gKEDPdA803:42
wpwrakah, nice. now all that's missing are md5sums on the server, right ?03:42
stekernpretty funny, it's a Logitech keyboard, but it identifies itself as a Novatek product03:43
wpwrakelectronics industry: through incest we thrive ;-)03:44
xiangfuwpwrak, yes. there is no md5sums under: http://milkymist.org/updates/current/03:45
stekernwould have thought that Logitech are large enough to put their own VID:PID in there03:45
wpwrakthe interface for the mouse is too cryptic to be understood by our stack03:45
stekernmouse? that's the keyboard03:45
wpwrakah, i thought it was a combo device03:46
wpwrakwhat does the mouse look like ?03:46
stekernit works fine under 1.0.103:46
wpwrakand is the keyboard full-speed or low-speed ?03:46
wpwrakso we have a regression ? hmm03:46
stekernshould be low-speed03:46
wpwrakmight be interesting to see the report dump for keyboard and mouse: http://lists.en.qi-hardware.com/pipermail/discussion/2011-November/009206.html03:49
wpwrakunfortunately a bit messy, the process03:49
stekernnaah, didn't look to bad of a mess03:52
wpwrakafter soc, rtems, and fn, you're of course in perfect shape for it ;-)03:55
stekernthat's the keyboard04:05
stekernhere's the mouse: http://pastebin.com/uxETdFhX04:05
wpwrakthe mouse looks very normal04:06
wpwrakdoes the BIOS enumerate the mouse ?04:07
stekernthat's with only the mouse connected04:10
stekernso I was wrong when I said that only the KB shows errors in BIOS04:10
wpwrakokay, that explains why it doesn't work :)04:12
wpwrakthe only thing nasty i get are a few rx timeout errors. but maybe that's the same04:13
stekernI could run a bisect on it to see what broke it, but do you have any suspicions yourself?04:21
wpwrakfor now, i'll chase my rx timeout. some changes affect others in strange ways, so i first want to bring a bit of order in this stack04:27
stekernok, yeah makes sense. might be that finding the guilty commit isn't going to give information about the root cause, if it's just coincidential04:29
wpwraki suspect it may be the loop unrolling, which changed timings all over the place. so anything else what was marginal may have fallen over, etc.04:31
wolfspraulgood news - we sold a Milkymist One to a low-volume/artsy furniture maker08:58
wolfspraulshe has something called 'media wall' that integrates a projector/screen, and dmx-controlled lights and lasers08:58
wolfspraulso xiangfu has some work now ;-) he will get that same dmx equipment and try to integrate it all well into the media wall08:58
wolfspraulfirst public appearance is next Friday, where we will also try to take some pictures or videos08:59
wpwrakwolfspraul: dmx furniture. very wow ;-)13:05
wpwrakwolfspraul: btw, how's the M1 stock forecast ?13:12
wolfspraulyeah, and about pricing. she paid 500 USD without even asking how much it is until after she made the decision to try this out in her media wall.13:12
wolfspraulwhich makes sense, if the speakers in the media wall cost a little over 6000 USD, for comparison13:12
wpwrakhehe ;-)13:12
wolfspraul"stock forecast" :-)13:13
wolfspraulno forecast, we sell one by one and unless we find new distributors/resellers we will have plenty of stock for a while13:13
wpwraknot worried about running out of stock yet ? :)13:13
wolfspraulsold around 40 now, as before13:13
wolfspraulnot worried13:13
wpwrakah, good.13:13
wolfspraulI am worried about pusing to rc4 as hard and fast as possible13:13
wolfspraulcannot let down13:13
wpwrakironically, selling too well is a risk for M1rc3 ;-)13:14
wolfspraulthen worst case we take some rc3 boards out and think of alternative use cases for them13:14
wolfsprauloh we can handle it either way13:14
wolfspraulthe important thing is what sebastien calls more 'publicity'13:14
wolfspraulI'd rather say we need to get the product into the hands of more excited users who spread the word13:14
wpwrak(pub) yup, but let's be at our best, with sunday clothes on and prepared. public attention can be a tad embarassing if they catch you with a 4 days beard and in your underwear13:15
wpwrakyes, that too. we still have to work the music side. so far, we went mainly to the IT geeks13:16
wpwrakbtw, any more reactions from heise ?13:16
wolfspraulno no :-)13:16
wolfsprauland I wouldn't expect that as I said13:16
wpwrakokay. there's no hurry there13:16
wolfsprauljournalists are flooded with stuff, just the other day rejon and I went to dinner with one who said he deletes everything sent to him13:17
wolfspraulthe way he writes his stories is by going out and searching and finding them himself13:17
wolfspraulI'm not surprised hearing that13:17
wpwrakfunny :)13:18
wpwrakat least not the lazy type13:18
wpwrakfor heise, perhaps you could try my text. that one puts things into the right perspective. ah, one thing is missing - shuold also give a forecast for the "final" version (i.e., the one not just for freaks and developers)13:19
lekernelwhat's missing?13:21
lekerneldocumentation? what else?13:21
wpwrakfull-speed usb, usb-midi, more midi variables, better handling of them, more efficient midi processing, ... (and that's just the midi side ;-)13:23
