#milkymist IRC log for Sunday, 2011-11-13

mwallewpwrak | grmbl. when i build my own bios, all i get is a newline on the serial console :-( << wrong branch?00:00
wpwrakprobably. need to check the voltage levels, though00:00
mwallesounds like youre using the new branch (with the new serial driver) on an old (read stable) mm soc00:00
wpwrakmwalle: master ?00:00
wpwrakah. yet another compatibility barrier then :)00:01
mwallewpwrak: master only works on the latest bitstream00:01
wpwrakah yes, i see there's a fairly large diff00:03
wpwrakhmm. why am i not running into problems with FN, too ? there, i'm using master on the old soc without trouble00:04
lekernelfn only supports the "old" soc atm00:05
wpwrakaha !00:07
wpwraksneaky little versions :)00:07
lekernelomg cvs is so slow00:12
lekernelwhy do people still use that ...00:12
wpwrak;-))00:24
wpwrakand even better, it's slow _each_time_ you use it ;-)00:24
GitHub100[scripts] sbourdeauducq pushed 2 new commits to master: http://git.io/F3C4oA00:30
GitHub100[scripts/master] Use libpng 1.5.6 - Sebastien Bourdeauducq00:30
GitHub100[scripts/master] Merge branch 'master' of github.com:milkymist/scripts - Sebastien Bourdeauducq00:30
GitHub127[scripts] sbourdeauducq pushed 1 new commit to master: http://git.io/nof-5Q00:36
GitHub127[scripts/master] Use curl 7.22.0 - Sebastien Bourdeauducq00:36
GitHub124[flickernoise] sbourdeauducq pushed 8 new commits to stable_1.0: http://git.io/oARqXQ00:44
GitHub124[flickernoise/stable_1.0] RTEMS CVS head support - Sebastien Bourdeauducq00:44
GitHub124[flickernoise/stable_1.0] Use new YAFFS API - Sebastien Bourdeauducq00:44
GitHub124[flickernoise/stable_1.0] YAFFS: flush task - Sebastien Bourdeauducq00:44
lekernelok let's trash the old code :)00:44
GitHub145[flickernoise] sbourdeauducq pushed 1 new commit to stable_1.0: http://git.io/rlVNEg00:55
GitHub145[flickernoise/stable_1.0] Simple mode: reset timer on manual patch switch - Sebastien Bourdeauducq00:55
wpwrakoh, good !00:56
wpwrakthe LV3 can supposedly use 8 bit values for controls instead pf the usual 7. does it do this out of the box ?01:11
wolfspra1lwpwrak: did I understand that correctly with hhkb pro resetting your m1 now?01:11
wpwrakwolfspra1l: i still need to check the voltages. but it looks that way. perhaps oversized input caps.01:12
wpwrakwolfspra1l: btw, you wanna check the route of M1 on its way to faderfox ;-)01:12
wolfspra1lover the us?01:13
wpwrakmuch better ;-)01:13
wolfspra1lchecking, but... I trust the fedex computer system to know how to best keep their 700+ airplanes filled all the time :-)01:14
wpwrakyes, that's the impression it get, too. keep the planes filled, no matter how ;-)01:14
wolfspra1lah ok, asia route01:15
wolfspra1lwell that's the shorter one01:15
wpwrakand a night in paris :)01:15
wolfspra1lguangzhou is becoming a huge hub for fedex01:15
wpwrakwell, evening :)01:15
GitHub152[flickernoise] sbourdeauducq pushed 1 new commit to stable_1.0: http://git.io/rjSORA01:15
GitHub152[flickernoise/stable_1.0] filedialog: do not allow empty filenames - Sebastien Bourdeauducq01:15
GitHub132[flickernoise] sbourdeauducq pushed 1 new commit to master: http://git.io/R-nRXw01:16
GitHub132[flickernoise/master] filedialog: do not allow empty filenames - Sebastien Bourdeauducq01:16
wpwrakyeah, the asian part is funny but kinda makes sense. the detour, on the other hand ... ;-)01:16
wolfspra1lwhat detour?01:16
wolfspra1lit's almost direct now01:16
wpwrakkoeln -> paris -> koeln01:16
wolfspra1loh, interesting01:17
wolfspra1lmaybe a packing reason?01:17
wpwrakisn't it ? ;-))01:17
wpwrakhmm. maybe.01:17
wolfspra1lit may be packed in a larger container and be such a minority that the first time it's at Koeln it's not worth to open the larger one01:18
wpwrakyeah, that could be the case. or they just put it on the wrong pile :)01:18
wolfspra1lunlikely, but ok01:18
wolfspra1lbut yes, interesting case :-)01:19
wolfspra1lI've seen others where a package from the US to China first went to the Phillipines, then back up01:19
wolfspra1lthat's a thousands of miles detour01:19
