#milkymist IRC log for Thursday, 2011-03-10

Fallenouok :)00:02
Fallenoulekernel: why is the video echo 512x512 -> 1024x768 30fps 16bpp taking 900 Mbps ?00:08
Fallenoutaking transparency into account makes 24 bpp right ?00:08
lekernelvideo echo is translucent, so you need to read-modify-write the framebuffer00:08
lekernelno, everything is 16bpp00:09
FallenouI arrive at  720 Mbps00:09
Fallenouoho k00:09
lekerneland read the 512x512 framebuffer too00:09
Fallenouok got it00:11
lekernelwolfspraul: hi00:30
lekernelthe video input on my mm1 (which was the one we did the ESD tests with) just broke down on me. it detects a SECAM 525 signal even when there's nothing connected, and the picture is always black no matter what you do (didn't do other investigations atm but will check tomorrow)00:31
lekernelhave you seen anything like that already?00:31
wolfspraulI vaguely remember at one time (and only once) during rc2 testing, we had a case where something SECAM was detected.00:39
wolfspraulwe ran the test again and it went away, so we didn't investigate further00:39
wolfspraullet me try to grep the production testing log files, if I can fine them. one sec...00:39
wolfspraullekernel: you there?00:42
wolfspraulI have one AD_RESULT of 7, the others are 4.00:42
wolfspraulwhat does AD_RESULT mean again?00:43
wolfspraulaw_: good morning00:43
wolfspraulSebastien just posted a bug he ran into, let me repost...00:43
wolfspraul"the video input on my mm1 (which was the one we did the ESD tests with) just broke down on me. it detects a SECAM 525 signal even when there's  nothing connected, and the picture is always black no matter what you do (didn't do other investigations atm but will check tomorrow)00:43
wolfspraulI remember that one time during testing, there was something SECAM we ran into, right?00:44
aw_wolfspraul, good morning00:44
wolfspraulbut only once, and then never again, if I remember correctly00:44
aw_yes, it did show up one time. exactly correct.00:45
aw_your mm1 broke down? and now it detects always a SECAM 525 without video-in plugged?00:47
wolfspraulaw_: I grepped through the production test logs, I find one AD_RESULT of 7, the others are all 400:47
wolfspraulbtw, for the next run I want to do even more extensive logging, and we should archive 100% of the logs00:48
wolfspraulyou never know later what you might want to search in the logs :-)00:48
wolfspraulthe AD_RESULT of 7 is here http://downloads.qi-hardware.com/hardware/milkymist_one/production/rc2/test_results/14-results00:49
aw_yes. if more logs can be generated by test program, it always good info.00:49
aw_yes, i remember that one.00:49
wolfspraulmaybe lekernel is already sleeping. not much we can do right now, let's see what he finds tomorrow.00:50
wolfspraulmaybe another improvement/fix we can roll into rc3 :-)00:50
wolfspraulwhat does AD_RESULT 7 mean?00:50
wolfspraultwo more bits set (1 and 2), the others are all AD_RESULT 400:50
aw_p18 of video decoder 7181B has autodetect function to recognize what video format it is. 7 is the SECAM 525 format detected.00:55
wolfspraulwe gave sebastien feedback, now we need to see what he finds on his board00:55
wolfspraulaw_: watch out for more secam detections... Back then it was only once and went away, so we didn't pay more attention.00:56
wolfspraulmaybe we should list it on the 'rc2 known issues' list?00:56
wolfsprauleven if it's rare etc... ?00:56
aw_wolfspraul, okay let me add. btw the PAL-B/G/H/I/D is AD_RESULT 4.00:57
wolfspraulyes I see it now - thanks for ds link00:58
wolfsprauldo you have any idea why the chip auto-detects the wrong format?00:58
wolfspraulI'm curious to see whether the problem goes away when Sebastien reboots his m1, or whether it persists.00:59
aw_such problem may need to check video out format from the device he set firstly, i quite don't  know how it detected wrong.01:02
wolfspraulalright. let's just add it to 'known issues' and wait until we hear more from Sebastien.01:04
aw_if with a standard video pattern generation and given test more times via test program on his board probably can reproduce such wrong detection.01:04
aw_wolfspraul, added.01:13
wolfspraulperfect. we keep an eye on it.01:14