wpwrakthen we have of course dvi out that will only come in m1rc4. get rid of the "vga only" taunting :)13:27
wpwrakbtw, i've noticed some color borders inside gradients lately. i think they're a new "feature", but haven't checked against an older version yet13:29
lekernelwhat gradients?13:30
wpwrakwhen i have adjacent areas of similar color (or different shades of the same color), there's sometimes a ~1 pixel band at the edge that's neither of the two colors or something in between13:32
wolfspraullekernel: how about the LV3 ?13:32
wpwraklemme see if i can snap a picture13:32
lekernelwhere? GUI? patches? with which version?13:33
wolfspraulit sounds it needs more work on the full-speed, usb-midi and variable/flickernoise side13:33
wpwraklatest version :)13:33
wolfspraulhave to test picture support a little, I think we need to work on how to get pictures to the m1. but looks like we have a great start now.13:33
lekerneland latest version of what? soc? fn?13:33
wolfspraulfor the media wall, I need a way to disable standby, so that after a power disconnect the m1 comes back rendering by itself13:33
wolfspraulhow about the 5-6-5 greenish problem - any plans to work on that?13:34
lekernelwolfspraul: needs full speed support, which is a massive problem with USB..13:35
lekernelit doesn't enumerate at all atm13:35
lekernelcomplete silence13:36
lekernelI don't know what wpwrak has in mind with "better MIDI support", the only problem I see is it'd need the ability to listen on multiple channels, which is minor13:37
wolfspraullekernel: ok so what are your next plans? I can always give you long wishlists if you want :-)13:42
wpwrakshould be latest of everything13:42
wolfspraulI hope we get the lv3 to work, for example. we can start selling as a bundle and cross-sell a little13:42
wolfspraul(or other faderfox controllers, not just lv3)13:42
wpwraklemme show it ...13:43
lekernelwolfspraul: if you can justify them...13:43
wolfspraulif we cannot get usb-midi to work, then we have the option of taking some of his old din-midi controllers, but he has limited stock and availability13:43
wolfspraullekernel: what are your next plans?13:44
lekernelget a release with image support (and a few things) out, see http://www.milkymist.org/wiki/index.php?title=Flickernoise_roadmap#1.113:45
lekernelI did most of this except background DHCP, /ssd block, and black screen/progress bar at startup13:46
wolfspraulcan we remove standby, or add a way to disable standby in the gui?13:46
lekernelremove the poweroff state altogether?13:47
lekernelthat's easy13:47
wolfspraulyes, remove imho13:47
lekernelexisting M1s will need a JTAG reflash though13:48
wpwrakcolor effects: http://downloads.qi-hardware.com/people/werner/tmp/colorfringe.jpg13:48
lekernelwpwrak: ah, in the effects then13:48
wpwrakhere's also the .xcf, so you can get rid of the frames: http://downloads.qi-hardware.com/people/werner/tmp/colorfringe.xcf13:48
lekernelwpwrak: that appeared with soc 1.1, right?13:48
wpwraki think so, yes. i noticed it after image support, but maybe it was there before13:49
wolfspraulah they need a jtag reflash because the standby bitstream is not updated over the web update, right?13:49
wolfspraulwell, I think that's ok13:49
wolfspraulwe have a very limited user base and the few that are active are easily reachable13:49
lekernelmaybe it's a small (as in number of lines of code to modify, but that's often inversely proportional to the time wastage potential) bug introduced with prefetching13:49
wpwrakand they all have jtag :)13:49
wolfspraulplus this is a specialized feature likely only valuable to new customers anyway, someone who wants to embed m1 in a larger installation13:49
wolfspraullekernel: so yes, remove standby would be great I think13:50
wolfsprauljtag reflash should be ok13:51
wpwrakitsa bti strange that it precisely follows gradients. but yes, prefetching would be the sort of major change that could have such an effect13:51
wpwrakin any case, it makes one look for vga hardware problems. cable reflections and such :)13:51
wpwrakwolfspraul: i think all the rest of the users will also celebrate the removal of standby :) though it would be nice to have a FN -> off state. but that could perhaps be entirely inside FN.13:52
wolfspraulwhy is a black screen on startup better than no signal? (just curious)13:52
wolfspraul(reading the 1.1 changelog)13:52
wpwrakno signal looks like "it shut down again"13:53
wpwrakand "intelligent" displays may go on a rampage then, trying to find out which other of their 1000 input sources have a valid signal13:53
wolfspraulah ok13:55
wpwrakhere's an experimental midi2osc mapping for the LV3: http://projects.qi-hardware.com/index.php/p/wernermisc/source/tree/master/midi2osc/LV314:01
wpwrakthe comments also explain how some of the things work14:01
wpwrakit doesn't use MIDI to alter LEDs, though14:01
wpwraki don't even know if we have OSC -> MIDI. checking ...14:06
wpwrakhmm, i guess not. even though there is a send callback14:08
lekernelif we remove standby, how do we sync the flash filesystem safely?14:14
wpwrakhave the standby loop in FN ?14:15
wolfsprauldon't understand. you mean on software shutdown?14:17