wolfspra1lsome Taiwan notebook makers have explored sending notebooks (and other consumer electronics) to Europe by train01:20
wolfspra1lbut I think except for a few test trains they haven't really started using that yet01:20
wpwrakphew01:20
wolfspra1llots of problems, politically, geographically, tax-wise, delays, climate, many unknowns01:21
wpwrakindeed. by boat may be easier :)01:21
wolfspra1lI think this was about replacing air01:21
wolfspra1ltrain can make it in a week or so01:21
wolfspra1lof course boat is unbeatable in price01:21
wolfspra1lbut 4-6 weeks, and needs some scheduling to catch a boat etc.01:22
wolfspra1lthey thought they can establish a train route as an intermediate01:22
wpwrakyes, boats are slow01:22
wpwrakinteresting idea01:22
wolfspra1lhey, this is something supportive for PIONEERS...01:23
wolfspra1lone sec :-)01:23
wolfspra1lhere you go "The concept for what became Federal Express came to Fred Smith while he was studying at Yale University. For a class there, he submitted a paper which argued that in modern technological society time meant money more than ever before and with the advent of miniaturized electronic circuitry, very small components had become extremely valuable. He argued that the consumer society was becoming increasingly hungry for mass-produced electr01:25
wolfspra1lthat must have been around 197001:25
wolfspra1lnow the interesting part :-) "He submitted the paper to the professor teaching the course, who gave the paper the grade of "C". Despite the professor's opinion, Smith held on to the idea."01:26
wpwraki guess he got that right on all accounts :)01:26
wpwrakand the professor ... not quite so :)01:26
wpwrakkinda like "linux is obsolete" by a. tanenbaum ;-)01:26
wolfspra1lyes. so the story is... if you do something truly new, you cannot assume people will agree with you or support you, no matter how much you believe in your story and how good it may eventually be.01:27
wpwrak(train) i'd worry about theft. trains stop a lot. easy to open a container and lighten the load a bit. or just grab the entire wagon. with a ship or a plane, you can be pretty sure that what is aboard when they leave will still be there when they arrive.01:28
wpwrak(new things) yup. a lot of people simply refuse to think outside their box01:28
lekernelupdates available. please test01:29
wpwrakwolfspra1l: this site may also be a place to pimp the M1: http://createdigitalmusic.com/01:31
lekernelthere's also createdigitalmotion...01:34
lekernelI'm on it, it's just slow and messy01:35
wpwraklife is ;-)01:35
wolfspra1lwpwrak: thanks, added to links [createdigitalmusic]02:33
wolfspra1lthis list needs so much work, argh http://en.qi-hardware.com/wiki/Milkymist_One_marketing#Links_to_Shops.2C_Communities.2C_Companies.2C_Projects.2C_Technologies02:33
wpwraknot only the list ;-))02:39
wolfspra1lnot bad http://worrydream.com/ABriefRantOnTheFutureOfInteractionDesign/02:40
wpwraka day's read ;-)02:43
wpwraklekernel: when i look in my crystal ball, i see that you02:44
wpwrak're using a mouse that has a wheel :)02:44
wolfspra1lno no, not much text [lots of pics]. this is about controllers imho02:48
wolfspra1l"Is that so bad, to dump the tactile for the visual? Try this: close your eyes and tie your shoelaces. No problem at all, right? Now, how well do you think you could tie your shoes if your arm was asleep? Or even if your fingers were numb? When working with our hands, touch does the driving, and vision helps out from the back seat."02:50
wpwrakyes, i skimmed over the pictures. i get the message that tactile feedback is good :)02:52
wpwrakwhee. brownout ! but nothing crashed :)02:55
wpwrakheh, it seems a lot of fun could be had on the lm32 by loading a non-zero value into r0 :)02:59
rohre03:28
wpwrakhad a peek at USB hubs ... alas, not so transparent. the host has to query port status and actively enable ports05:13
wpwrakand another interesting discovery: the USB host only does low-speed. that explains a few more compatibility issues ...05:47
wpwrake.g., the HHKB is full speed05:49
wpwraknow let's see if i've done a good job at breaking USB ...06:16