Fallenouthe slides for the OSHUG #801:18
Fallenoutomorrow (humm in a few hours)01:18
Fallenou30 minutes talk, it's gonna be quick :)01:18
wolfspraulFallenou: do you think there is anyone there who wants to buy an m1 board?01:18
FallenouI have really no idea01:19
wolfspraulespecially someone who seriously needs it and would start serious work with it right away? :-)01:19
Fallenoumaybe I will be able to tell you tomorrow after the talk01:19
wolfspraulthe reason I'm asking is that I only have 2 left now.01:19
wolfspraulalthough there may be some more at Bearstech, Tuxbrain, etc.01:19
wolfspraulwe work hard on rc3, but we may be out of stock for a number of weeks, which is bad :-) (I should never run out of stock)01:19
FallenouI will try my best to make them want one ;)01:19
wolfspraulyes, great!01:19
wolfspraullet's make sure those 2 go to really great guys01:20
Fallenouyes it would be goo to have feed backs01:20
wolfspraulthanks a lot for your help btw!01:20
Fallenouand contributions01:20
wolfsprauloh you bet01:20
Fallenouyou're welcome :)01:20
wolfspraulhow many people will come to that talk?01:21
wolfspraulis it broadcast in any way?01:21
wolfsprauldo you have milkymist stickers? (bit late too ask :-)) but maybe for next time...)01:21
wolfspraulnice PDF. what is the license of the pdf?01:23
Fallenouwhatever you want :p01:23
Fallenouit's a mix of several presentations lekernel did01:23
wolfspraulok so let's just say cc-by ?01:23
FallenouI tried to take the best of several slides he did01:23
Fallenouit's OK for me :)01:24
wolfspraulif possible, just add a little cc-by somewhere01:24
Fallenoushould I write it somewhere ?01:24
wolfspraulalso attribution - who are the authors01:24
Fallenouhum ok will do that tomorrow01:24
FallenouI doubt lekernel will sue me for using his slides for presenting his project :p01:25
wolfspraulit's enough you have a tiny "cc-by Sebastien Bourdeauducq & Yann Sionneau" somewhere01:25
wolfspraulno sure, of course not01:25
wolfspraulbut the problem spreads01:25
Fallenouok ok01:25
FallenouWill put that01:25
wolfspraulonce we are sloppy on marking the license SOMEWHERE, and attributing the authors, then later we will never clear it up anymore01:25
wolfspraulthen we have lots of content with unclear licensing and authorship - BAD01:25
wolfspraulbecause people will ask "who wrote it?" and we don't even know01:25
wolfspraulthey will use that to cast doubt over the whole project01:25
wolfspraulso we need to be a little diligent along the way01:26
Fallenousure I understand your concern01:26
Fallenouyou're totally right01:26
wolfsprauldo you have notebook stickers?01:26
Fallenouhumm no I have no stickers01:26
wolfspraulif not, and if you have another conference again one day, please email me your postal address and I'll send you a pack01:26
Fallenouno flyers01:26
Fallenouno board01:26
Fallenouok will do01:26
wolfspraullet's start, one by one01:26
wolfspraulemail me your postal address, and the stickers are on the way01:26
Fallenouok thanks I'm sending the e-mail right now01:28
wolfspraulgreat, thank you01:28
wolfspraulwhat software did you use to make that presentation? looks like beamer/tex?01:28
Fallenouyes it is01:28
wolfspraulif possible, remember to upload the original somewhere01:29
wolfspraulthen it's easier for others to remix later01:29
wolfspraulyou could even have a url to the original in the pdf itself01:29
Fallenousure I had big troubles remixing one of lekernel pdf01:29
Fallenousince the sources were lost01:29
wolfspraulI can imagine :-)01:29
wolfspraulit's always the same little things, isn't it? :-)01:29
Fallenoubut for the others it was OK01:29
Fallenouhehe yes :)01:29
wolfspraulput a URL to the tex original into the pdf, that would be perfect01:29
FallenouI'm going to sleep01:29
Fallenoui'm really tired01:30
wolfspraulgreat presentation01:30
wolfspraulgood luck!01:30
Fallenouit's 1:30 am in London atm :)01:30
Fallenouand i'm working tomorrow :p01:30
wolfspraulI'm sipping my morning coffee01:30
wolfspraulactually get another one now :-)01:30
xiangfuHi aw_01:43
aw_hi xiangfu01:43
xiangfu if the "D2" and "D3" 'shine' when plug the power.  what I should do?01:44
aw_at first power-up normally?01:44