wolfspraulI think we can safely say most people don't use that, 90% or more will unplug the DC cable.14:18
wolfspraulso even if you remove the safe syncing, you actually don't remove it because people don't use it14:19
wpwrakthat's probably a very correct assessment :)14:20
lekernelso? just rely on the 10s periodic write cache flush?14:20
wolfspraulsame as we do now when people unplug DC, yes14:21
wpwrakmaybe make it faster :)14:21
wpwrakpeople will not know about those 10 seconds. they just have "things go wrong" from time to time14:21
wpwrakconclusion if this happens too often: m1 is unreliable14:21
wolfspraulno difference from now. people don't use the software shutdown, if Sebastien uses it maybe he's the only one...14:22
wolfspraulsince everybody loves sales numbers14:36
wolfspraulwe sold 43 rc3 to date14:36
wolfspraulreal sales14:36
wolfspraul5 to the US14:36
wolfspraul1 to China, 2 to Taiwan14:36
wolfspraul2 to Australia14:37
wolfspraul33 to Europe14:37
wolfsprauldoes this add up?14:38
wolfspraulgood results I think. if we keep pushing with strong updates and regular sales activity and spreading the news, it might actually take off :-)14:39
wpwrakreal sale = to customer. or also including to distributors ?14:39
wolfspraulalso to distributors14:39
lekernelno news from MSL by the way?14:39
wpwrakokay, so some more are sitting at distributors. maybe ~30 in the field then ?14:40
wolfspraulcould be, yes14:40
wolfspraullekernel: I think the news from MSL is very clear14:40
wolfspraulat least to me14:40
wpwrakso that would mean that, if the rate stays roughly the same, you'll have stock until the end of the year14:40
wolfsprauluntil today they haven't replied to very simple questions about payment terms, even though I asked 3 times by now14:41
wolfspraulM1 doesn't fit their business model, imho14:41
wolfspraulI'd say - move on14:42
wpwrakprobably won't be able to sneak m1rc4 in before CNY. ah well, it's just one week plus ripples.14:42
wolfspraulwpwrak: oh I don't calculate like that14:42
wolfspraulM1 sales are still very slow and the result of heavy one-on-one efforts14:42
wolfspraul90% of them14:42
wpwrakyeah, that's true14:43
wolfspraulthe product has to get into the hands of someone who truly likes it14:43
wolfspraulnot just in words, but in actions14:43
wolfspraulwe cannot accelerate that happening much beyond what we do already, and our own belief in the product, and improving the product14:43
wolfspraulit's on a very good track imho14:43
wpwraklekernel: have you thought of touring some clubs in berlin ? offer them a free night of vjing. if they accept, spread the word on the online fora vjs hang out on, so they come and have a look14:44
wolfspraulah yes, quite a few things you could do in Berlin14:45
wolfspraulsuch as that de-bug.de music magazine, show m1 to them14:45
wolfsprauland then this schneider's laden14:45
lekerneli was handing out brochures yesterday again *g*14:46
wolfspraulplease continue to believe in the product14:46
wolfspraulit has a chance if you like it and push forward, really14:46
wolfspraulI think so :-)14:46
lekernelactually, my scopesession presentation did have some impact, at least locally14:46
wolfsprauljust had that good experience at the furniture maker today14:46
wpwrakthe "free vj night" should be a very slow night. one on which the other vjs aren't busy :)14:47
wolfspraulthe next level can only come if we run into someone with *serious* business interest14:47
wolfsprauland sooner or later we will be a little lucky and run into that person14:47
lekernelbtw, I have totally run out of stickers, and I only have 10 or so brochures14:47
wpwrakwolfspraul: speaking of which, how's the investor hunt going on your side ? :)14:47
wolfspraulin the meantime we should believe in the product ourselves, use it, and continuously improve it14:47
wolfspraullekernel: that can be fixed ;-)14:48
wolfspraulI am also not 100% sure what is more important now - spend working hours on sales activity, or spend working hours adding features14:48
wolfspraulusb-midi is probably still something super important, most people from the dj/vj scene will ask for it at some point14:49
wolfspraulbut one by one14:49
wolfspraulwpwrak: investor hunt, ask Jon14:49
wolfspraulI'm on the customer hunt14:50
wolfspraulJon has a good presentation at Intel in a few days14:50
wolfspraulhe will use his M1 for presentations now ;-)14:50
wolfspraullekernel: can you look into presenting m1 at de-bug.de or schneider's laden?14:52
wolfspraulwe should also see what's going on at bearstech14:53
wolfspraulthey had all these great plans with 10 units a month etc., and I accepted all their terms, but after my mail silence :-)14:53
wolfspraulon the other side I am sure when they run out, they know where to ask...14:53
lekernelguyzmo ?14:53
lekerneland de-bug/schneider's lader, yes ...14:54
wolfspraullekernel: did I send you 1 or 2 rc3 ?14:55
wolfspraulmy records show 1, but my memory somehow says 214:55
wolfspraulah ok14:55
wolfspraulgood you can use that to show around, or even sell if you can :-)14:55
lekernelI haven't tried to sell it... I need it to show around14:56
wpwrakmmh. that's what i thought :-( then we're doomed.14:56
