#milkymist IRC log for Saturday, 2011-09-10

kristianpaulxiangfu: when you boot uclinux on your M1 the VGA out is working?00:44
wpwrakgrmbl. i hate it when my final strictly pro forma test produces worrying data. in a scenario what is fully exposed to NOR corruption, so far, loop2 has been doing 1300 cycles without trouble. i've never had it do more than 700 this far ...02:23
wpwrak(loop2 is the ones that boots RTEMS but doesn't wait until FN is up)02:25
qi-botThe build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-openwrt.minimal-09102011-0858/07:53
xiangfukristianpaul: VGA not working. under Linux.08:18
errordeveloperlekernel: have you heard of this http://vrq.sourceforge.net/ ?12:11
lekernelerrordeveloper, no. interesting :-)13:37
lekernelhow to wait forever in verilog? like "wait;" in VHDL... does "wait(0);" work?13:37
lekernelit does :)13:39
errordeveloperI'm working on a script for PD rith now, just bumped into the vrq thing13:42
lekernelmh? how does it relate to vrq?13:42
errordeveloperthere no relation13:42
errordeveloperI was googling something and vrq came up somewhat :{]13:43
wpwrakand of course it's mainly galling that the NOR corruption seems to have vanished now, even when i remove locking. this means that there is some yet unexplained parameter that affects all this. (or i'm having just incredibly bad luck and i've spent the last days way out on the tail of the bell curve)14:16
wolfspra1lthat sounds depressing14:22
wpwrakquite :-( i made some small changes to the test setup, namely connected a scope to monitor the power and tightened a loose nut in labsw. maybe these shifted things out of the "problem range"14:28
wpwraktightening the nut may have shifted some cables. i now notice fewer failures to switch (caused by interference). so maybe it wasn't the regular power cycling that triggered the corruption by these very brief surges. intentionally producing these is next on the list.14:29
wpwrakat the moment, i'm testing with the scope removed (so DC IN isn't grounded). but that doesn't seem to have an effect. still need a few hundred more cycles, though.14:30
wpwrakif it's the brief surges that triggered the NOR corruption, i should be able to make it happen up to about 10x more often that my initial tests showed. but we'll see.14:31
wolfspra1lby now anything I say is this rumor-land, compared to your mountains of hard test data14:34
wolfspra1ls/is this/is just/14:34
wolfspra1lbut I still am dubious whether electrical issues can trigger a write14:35
wolfspra1lthat's a pure guess though14:35
wolfspra1lmaybe someone is just writing :-) (in software)14:35
wolfspra1lor is that ruled out already?14:35
wpwrakit could be just sw14:35
wolfspra1lthat's my uneducated lazy guess14:36
wpwraka possible failure scenario could be a 0x40 bit pattern present on D0-D7 plus some glitches on the write signal. that would provide the first write to enable word writing and then the second write to provide the actual data.14:37
wpwrak(you need two write cycles to actually program a word into the NOR)14:38
wpwrak1304 cycles and counting. next test due at ~2000 (which would be 3-4x the number of cycles i needed originally to see a corruption)14:42
wpwrakif the "quick surge" test doesn't do anything either, i'll go back to include rendering in the loop, i.e., the very beginning14:43
wpwrakif that still doesn't do anything, i'll check the lock bits again. need to see if there's a non-destructive test for this. (e.g., write 0xffff and just check the status)14:45
wpwrakof course, i've already made sure twice the lock bits aren't set, so i don't really expect to find something unusual at that step14:46
wpwrakif things still are weird, i think it's a for a bottle of wine. that followed by a nap often helps to boost out of the box thinking ;-)14:46
wpwrakwolfspra1l: btw, how are the launch preparations going ? did you already book the airtime on youtube ? ;-)14:50