xiangfuaw_: no.01:45
aw_but you described 'shine' when plug the power. or power on   for a while then off ...and on? or other?01:46
xiangfu1. the milkymist can power on normally. I can see the 'control panel'01:47
xiangfu2. I try to connect nanonote and milkymist by using UART. since they are both 3.3v01:48
xiangfu3. reboot milkymist01:48
xiangfu4. a lot of message output in VGA monitor. (cann't read, too fast)01:49
xiangfu5.milkymist one can not boot to 'control panel'01:49
xiangfu6. I poweroff the milkymist one. dis-connect the nanonote with milkymist01:49
xiangfu7. un-plug the milkymist power cable.01:50
aw_actions between 2. ) and 3). you power cycle the M1 board or clicking 'Reboot' button on 'control panel'?01:50
xiangfu8. plug again. the D3 and D2 shine. I can not boot milkymist now01:50
xiangfubetween 2) and 3). clicking 'Reboot' inside Flickernoise01:51
aw_hmm...good capture! That was my felt strange too.01:52
xiangfuhi aw_ by un-plug the power cable for awhile . the milkymist one boot again.01:52
aw_well...once you got no configuration condition. yes, exactly ...just waiting for a while ..then M1 is booted again.01:53
aw_'for a while' means that I quite don't know how long it should ...i even wait for days.01:54
wolfspraulaw_: do you think with our diode+reset ic fix the problem xiangfu just described would not appear anymore?01:54
wolfspraulxiangfu: don't wait for days :-) you can also try reflashing with jtag maybe?01:54
aw_yes, the reset + diode should fix this problem..01:54
wolfspraulAdam can try waiting for days because he has many boards, but that's a little extreme for a software guy who only has 1 board. :-)01:54
aw_but I still have the thing strange thing likely I posted here two days ago.01:55
aw_so I do really don't know why clicking 'Reboot' button will cause this?01:56
wolfspraulaw_: yes but I think that's a software bug, didn't see Sebastien's reponse though01:56
aw_this is a symbol that shows we still may ran into it again..I'll also watch out this such case!01:56
kristianpaulxiangfu: you can try "detect" command in urjtag and see if works (when the mm1 board dint powerup)02:09
xiangfukristianpaul: ok. will try that when next time cann't boot.02:09
Action: kristianpaul want a can't boot state02:10
kristianpaulSo far i just have the reboot bug02:10
wolfspraulkristianpaul: do you see the reboot bug?02:11
wolfspraulhow common is it? is it easy to make it go away?02:12
wolfspraulI suggest you don't try extensive testing around this bug, just stay away from it. We have damaged 2 boards trying to track this down with extensive/aggressive testing.02:12
wolfspraulthat may be a result of our testing, but still, no need for many people to dig around there at the same time.02:12
wolfspraulI'm quite confident rc3 will be much more robust in boot/reboot behavior, on the hardware side.02:13
aw_kristianpaul, yes..stay way from extensive testing.02:13
wolfspraulthen we need to see whether we can also make some improvements on the software side somewhere, but one by one.02:13
kristianpaulwolfspraul: i stay away, i just cut powercycle board and still working okay02:14
kristianpaulwolfspraul: also track not my interest, anway it hapen kinda ramdon ..02:15
wolfspraulI think it should all be safe and stable, no worries.02:15
wolfsprauljust don't play with power experiments, we are too remote to recover if several people around the world break their boards.02:15
wolfspraulit may take us several weeks/months to get everybody back up and running :-)02:15
wolfsprauldon't want to waste time02:15
wolfspraulwe have a really nice improvement coming in rc3 in this area, I think02:16
wolfsprauland the rc2 people know they are true pioneers :-)02:16
Action: kristianpaul just play with fpga for now 02:18
lekernelaw_: what is the "reboot" bug you are talking about?10:23
lekernelxiangfu: the "LED shining" bug should be fixed in rc3, for now just power cycle until it's gone10:24
lekernelxiangfu: what are the last messages that get displayed on the VGA monitor when you get the reboot bug?10:26
lekernelbtw, they get also sent to the serial port, so you can connect a computer to log everything10:26