wpwrakapparently yes06:18
rohheh.06:40
rohwhat are you trying to do? add usb-hub support?06:40
wpwrakno, support for my RF keyboard+mouse combo06:41
rohheh06:41
wpwraknow .. where did i make the stupid little mistake that breaks it ...06:51
wpwrakah, found one ... let's see what still work ...06:54
wpwrakah. the acme keyboard (non-RF) doesn't work because it has some weirdo 2nd interface the new code didn't like. fixing ...07:02
wpwraksolved07:04
wpwraknow ... seems that my RF kbd uses a funny report format. let's analyze it ...07:04
wpwrakwhoa ! now there's a rich report descriptor07:19
wpwrakhttp://pastebin.ca/209411907:20
wpwrakand this also explains why the bestie only moves in one direction - the values are 16 bit07:23
wpwrak.. apparently with LSB = 0. sweet. make it a little messier. because we can.07:41
wpwrakah no, the other way around. MSB = 0. and they added a report ID. that's why everything is shifted a little.07:45
wpwrakhehe, works :-)07:47
wpwraklekernel: how do you like the special-casing of my USB combo keyboard/mouse ? :)08:25
wpwrakbtw, i'm done battling with USB for now. so there's no risk of conflicting patches if you add usb-midi08:50
lekerneljust a crazy mess... and it never ends10:01
lekernelthat's why I stopped at supporting just the most basic devices10:01
lekernelbtw, full-speed supposedly works on the hw... it's just that navre doesn't enable it10:01
lekernellol @ the end of http://worrydream.com/Apple/10:14
lekernelthis website is very good at crashing X.org. I wonder if he's also making a statement here...10:25
GitHub33[milkymist] sbourdeauducq pushed 8 new commits to master: http://git.io/tbYdkg10:49
GitHub33[milkymist/master] softusb-input: also accept 3 byte reports, e.g., from wheelless mice - Werner Almesberger10:49
GitHub33[milkymist/master] softusb-input: use symbolic constants for packet ID values - Werner Almesberger10:49
GitHub33[milkymist/master] softusb: use macro to construct (address, EP) value - Werner Almesberger10:49
lekernelwolfspraul: reminds me of http://www.wisibility.com/post/2011/08/30/Inkling%2C-la-nouvelle-tuerie-de-Wacom%2C-entre-sketch-papier-et-num%C3%A9risation-automatique-! (en fran├žais)11:37
qi-botThe Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-11132011-1104/12:02
GitHub194[flickernoise] sbourdeauducq pushed 1 new commit to master: http://git.io/qbq96w12:43
GitHub194[flickernoise/master] simple mode: random delays - Sebastien Bourdeauducq12:43
kristianpaulhands -> body, actually i had seen lotsof performances driven/synced (dont know how) with sound and video (multiple walls)12:55
kristianpaulhttp://www.heinry.fr/olivier/index.php/post/2011/09/17/Fotograma-Completo-Principal-%C3%A0-Laboral-Centro-de-Arte-%28Xixon-aka-Gijon%2C-Espagne%2912:55
wpwraklekernel: (full-speed) okay, you'll need this for usb-midi :)15:14
mumptaire15:31
lekernelwpwrak: I'm also getting what looks like lost midi messages when I press many keys at once... we should perhaps add a bit of HW buffering and do the message separation in the driver16:20
GitHub16[linux-milkymist] larsclausen pushed 1421 new commits to master: http://git.io/9BoFUg16:20
GitHub16[linux-milkymist/master] "xfs: fix error handling for synchronous writes" revisited - Ajeet Yadav16:20
GitHub16[linux-milkymist/master] rtc: ep93xx: Fix 'rtc' may be used uninitialized warning - Axel Lin16:20
GitHub16[linux-milkymist/master] rtc: rtc-twl: Switch to using threaded irq - Ilkka Koskinen16:20
wpwrakwhat would you think of going one step more radical and moving the message separation into the FPGA ? so that it would push out words, not bytes16:21
lekernelnot sure this would really help16:26
lekerneland it would use resources, and we won't be able to share the code with the uart anymore16:26
GitHub14[flickernoise] sbourdeauducq pushed 1 new commit to master: http://git.io/j6ZX8A16:28
GitHub14[flickernoise/master] Patch mashup support - Sebastien Bourdeauducq16:28
wpwrakyes, i would get a bit more complicated. for the hw buffering, that would be a proper FIFO where you can loop until it's drained ?16:28