lekernelwpwrak: ?14:56
wpwraklekernel: i dont expect jon to be able to find useful investors for the project. and wolfgang isn't even trying.14:57
wpwraklekernel: what jon finds are people who have interesting projects of their own. but M1 isn't part of their plans.14:57
lekernelthe mozilla thing isn't too bad imo14:58
lekernelif there can be little customization14:58
lekerneland if they're happy with PAL resolution video14:58
wpwraklekernel: and anything more flexible takes a lot of time. if jon would have been working on this since, say, 2-4 years ago, he wuold have a good lineup by now14:58
wolfspraultoo much theory14:58
wolfspraulthe next step for m1 is to find an individual that helps us take it to the next level14:59
wpwrakyes, a knight in shining armor ;-)14:59
wolfspraulno it always works like this14:59
wpwrakindeed. that's how gotham city got save every single time (-:C14:59
wolfspraulsomeone has to like the product and team14:59
wpwrakindeed. it's better if your investors don't hate you just yet :)15:12
wolfspraulwas off by 1, 42 sold15:13
wpwraklucky number :)15:14
wolfspraulregarding investors, I can only repeat15:24
wolfspraulthere is no money to fund a hobby15:24
wolfspraulMilkymist One has to fight for its position in a competitive market15:24
wolfsprauland an investor, or customer, has to believe in the strength of the product15:24
wolfspraulthat's the only honest and long-term functioning way to do it15:24
wolfspraulthe world can easily live without Milkymist One, or the entire Milkymist - guaranteed15:25
wolfspraulnot many people will miss it15:25
wolfspraulso we have to make a big point about why it's great and cool and will have a future15:25
wolfspraulyour summary of "don't expect Jon to find" and "wolfgang isn't even trying" is both wrong15:27
wolfspraulthere's a lot of rich folks in Argentina I read - why don't you chat them up in your favorite nightclub and make them fork over a cool million or two? :-)15:28
wolfspraulinstead of spending it on a yacht upgrade, for example15:28
wolfspraulso rather than chasing this kind of illusion, I work my way bottom up. it feels great and we make progress and eventually we get the due rewards.15:28
wolfspraulwhat those rewards will exactly be is yet to be found-out :-)15:29
wolfspraulI always think we are just one small step away from some big break-through, of course I have no problem feeling like this for years :-)15:29
lekernelillusion? it was the case with openmoko, no?15:29
wolfspraulopenmoko was privately financed much like milkymist is15:30
wolfspraulof course milkymist has a lot less financial power, but the financing background is the same15:30
wolfspraulwe should actually get outside investors, I agree with werner15:30
wolfsprauleven if they would bring in some harsh dose of realidad :-)15:31
wolfspraulat least since our tech is all free, this option always remains open, which is great15:31
wolfspraulat this point I would work down the many leads we have in our minds, magazines, journalists, retail stores, friends, etc.15:32
wolfspraullekernel: you can shoot emails in all directions that we have M1 ready now and people should buy it and be inspired about what it can do today and tomorrow...15:33
wolfsprauland then either they take the bait or not15:33
wpwrak(finance a hobby) doesn't matter if it's a hobby :) you're not selling the fun of doing it but the potential as a product. i'd say M1 is in rather good shape for that, with most of the R&D risk already removed. still needs some more work and has to spend more time with with focus group, but the project should look rather tasty to investors.15:33
wolfspraulif not, we need to think why not and what we can improve15:34
wolfspraulyes I don't disagree15:34
wolfspraulso no need to talk to each other here, we need to go out and find new friends15:34
wolfspraulwhich Jon for example is doing every single day15:34
wolfsprauleither people like it or not - at some point you also have to give some space for people to simply not be interested, not get it, think it's a fad, etc.15:35
wolfspraulthat's life15:35
wolfspraulif the entire Milkymist disappears from Earth tomorrow, not many people will miss it. we have to change that first.15:36
wolfspraulwhich I believe is happening15:36
wpwrak(access to rich people) that's what takes time. if they're not good at fending off the 99.9% who'd want some of that money, too, they're not rich anymore :)15:36
wolfspraulI was really relieved to get the good feedback from n0carrier, for example15:36
wolfspraulwpwrak: correct!15:36
wolfspraulso we have a live study now with n0carrier - how long will his Milkymist passion last?15:36
wolfsprauland when is the product drawered15:37
wpwrakso i dedicate only a small fraction of my time on that. i'm not completely dismissing this route, but it's a very very long shot15:37
wolfspraulI hope it lasts, and I hope we can keep making it worthwhile for him15:37
wolfspraulbut it's not for granted, the rest of the world is also working15:37
wpwrakn0carrier is a very promising start. we need a lot more feedback from the field. that's why i'm asking about the possibility of taking M1 to clubs, to demo it there.15:40