aw_lekernel, from xiangfu's asking, his posts on step1 ~ 8 , i was thought that it's just no configuration only.10:32
lekernelit's only 6-8 that are related to the configuration problem10:34
aw_but from his steps 3 ~ 5, it's he clicked "Reboot" button from 'control panel'.10:34
lekernelif there are messages on the VGA monitor, the fpga obviously was configured10:34
aw_yes, i know10:34
aw_i don't know how it could happen from clicking "Reboot" only..then he can not enter control panel.10:35
wolfspraullekernel: did you see my answer regarding secam-525?10:35
wolfsprauldoes the secam problem persist through power cycles, or did you have it only once?10:36
lekerneland the problem persists10:36
lekernelit's permanently broken now10:36
wolfspraulsounds like we need to track this down and fix it before rc3 then.10:37
lekernelyup, otoh permanent problems are easier to track down10:37
wolfspraulok what do you plan to do now?10:37
wolfspraul(regarding secam-525)10:37
lekernelcheck solders, clock, and video chip reset sequence10:38
lekernelreplace the video chip10:38
lekernelI hope I have spares... checking10:38
lekernelhrm, no...used them for rc1 tests... i'll order some10:39
lekernelcorrection: I have one left :)10:40
lekerneldamn this video input circuit is such a problem magnet...10:42
lekernelwrong bits connected, reworked chips burning out, broken crystals, and now that... :/10:43
wolfspraulno problem. EVERY feature is a risk, source of pain, etc.10:44
wolfspraulMilkymist One is a really complex board, I think I knew beforehand what we got ourselves into :-)10:44
lekernelbtw, i'd be up for some directed ESD testing on that part10:44
wolfspraulat least on my side I'm all patient, one by one we fix the issues we find10:44
lekernelwhile it sample some video, zap it with the ESD gun, send radio interference, etc.10:45
lekernelwhen are the wpan boards ready for EMC tests? perhaps I could do both at the same time10:46
wolfspraullet's first focus on the particular secam-525 now, that sounds like a high-priority problem10:46
wolfspraulI'm not sure changing the video chip right away is the best approach. What if it works then? what do we learn?10:48
lekernelno, not right away10:49
wolfsprauldid you do anything noteworthy as the last step before the secam-525 problem started?10:54
lekernelno, I was just using it10:54
lekernelthen the picture from the video input looked like random data10:54
lekernelone second after everything froze (probably the video chip sent garbage data that tickled some bugs in the video input core or software)10:55
lekerneland after rebooting, it's always the secam525 problem10:55
wolfspraulwhile using, ok. how did you 'use' it? just let it run?10:56
lekernelwhich I did for dozens of hours before without running into this issue, btw10:56
wolfspraulaw_: do you have any ideas what Sebastien should check?10:57
wolfspraulseems he has a list already, so we just wait and see...10:57
aw_lekernel, do you have camera with output format can be set as NTSC then we see if decoder can detect it?11:20
aw_wolfspraul, before rc3, we knew that I can not place a new bigger smd xtal on bottom side, i can imagine that those parallel bits wires goes through nearby the  new place which will be re-routed surrounding.11:26
aw_the bottom side now is for placing metal shield instead.11:27
lekernelno, I don't have anything NTSC11:38
lekernelanyway the video in doesn't detect ANYTHING now11:38
lekernelit's stuck to secam525 no matter what you do11:38
lekerneland whether there's a signal or not11:38
wolfspraulaw_: what is your point about crystal?12:21
wolfspraulyes, new (stronger mechanical) crystal must be on top side12:21
aw_wolfspraul, the parallel bits wires routed is always the sources of radiated wave. So have you let video decoder worked while you did a pre-scan and test in German(llaboratory)?12:52
wolfspraulprobably not12:53
wolfspraulat least we didn't have a video camera inserted during the tests, so unless the video decoder is always working (on) anyway, then no12:53
lekernelthe video decoder chip should keep transmitting a blue picture, even without input12:54
aw_wolfspraul, that could be worse case if do it. Not sure, because the decoded pixel works is a pretty high frequency running unless s/w ported already.12:54