wolfspra1lneed to discuss with Jon14:51
wolfspra1lright now I am working through a list of people who have expressed interest14:51
wpwraksort of the first wave ? while adam is assembling cases14:53
wolfspra1lyes, correct14:54
wolfspra1lfirst 'wave' maybe a little too much14:54
wolfspra1las expected most people bail once it comes to paying14:54
wolfspra1lthat's how I clear out my Inbox :-)14:54
wolfspra1lbut some don't, and we have shipped out units14:54
wolfspra1lAdam builds up inventory14:54
wpwrak(inventory) good. don't want to build up long lead times instants after the great announcement :)14:56
wpwrak(bail out) there's that great distance between the mouth and the wallet ...14:56
wolfspra1lyes I love how it clears my Inbox14:56
wolfspra1lfresh air14:57
wpwrakwhat's the serious vs. vapour ratio ?14:58
wolfspra1lratio again may be too much, I think I shipped out 2 units to paying customers :-)14:58
wolfspra1la few more are in process14:58
wolfspra1lsome definitely went MIA after receiving the mail that they can now order :-)14:59
wolfspra1lall fine14:59
wolfspra1lyes, launch still coming, asap14:59
wolfspra1lthat depends on journalists too, in fact mostly14:59
wpwrakwhat's the process with journalists ?15:05
wolfspra1lneed to email some and see what they say, coordinate with Jon15:06
wolfspra1lnext week probably15:06
wpwrakgood. so, the public launch would be in about one week ?15:06
wolfspra1lwe don't have that much control over it, it requires journalists to cooperate15:07
wolfspra1lbut yes, sure, it's getting close15:07
wpwrakdo you want them to just reproduce your press statement or do you plan to provide them with a hands-on experience ?15:08
wpwrakor, somewhere in the middle, do an interview ?15:08
wolfspra1lyour quality standards come from a time long gone :-)15:09
wolfspra1lalmost nobody writes about things they actually tried anymore15:09
wolfspra1lwhat you read are 'stories', imagined stuff :-)15:09
wpwraksome do :) depends on whom you're targetting15:09
wolfspra1lwith the NanoNote, we did have 2 or 3 stories from journalists that actually worked on their piece15:09
wolfspra1lthey bought a NanoNote, there was an email exchange with questions etc. and eventually the article15:10
wpwrakwow. quite a lot then.15:10
wolfspra1lbut the majority, and especially the 'launch', was not like this and I don't expect it for M115:10
wpwrakthe majority will probably already fail to get your "have to buy to play" rule15:11
wolfspra1lnah they have not time and no interest15:11
wpwrakof course, with M1 that gets even worse - on both sides15:11
wolfspra1lno. their time is far more valuable.15:11
wolfspra1lit's all made up.15:12
wolfspra1lactually I've wondered this myself15:12
wpwrak(time valuable) do they see it like that, too ?15:12
wolfspra1lhow can a journalist write day after day mostly on hearsay? crazy15:12
wolfspra1lthat's why the entire tech journalism is in the state it's in, deplorable imho15:12
wpwrakwelcome to corporate planet earth :)15:12
wolfspra1land who reads it... bored geeks15:12
wolfspra1lyou just go to engadget.com, and you go through, and I guarantee you, it's all hearsay, every last story15:13
wolfspra1lit's more like a 'reality show' :-)15:13
wolfspra1lwhich of course is also not about reality...15:13
wpwraka rarely venture to engadget ... :)15:13
wolfspra1lyes hard to read15:13
wolfspra1lthey should spice it up more15:14
wolfspra1lget some good screenwriters from actual reality shows, to add some drama15:14
wpwrakheh ;-)15:14
wpwraksome martial arts fighter from samsung against a nazghul from apple. that could be fun :)15:15
wolfspra1lwell, the raw material is tough15:15
wolfspra1llots of overweight middle-aged men15:15
wolfspra1lbut they are quite creative to turn the lamest raw ingredients to something thrilling on TV...15:16
wpwrakreminds me to have lunch :)15:16