wpwrakthis will also help to refine the communication with the focus group15:41
wolfspraulof course15:41
wpwrakthink of your favourite VJ coming to you with an IT project he dreamed up. i'm sure you'll poke it full of holes within mere minutes.15:42
wolfspraulI will make another sales attempt at Lantern Club in due time15:42
wolfspraulyou want to try to sell some in Buenos Aires? :-)15:42
wpwrakpart of it will be that he'll be sending all the wrong competence signals. signals you've spent a decade and more detecting and interpreting.15:42
wpwrakBUE is difficult due to very high taxes and import restrictions. also, i won't try to start pushing it if my involvement is about to end.15:44
wpwrakbut yes, i considered the possibility of showing it around here15:44
wolfspraulyou will quickly realize it has marginal value only right now, an exotic novelty15:45
wolfspraulso you need *A LOT* of love on the other side to proceed thinking about it more, or even buying one15:45
wolfspraulnot sure how to overcome this yet, I keep learning15:46
wpwrakcorrect. the main objective would not be to sell it on the spot but to first get feedback. establish contacts, etc. prepare for the "real launch", maybe in april or may.15:46
lekernelhave an inspiring concept? like mixing an ipad with a wacom tablet for the next Milkymist UI :)15:47
wpwrakwhy the wacom ? :)15:47
lekernelbecause you can do tons of stuff with the pen15:47
lekernelpressure sensitive15:48
wpwrakand yes, in my mental image, i see a touch screen on the top of the perfect milkymist :)15:48
lekernelrotation sensitive15:48
lekerneli.e. you can touch a control with the pen, and rotate it to change its value15:48
wpwrakyou may have noticed that i started to use "milkymist" and "milkymist one" as distinct terms :)15:48
lekerneltilt sensitive15:48
wpwraklekernel: you're still control-starved ;-) you shuold play a bit with the LV3. that'll cure this :)15:49
wolfspraulyeah please get the lv3 to work15:49
wpwrakyou can already use it (via a PC)15:49
wolfspraulsince we already started this with faderfox, a good direction I think15:49
wolfspraulof course maybe we have to give up, but sounds a little premature to conclude this right now15:50
wpwrakusb-midi is a given. we need that anyway.15:50
wolfsprauloh btw, I think I mentioned it earlier - xiangfu will start to look into dmx this week15:50
wolfspraulhe will buy some equipment and off he goes, and hopefully can produce some results quick, in the form of documentation, patches, fixes15:51
wolfspraulI hope he doesn't immediately run into major low-level bugs...15:51
lekernelworst case we'll just throw in some proprietary USB ASIC and let some engineers in the dark care about petty USB idiosyncrasies15:51
wpwrakLV3 depends on whether we find it convenient to use. that'll take a bit of time. you need to spend a day or two with each control concept before your brain has adapted to it15:51
wpwraklekernel: have you joined wolfgang's defeatist's club ? :)15:52
wolfspraulthat kind of logic makes people think whether the entire Milkymist concept may be wrong btw :-)15:52
wpwrak(xiangfu) yeah, you mentioned that you landed him a DMX contract :)15:53
wolfspraulRaspberry Pi! :-)15:53
lekernels/Milkymist/open source hardware/ :-)15:53
wolfspraul"dmx contract" sounds a little hyperbole15:53
wolfspraulbut it's a paying customer and we use the chance to dive into dmx15:53
wolfspraullekernel: Arduino is doing well, and maybe Raspberry Pi next? :-)15:54
wpwrakyour customers must love you. they buy a USD 500 box and get USD 5000 worth of consulting for free ;-)15:54
wolfspraulI don't know any/enough details on the usb situation, so it's hard for me to say15:54
wolfspraulI can just watch and follow the arguments15:54
wolfspraulbut if it's too hard to get full-speed usb to work, hmmm....15:55
wolfspraulthat will make a lot of people think15:55
wpwrakbut well, if it gets DMX properly tested and debugged, ...15:55
wpwrakhey ! full-speed "works for me" ;-))15:55
wolfspraulsure I know15:55
wpwrak(doing my best sebastien impersonation :)15:55
wolfspraulI'm just saying how others may look at things and interpret them15:55
wpwrakthe remaining bugs will die, too15:56
wolfspraulI think it's just bugs bugs bugs, and details we don't understand, maybe hard to fix etc.15:56
lekernelwolfspraul: both arduino and raspberry pi rely on some proprietary ASIC company to take care of >99% of their project's technical details15:56
wolfspraulbut if we give up - after how much effort? - then what should others think?15:56
lekernelthis isn't what I call open source hardware15:56
wpwrakin an open source project, bugs are always visible to everyone. nothing new there in the last 20+ years15:56
wolfspraulbut they are not struggling with full-speed usb, not even with high-speed usb15:56
lekernelof course - the pesky details have been handled for them by the ASIC manufacturer15:57
wolfspraulI am just describing how others may perceive things.15:57
wolfspraulwe can say "usb is low-speed only"15:57
lekerneljust like on our USB-JTAG cable15:57
wolfspraulno problem15:57
wolfspraulbut it's a really tough thing to say15:57