lekernelthat being said, I wouldn't be against some further EMC tests on it12:55
lekernelaw_: the decoder chip was supposed to be running during the tests we did12:55
aw_lekernel, good.12:55
lekernelbut we didn't check it kept performing OK12:56
wolfspraulaw_: do you think this could be related to the secam detection problem? are you saying we should change/improve the wire routing in rc3?12:56
lekernelthe pixel port is unrelated to the secam thing12:57
lekernelthe secam detection comes from the i2c bus, not the pixel port12:57
wolfspraulI'm a little surprised that lekernel's secam problem persists now.12:57
aw_so wolfspraul , so you think that I also need to check or find laboratory to test.12:57
wolfspraulroughly guessing if we are lucky it's just the crystal, which we plan to make stronger (hopefully) anyway12:57
lekernelthe only possibility is involving EM radiation from the pixel port messing up the video chip - but that would be surprising imo12:57
wolfspraulaw_: no lab, for what purpose?12:57
wolfspraullet's first see what lekernel finds. maybe he replaces the crystal and it's back to normal.12:58
aw_lekernel, no, i was not saying things related to secam problem, i just thought that if we enabled that pixels worked when we did EMC test. ;-)12:58
wolfspraulin which case we can probably just proceed with rc3 as planned.12:58
lekernelI already replaced the crystal once on this board, but left it upwards12:58
wolfspraulaw_: let's not throw unconnected things together now.12:58
wolfspraulI want to focus on the secam problem now.12:58
lekernelI did not handle it roughly, though, and the board was put in a case soon enough12:59
aw_wolfspraul, ot's okay on focusing on secam.12:59
wolfspraullekernel: sure, I am just speculating. I'm just saying if replacing the crystal fixes your secam problem, then we will probably continue with rc3 exactly as planned.12:59
aw_i just had have many fixes on video pixels work layouts while EMC test failed13:00
wolfspraulbut if not, then maybe we have another root cause to track down. or if we are worried that we don't understand the crystal failures, maybe we need to investigate the cause of that?13:00
wolfspraulaw_: please explain that more clearly.13:00
lekernelusually, crystals heat before breaking down electrically13:00
wolfspraulyou say you worked on products before, and they showed certain bugs, under EMC testing?13:00
lekerneland this one doesn't (I don't even think the video chip oscillator as enough power to break down a crystal)13:01
wolfspraullekernel did not do any EMC testing, he just ran his board, when after hours suddenly he video-in started failing.13:01
aw_wolfspraul, yes, i ran into EMC test failed on those bits wiring routes. Unfortunately current wires on the top side now. So if having a pre-teset for EMC is good. I know this is irrelevant to secam.13:05
Fallenouthat's fucking insane13:40
Fallenoujust because the newlib maintainer has the honesty to say "i'm not sure if I have all the licenses listed for my code"13:40
Fallenoufedora says "I can't be sure newlib is free software"13:40
FallenouI guess for big projects with a lot of contributors it's the same13:41
rohyes and no. there is a reason most projects only let you commit code with trace.13:53
rohas linux does.13:53
wolfspraulFallenou: I can understand Fedora's position.13:56
wolfspraulthe problem with the answer of the newlib maintainer is that it will be torn apart in a legal conflict.13:57
wolfspraulit's just unacceptable, the system doesn't work like this.13:57
wolfspraulthen he can also be super-honest and say "I don't give a damn about licenses, I'm a pirate"13:57
wolfspraulhe may still get some sympathy points, but his legal standing is in the bin13:58
wolfspraulso what he needs to do, if he cares about legal strength, is to work on those license unclarities. try to track down authors for code chunks that are copyrightable, remove code with unclear origins, etc.13:58
wolfspraulor give up on the legal stuff, and become a happy pirate. but stay out of courtrooms then :-)13:59