wolfspra1lso? can anybody help the tech press?15:16
wolfspra1lprobably not15:16
wpwrakyou used TV and thrilling in the same sentence ? i'm impressed :)15:16
wolfspra1lbut anyway, we do need a launch, and we need some luck15:17
wpwrakyeah. let's hope there is poetic justice and all the hard work does pay off :)15:19
lekernelwolfspra1l, we shouldn't focus only on the tech press15:22
lekernelthe music/arts press is worth trying too15:22
rohif apple continues on that course they'll end up in simpsons again.15:23
wpwrakroh: do you think jobs has a simpsonphobia ?15:36
lekernelwolfspra1l, also, again: why don't you enable the sharism webshop again? it would take you 10 minutes, looks more professional, and you don't have to deal manually with people bailing out again15:36
rohnaaah. but its still entertaining if even in there somebody makes fun of em15:36
wpwraklekernel: agreed. a web shop that just says "closed" looks very very bad15:36
wpwrakmight as well board it up15:37
lekernelI have totally removed the link from milkymist.org/buy.html ... and said to email sales@15:37
rohchange the 'closed' to 'open soon'15:37
lekernelhoping this would last only a few days...15:38
wpwraklekernel: probably a good choice. that "OUT OF STOCK - GO AWAY" is quite anti-marketing ...15:38
lekernelyes, it's less bad15:38
wpwrakit' basically says "we don't have it. we don't have any plans either. we don't want your questions either. we don't even know if we're in business anymore."15:39
kristianpaulrtems internals had changed a bit since last git pull :-)19:35
kristianpaulhe, finally rtems_nfsfs_initialize is  rtems_nfs_initialize, wich make more sense :-)19:58
kristianpauloh, yaffs had changed too including a flush task? ..20:03
kristianpaulthe only feature needs a RTEMS_FILESYSTEM_READ_WRITE is telnet user/paswwd to be stored, and wallpaper preference i guess too20:07
larsclekernel: hm, so how does the lm32 interrupt controller work? i send a trigger pulse to mark an interrupt as active and get an ack pulse back, when the interrupt is cleared?21:44
mwallelarsc: ack pulse?21:50
larscmwalle: there doesn't seem to be one21:51
mwallelarsc: who needs an ack anyway?21:51
larsclooking at the source code, it looks as if the interrupts are level triggered21:53
mwallemm/bios sourcecode or lm32 verilog code?21:53
larscverilog code21:53
mwallethe cpu interrupt line is latched until the cpu actually executes the interrupt iirc21:54
mwalleand is then cleared21:54
mwallehave a look at my serial break interrupt logic21:55
larsci meant the signals which go into the interrupt controller21:56
mwalleIP stays asserted until you ack it, if any(IP & IM) => cpu_interrupt21:58
mwalleIE is cleared and saved in BIE/EIE if interrupt is executed21:59
larscbut IP will also stay asserted if the source has not been cleared22:00
mwalleso you can either use pulses or levels, in the latter case you have to make sure to ack your core before you ack IP22:01
larsci was just confused, because the way i understood it in the discussion a few weeks ago was that it was edge triggered22:02
mwalleno its actually level triggered ;)22:03
larscwhich makes things easier22:05
kristianpaulargg obj/main.o:(.data+0x148): undefined reference to `shellext'23:09
wpwrakedge triggered is if you want to live on the edge - permanently ;-)23:26
larscit is usefull for certain external signals, but not really for internal signals23:50
wpwrak(external) you mean where there's no means to ack ?23:51
larsclike buttons23:51
wpwrakheh :)23:51
wpwrakbuttons are tricky. need rather lavish preventions against all sorts of upsets if your environment is just hostile enough.23:52
larscwe can do debouncing in software23:53
wpwraksometimes :)23:58
wpwraksome of the bounces are of a different nature, though ...23:59
wpwrak(i.e., interference)23:59
--- Sun Sep 11 201100:00

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