wolfspraulbasically it means keyboard and mice only (and even there we still seem to run into bugs)15:57
wpwraknaw, full-speed will be there. that's a problem that can be solved.15:57
wolfspraulhopefully, yeah15:58
wpwrakjust way a few more years, and mice and keyboard will switch to full-speed too.15:58
wolfspraulotherwise din-midi controller can fill in for a while15:58
wolfspraulbut we have to believe that we get full-speed usb to work15:58
wolfspraulotherwise we cannot find new people to join, customers, let alone 'investors' - how do we explain to them what we do here?15:59
wolfspraulthey will think we are crazy :-)15:59
wpwraki don't see any fundamental problem. it's not as if we'd be pushing the limits of the FPGA anywhere.15:59
wpwrakmost likely, CRC will have to go into the FPGA. LFSRs are made for that.16:02
wolfspraulhave you tried your lv3 directly on m1?16:02
wpwrakworst case would be that sending of ACK would also have to go into the FPGA. but i don't think that'll be necessary. the timing with navre is tight but not hopeless.16:03
wpwrakwe don't have usb-midi protocol support on M1 yet16:03
lekernelwpwrak: we can increase the frequency to 72MHz too16:03
lekernelwpwrak: does your lv3 enumerate at all?16:03
wpwraklet's see ...16:03
lekernellast time I tried all I got from mine is silence16:03
wpwrakhmm. and maybe i shuld flash a soc that actually works :)16:04
wpwrakRX timeout error and  SETUP not ACKed .. then VID: 0853, PID: 010016:05
wpwrakbut that's the hhkb. without it ...16:06
wpwrakjust timeout and no ack16:07
AlarmHello I have questions to ask about the organization of files in the SDK Milkymist. I'm confused16:09
Alarmsaw the list of files to flash the M1. I do not know what is the use some files?16:13
Alarmfor example standby.fpg ?16:14
lekernelstandby.fpg is the first bitstream to be loaded, which then loads the regular or rescue bitstream depending if btn1 is pressed or not16:14
GitHub38[milkymist] sbourdeauducq pushed 1 new commit to master: http://git.io/RwBYPw16:18
GitHub38[milkymist/master] standby: auto-on - Sebastien Bourdeauducq16:18
lekernelwolfspraul: done16:18
wpwraknow you have the headline for the quarterly news :)16:18
wolfspraulah don't mention that again16:19
wpwrakyou will not write them, will you16:20
wolfspraulthere seems to be an infinite stream of interruptions preventing me from finishing, yeah16:21
wolfspraulfor example now I hope I get a download link from n0carrier16:21
wolfspraulor rather Emily16:21
wolfspraulbut when/how/where?16:21
wolfspraulwait or not wait16:21
wolfspraulnot even mention the elusive Warsaw videos...16:22
wpwrakaah, you're planning on the news to end history. i see :)16:22
Action: lekernel has given up on the Warsaw video already. one thing less on my mind...16:22
AlarmIf I understand correctly, there are three bitstreams: fjmem.bit standby.fpg and soc.fpg16:27
lekernelfjmem.bit is only temporarily loaded to give flash access over JTAG16:30
lekernelstandby, soc and soc-rescue are permanently written to the flash16:31
Alarmbios.bin is it bsp (RTEMS)?16:43
stekernit's a bootloader/monitor16:44
GitHub183[flickernoise] sbourdeauducq pushed 1 new commit to master: http://git.io/3VMiFw16:59
GitHub183[flickernoise/master] shutdown: simplify - Sebastien Bourdeauducq16:59
lekernelshould we display the compilation progress bar in simple mode?17:23
wolfspraulin my opinion - no17:25
wolfspraulbut it depends on how long the compilation takes, which I guess changes depending on how many patches we have preinstalled17:25
lekernelwell, it will show the user that there's a GUI behind17:25
lekernelgoes against the "$500 screensaver" ...17:26
lekernelotoh it can be inconvenient live17:26
wolfspraulmost people that see the GUI will then want it on a separate screen17:26
lekernelbut we can say that serious live users will use the configured mode17:26
wolfspraulok, the progress bar is really a detail, at least I don't care17:26
lekernelthere's only one screen, you use the configured mode after you have designed your patch and set up things ...17:27
wolfspraulideally the compilation is so fast that no progress bar is necessary :-)17:27
lekernellots of patches mean some compilation time, especially if we load them with images17:28
lekernelhave you tried my latest patches? you can do really nice stuff with images  :)17:28
wolfspraulhaven't tried yet17:29
GitHub191[flickernoise] sbourdeauducq pushed 2 new commits to master: http://git.io/anPFdw17:35
GitHub191[flickernoise/master] fb: send black screen instead of no signal during startup - Sebastien Bourdeauducq17:35
GitHub191[flickernoise/master] Display progress bar in simple mode - Sebastien Bourdeauducq17:35
AlarmIs there a wiki page sure the different software layers of the M1?17:53
AlarmIs there a wiki page about the different software layers of the M1?17:53
lekernelhttp://www.milkymist.org/wiki/index.php?title=Flickernoise_build_instructions may give you some ideas17:55
guyzmolekernel ?19:10