kristianpaullekernel: (video decoder chip should keep transmitting a blue picture) oh what? hmm in my control panel i just remenber black screen.. oh dear i need confirm that..14:11
Fallenouok :)14:12
Fallenouwolfspraul: cc-by-sa forbids commercial use of the work15:15
Fallenouit's restrictive isn't it ?15:15
Fallenouisn't it better to put the sides under BSD or GPL ?15:15
wolfspraulFallenou: absolutely not. Unfortunately the CC people have created a lot of damage in their lust of world domination, so many people associate anything cc with non-commercial.15:28
wolfspraulwe cannot fix that unless a completely new alternative to cc emerges15:29
wolfspraulbut unless there is a -nc (non-commercial) in the line somewhere, commercial use is definitely allowed.15:29
wolfspraulso cc-by, cc-by-sa - both fine15:29
wolfspraulallows commercial use, allows modified derivatives15:29
FallenouI ve read otherwise on their own web site15:30
wolfspraulyes but I understand your feedback, I think it's horrible too15:30
wolfspraulno, that must be a misunderstanding15:30
wolfspraulthey are good politicians, they try to make everybody feel at home :-)15:30
wolfspraulat times that makes their message a bit confusing15:30
Fallenouo sorry15:31
Fallenouas you said15:31
wolfspraulcc-by-sa allows commercial use, definitely15:31
Fallenouive read by-nc-sa15:31
wolfsprauland it allows derivatives15:31
Fallenoumy bad15:31
wolfspraulyes correct15:31
wolfspraulstay away from -nc, -nd15:31
wolfspraulthe slides under cc-by should be just fine15:32
wolfspraulor public domain :-)15:32
wolfspraulif you want copyleft add a -sa15:33
Fallenouis it OK for you ?15:36
Fallenouoops a typo in sebastien s name15:36
Fallenoulast slide15:37
wolfspraulwow perfect15:37
wolfspraullicense, attribution, source url15:37
wolfspraulcan't be any better, I think15:37
Fallenouok great15:38
Fallenouthanks for your help15:38
Fallenoutypo fixed15:38
wolfspraulthanks for your attention to detail15:38
wolfspraulnow we hope you have a large and open minded audience :-)15:38
Fallenouyes I hope :)15:38
wolfspraulor you had already?15:38
Fallenouit's in 2h3015:38
wolfspraulI will be sleeping soundly then :-)15:39
wolfspraulgood luck!15:39
Fallenouhehe gn815:39
wolfspraulone thing milkymist needs a lot are good story tellers15:39
wolfspraulbecause there are many ways to tell this story, but it still needs to be done15:39
Fallenoui'll try to make them dream15:39
Fallenoushould be easy, they like ardware and open source  hardware15:40
Fallenouand this is a good project15:40
FallenouI hope some of them will come on IRC and ask questions15:40
wolfspraulgood point, maybe the last slide should have a URL to the webchat of this channel15:41
Fallenouhumm I will put #milkymist on Freenode15:42
Fallenouthey should be able to come15:42
wolfspraulsure, that's good too15:42
wolfspraulmaybe also the mailing list? but if someone just has a quick question, #milkymist on Freenode should be listed as a way to get it answered15:43
wolfspraulthe webchat url is a bit long http://webchat.freenode.net/?randomnick=1&channels=milkymist15:43
FallenouI think a web url is not necessary15:43
Fallenouhackers are used to come on Freenode15:43
wolfspraulyou could leave out the randomnick15:43
wolfspraulyou never know who is there (well I don't know)15:44
wolfspraulso as a rule of thumb, it's not a bad idea to dumb down ways to communicate15:44
wolfspraulbecause otherwise you will never even find out whether there were others that couldn't reach you, for whatever reason15:44
Fallenoua big fat url is really ugly for te slides15:44
wolfspraulremember that someone who just sees this passing by will only give the project a very short time maybe15:45
wolfspraulat least that's how I am, if I'm honest15:45
wolfspraulimagine you read an article somewhere, about something interesting, but you hear it for the first time15:45
rohcosmetic foobar. hide it behind some shortener if you like.. i dont care15:45
wolfspraulyou click on some lick, quickly want to find an FAQ or About page, right?15:45
wolfspraulif you cannot find it - forget it :-)15:45
Fallenouyes sure15:45
wolfspraulthat's because you cannot know in advance whether something later turns out to capture your interest for years or not15:46