guyzmolekernel - FYI, I was at the Educatec convention last week, and there has been a couple of teachers (engineering school and university) that looked interested in the milkymist (I'll recontact them next week to see if it's for real...) and there has been a couple of A/V technicians that were interested as well.19:14
AlarmI am back. what I want is an overview of the software layers19:55
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-11272011-1909/20:00
xiangfuAlarm, it like, standby.fpg --> soc.fpg --> bios --> flickernoise.20:03
xiangfuAlarm, there is no such wiki page about software layers.20:03
GitHub139[scripts] xiangfu pushed 1 new commit to master: http://git.io/Njg7Gg20:04
GitHub139[scripts/master] don't reflash empty file to M1 - Xiangfu Liu20:04
Alarmrtems is included in what layer?20:06
xiangfuAlarm, flickernoise20:11
xiangfuAlarm, rtems is like libs link to flickernoise. it is different with linux kernel. :)20:12
Alarmok I understand better. Standby.fpg and soc.fpg files  are synthesized by the Xilinx tools, but the bios file?20:16
lars_is compiled by a C compiler20:18
Alarmsource code bios ? where ?20:20
lars_in the milkymist git repo20:21
lars_in the bios subfolder20:21
xiangfuAlarm, https://github.com/milkymist/milkymist/tree/master/software/bios20:21
Alarmxiangfu, ok thank you I have seen. Files whose extension is "raw"? What is it?20:30
xiangfuAlarm, you mean http://milkymist.org/updates/2011-07-13/splash-rescue.raw?20:31
xiangfuAlarm, it's just a picture. for splash screen.20:31
wpwrak(progress bar) i think M1 wants an OSD :) and yes, it should be on (at least to some degree) in "simple mode". simply to make sure people know it's there.20:31
wpwrak(osd) it should show controls settings and other patch-related status (e.g., audio level, interface use, etc.)20:32
wpwraki would get rid of the "simple mode" as something that looks good to the public completely. that's the "USD 500 screensaver" once more20:33
Alarmxiangfu, "raw" is an image format. I did not know. sorry.20:35
wpwrakAlarm: bios is completely from inside milkymist.git; flickernoise is from milkymist.git, rtems, rtems-yaffs, flickernoise.git, plus some more libraries i just took from the SDK but never bothered to build myself20:35
xiangfuAlarm, it's not a image format.20:35
wpwrakyou could say that "raw" is the format the bare metal uses. but as such it's a class of formats. basically all formats that are not "nice", with platform-independent codings, headers, etc.20:37
xiangfuwpwrak, thanks :)20:41
AlarmHow drivers keyboard and mouse are implemented?20:48
wpwrakAlarm: there's bit-level support in the SOC. that then passes bytes up to an embedded AVR (navre). that one does the packet processing and above (it implements the entire USB stack)20:50
xiangfuand https://github.com/milkymist/flickernoise/blob/master/src/input.c20:51
wpwrakAlarm: that navre then sends the keyboard or mouse data to the lm32, which can then run bios or flickernoise. in the case of flickernoise, you can see the last steps of keyboard/mouse processing in the input.c whose path xiangfu just sent20:52
xiangfuwpwrak, you mean it's like bit-leval-from-soc --> AVR --> back-to-soc --> bios/flickernoise?20:53
xiangfuI though it is bit-leval-from-soc --> AVR --> bios/flickernoise.20:54
wpwrakwell yes, there's soc ("hw") support for the mechanism with which avr and lm32 communicate. haven't looked at the details yet20:54
wpwrakbut it's not as if there fpga did much with the higher-level data. it just passes it on.20:55
xiangfuwpwrak, got it. thanks.20:55
wpwrakin the case of it's ---(unsynchronized bit-serial stream)--> SOC ---(bytes)--> AVR --(buttons and coordinates)--[soc]--> LM3220:56
xiangfuwpwrak, why the last is LM32?20:58
whitequarkUSB stack in an embedded AVR when you have an entire SOC? sorry, sounds like a "wtf" for me... I don't quite get the idea20:59
xiangfuwpwrak, I think the soc(the soc bitstream) is LM32.20:59
wpwraklekernel: btw, here's how to play with the lv3: cd wernermisc.git/midi2osc/; make; ./midi2osc `sed 's/#.*//' LV3` address_of_m120:59
wpwraklekernel: you can also give the option -d to get debug output, so you can see what gets sent20:59
wpwrakxiangfu: LM32 is the cpu core. i think the entire soc is called "milkymist soc"21:00
wpwrakwhitequark: i don't like it either21:00
xiangfuwpwrak, now. understand.21:00
wpwrakwhitequark: i think lekernel didn't want to have to mess with rtems and didn't want to implement a complete host controller (for which rtems would have drivers), so he implemented the little he needed in the avr21:01
wpwrakwhitequark: there may also be historical issues. e.g., if rtems usb wasn't usable when he started working on usb, it would have made sense to avoid it21:02
xiangfuwpwrak, on more question: ((unsynchronized bit-serial stream)--> SOC ---(bytes)) this is done in FPGA, right?21:02
wpwrakwhitequark: also, for moving bits around, the lm32 isn't much faster than the navre21:02
wpwrakxiangfu: correct, yes. the fpga reconstructs the clock, finds packet beginning/end in the incoming data, and retrieves bytes21:03
xiangfuwpwrak, and AVR code is running in bios. (memory) right? just want make sure.21:04
xiangfuwpwrak, so the soc.fpg include LM32 and other stuff. (like this one)21:04
wpwrakxiangfu: i think navre has its own memory. but bios and flickernoise can change the program there. they do this when initializing. after that, the avr is on its own21:05
wpwrakxiangfu: yes, soc.fpg contains the usb serial engine, the avr core of navre, the lm32 core, the graphics processor, the UARTs, and so on21:06
xiangfuwpwrak, that is a System On Chip. understand now. thanks21:08
Alarmwpwrak, When you speak of "AVR", you refer to this family of microcontroller?21:12
wpwrakthere navre is an AVR core = you can compile programs for it with avr-gcc, etc.21:15
AlarmAs Xiangfu said, we better understand what a system on chip21:19
wpwraksoc = core + peripherals, yes21:20
xiangfuwpwrak, Hi I just checked those pending patches: http://www.milkymist.org/wiki/index.php?title=Pending_RTEMS_PRs21:30
xiangfuwpwrak, those three patches have applied: http://dpaste.com/662456/21:31
Alarmmerci a tous21:45
wpwrakoh, wow. lemme check21:47
wpwrakexcellent. series updated. thanks !21:50
lekernelwpwrak: what do you want to do then? implement OHCI in Verilog? good luck!21:51
lekerneland the result will probably be bigger than LM32 itself21:51
lekernelas I said, USB is designed to be messy to implement in both hardware and software21:52
wpwraklekernel: dunno yet. maybe ohci. maybe a mix with navre. whatever works best :)21:52
lekernelat least the AVR solution reduces greatly the design cycles21:53
lekernelyou can implement OHCI on top of softusb21:53
wpwraki haven't looked t ohci yet. i looked at bit at uhci and it didn't strike me as overly complex. a bit too many layers in hardware for my taste, but nothing shocking21:53
lekernelor rather, most of it - they _also_ made a complicated interrupt controller21:53
wpwrakheh ;-)21:54
lekernelbut nothing impossible to handle21:54
wpwrakthere's always the option of simplifying things a little :)21:54
wpwrakafter all, we control the host interface driver as well. in the world of open source, we're omnipotent :)21:54
lekernelalso, having the USB stack on LM32 will likely cause other problems, e.g. consume CPU resources to process the 1000 MIDI messages per second that the LV3 sends21:54
lekernelwhile the Navre will easily filter those out in a not too dirty way21:55
wpwrakdidn't you once mention multicore as an option ? :)21:55
lekernelso you want to replace the AVR with another LM32?21:55
wpwrakand yes, devices needlessly hammering the core could admittedly be a problem21:56
wpwrakhehe :)21:56
wpwrakas long as the 2nd lm32 runs a stack we don't have to maintain, it's a win, no ? :)21:56
lekernelit's a dream21:56
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-11272011-2105/21:56
wpwrakif we have a softusb core in there, we could perhaps also let it act as a filter in some special cases. that still means that those other 10'000 cases with tricky over-designed protocol stacks on top of usb get handled by someone else's software21:59
lekernelmy position regarding Linux hasn't changed22:01
lekerneldemonstrate FN without regressions on it :-)22:01
wpwrakone step at a time :)22:01
wpwrakright now nobody is really working on linux, so lack of progress there isn't suprising22:02
wpwrakbug 1961 seems to move forward, too. very good. soon, i can scrap my patch stack :)22:03
mwallewpwrak: at least im working on linux. part time ;)22:04
wpwraklinux += 0.2; /*  ! ;-) */22:05
wpwrakwhile (rtems > linux) linux += 0.2;   /* let's see how many cycles this takes */22:05
mwalleqemu needs some love too22:10
wpwraki never quite dug qemu. for me it's either real hardware or cross-platform apis, so that i can run the thing i'm interested in natively on linux22:14
GitHub168[scripts] xiangfu pushed 1 new commit to master: http://git.io/VFBDjA22:17
GitHub168[scripts/master] compile-miplymist-firmware.sh: apply werner's patch in build host - Xiangfu Liu22:17
mwallei thinks its a handy debug/development tool :)22:17
GitHub33[flickernoise] sbourdeauducq pushed 1 new commit to master: http://git.io/mZD3YQ22:37
GitHub33[flickernoise/master] Always background DHCP and keep trying on failure - Sebastien Bourdeauducq22:37
lekernelwolfspraul: so, another source of boot slowness is gone...22:37
wpwrakwhee !22:37
lekernelgrmbl. the removal of standby has brought in an unexpected and unwelcome bug22:44
lekernelsometimes, shutting down puts the FPGA in some weird state... it seems it's configured, but in theory not with the SoC (it won't respond to GDB) and neither with standby (it should load SoC) ...22:45
lekernelwhat the hell is going on again22:46
lekernelof course, this happens intermittently, like all good time-wasting bugs ...22:47
rohi havent quite understood why there are so many different 'images' for the fpga at all22:48
lekernelloading soc from jtag when the board is in the locked up state boots, so I guess the standby bitstream fails to reboot the fpga to soc ...22:49
lekernelroh: because we want a backup bitstream22:50
lekernelthis is explained in http://www.xilinx.com/publications/xcellonline/22:50
rohhm. which article? i find nothing in release7722:52
rohchecking 76 now22:52
whitequarkwpwrak: the strange world of fpgas...23:15
--- Mon Nov 28 201100:00

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