wolfspraulso everybody has a rather short cut-off time for unknown/new things, otherwise you are too slow looking at new things15:46
wolfspraulwhen I see the same new thing popping up a second time in another place, then I typically give it some more time to study15:46
wolfspraulanyway, good idea about irc channel, just list it there...15:47
wolfspraulin whatever way you think works :-)15:47
FallenouI will cut the apple in two15:48
FallenouI will put a shortened link15:48
kristianpaullekernel: (video decoder chip should keep transmitting a blue picture) oh what? hmm in my control panel i just remenber black screen for video.. oh dear i need confirm that..19:17
lekernelwell, that's what the adv7181 datasheet says19:18
lekernelbut tbh I have not tested that a lot19:19
kristianpaulah ok19:20
Action: kristianpaul will do some tests son with ccd/cmos video cameras19:20
rejonlekernel gsoc app completely sent btw19:53
lekernelcool, thanks20:09
kristianpaullekernel: what's the max data trasnfer rate for the usb core considering that is a slave?20:50
lekernelI guess you mean by reading from lm32... I have no idea, benchmark it20:52
lekernelit's certainly higher than the 12MBps20:53
lekernelUSB limit20:53
kristianpauloh nice20:59
mwalleomg gdb sources are such a mess21:42
Action: lekernel is not surprised21:44
Action: mwalle has just found out that gdb supports some kind of monitor roms21:45
mwallesome cli based ones ;)21:46
mwalleunfortunately gdb supports only predefined break in characters21:47
mwallectrl+c, break and break+g21:47
lekernelhaving only one or two characters for break is probably a bad idea21:48
lekernelwe may transfer binary data through the UART for booting21:48
lekernelso we'd need at least 4 characters i'd say, to be safe21:48
lekernelmaybe even 821:49
mwallebut that means we need to patch gdb21:49
lekernelyeah, but we can certainly get that patch merged... shouldn't be much of an issue, unless the code is so messy that adding a break sequence is hell21:50
mwalleshould be a small patch21:51
mwallebut what happens with the received characters?21:52
mwalleeverytime you break in the application will receive some garbage21:53
wolfspraullekernel: any news on the secam problem?21:54
lekernelno, had no time to look at it today21:54
mwallewhat do you think about disabling the breakin for serialboot?21:54
mwallethats the only time you transfer binary data isnt it?21:55
lekernelthat's one solution21:55
lekernelthe other is to make the UART hold the data until the break sequence is complete or it receives a character that is not part of it (in this case it releases everything to lm32)21:56
mwalleat uart speed?21:57
wolfspraulok, I suggest you look at it soon, because depending on the outcome, it could be on the critical path for rc3.21:57
lekernelor, we can modify the boot protocol so that the break characters are escaped21:57
wolfspraulI will think what I can do on my side in parallel.21:57
lekernelwolfspraul: sure don't worry21:58
lekerneli'm on it21:58
wolfspraulsure. I think about what I can do too, but keep me posted.21:58
mwallei like (1) and (3) :)21:58
mwalle(3) more than (1)21:59
mwalleso gn8, tomorrow is graduation ceremony22:00
lekernelsame here, let's go for (3)22:00
Action: kristianpaul hates GMT -522:05
wolfspraulFallenou: how was the talk?22:06
lekernelwolfspraul: actually, the video chip seems totally destroyed and no longer responds to anything22:10
lekernelmaybe it's just the crystal22:10
lekernelwith the test program I get22:11
lekernelDecoder probe:22:11
lekernelVIN: I2C bus initialized22:11
lekernelUnexpected register value: 0xff                                          [  FAILED  ]22:11
lekernelI try replacing the crystal now22:11
lekernelhmm, no still doesn't work with new crystal22:29
lekernelbtw we'll need a ground point to connect scope probes to on rc322:32
lekernelor maybe even several (better for high frequency signals)22:33
kristianpaulI think adam comment about some tests point for rc3somwhere..22:38
lekernelfunny, I get higher amplitude oscillations of the crystal on the non working board22:41
--- Fri Mar 11 201100:00

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