lekernelyes16:28
wpwrak(mashup) video ? :)16:28
wpwrak(fifo) okay, that should be almost as good as hw MIDI16:29
lekernelphew, I'm a terrible film maker16:29
wpwrak;-))16:29
lekernelalso, this is a quite experimental option. many patch combinations look very shaky and/or sluggish16:30
lekernela few are interesting though16:30
lekerneland you can also use the other patch as a temporary disturbance, by pressing the key for a short instant16:31
lekernelusing this feature is simple: when you have a performance set up, just press multiple keys at the same time16:31
lekernelit works with MIDI and the USB keyboard16:31
stekern(hw MIDI) parsing MIDI in HW is a pita, going with FIFO and sw parsing is the "right thing", my 2 cents16:33
wpwraksw MIDI is certainly easier to tweak. e.g., we also don't have the "running status" yet16:35
lekernelit's probably not worth it anyway, you don't need hardware acceleration to process a 31kbps stream16:38
lekernelanything that can be done in software should be done in software16:38
stekernI agree, I have done it, but the only reason was that I didn't have a cpu to do the processing16:39
GitHub3[milkymist] sbourdeauducq pushed 1 new commit to master: http://git.io/k10Qew16:50
GitHub3[milkymist/master] tools: allow overriding the C compiler for distributions like Fedora who can't package clang correctly - Sebastien Bourdeauducq16:50
wpwrakin gcc we trust ;-)16:51
lekernelI don't16:51
lekernelgcc should die16:51
wpwrakwhy ? it never let me down16:51
lars_lekernel: $(CC)?16:51
wpwrakaye, $(CC)16:52
lekernelGNU make sets $CC16:52
wpwrakso you override it. CC=clang16:53
wpwrakif someone overrides it on the command line, you still get the overridden value16:53
stekernthe problem with a fifo is that you don't know beforehand how long the message is going to be (IIRC), so you can't trigger an interrupt on a fifo level. one way to solve it would be to poll the fifo at an interval that gives decent latency.16:53
lekernelyes, but then you have to modify the script to run on distros who ship a broken clang16:53
wpwrakbut that happens in your case anyway ?16:54
lekernelstekern: you just interrupt as long as you have > 0 bytes in the FIFO... then the ISR reads the FIFO until it's empty16:55
wpwraki mean, if you override it, it will be overriden. if you don't, you default to clang. so that's the same, isn't it ? only that you're using a non-standard name16:55
stekernlekernel: sure, but that would be the same as not having a fifo at all (if your system isn't really slow)16:55
wpwrakstekern: maybe you could add a delay after exiting empty state. we don't care about RT anyway.16:56
lekernelnah, the purpose of the FIFO here is to solve the problem of multiple bytes arriving when the interrupts are disabled16:56
stekernah, ok. in that case it'd help16:56
lekernelwpwrak: if I use CC=clang, you have to modify the script16:56
lekernelwpwrak: if I use CC?=clang, it won't work because gmake sets CC16:57
wpwrakso what would i do if i want non-clang ?16:58
wpwraki still have to modify the script to set COMPILER=gcc, no ?16:58
lekernelwpwrak: if you still don't hate GCC, you can try fixing this bug: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=4689816:58
mwallelekernel: even with make CC=clang ?16:59
lekernelwpwrak: COMPILER=gcc make16:59
lars_lekernel: if you want clang as default add a entry to your make.conf16:59
lekerneland it defaults to clang16:59
lars_btw CC defaults to cc17:00
lekernelyes, which is gcc on all distros17:00
lars_so?17:00
wpwraklekernel: CC=whatever make  should still work17:01
mwallenope17:02
lekernelI know, but I want it to default to clang.17:02
mwallemake CC=gcc17:02
mwallenot CC=gcc make17:02
wpwrakokay. make -e then :-)17:02
lars_lekernel: but why?17:02
lars_why not use the systems default17:03
lars_that will work for almost everybody17:03
lekernelbecause I don't like GCC17:03
lekernelperiod :)17:03
mwallelekernel: so set CC=clang, override it with make CC=gcc17:03
mwallelol ignoring arguments seems to be common practice on the uboot mailinglist17:04
lars_lekernel: so change your /usr/bin/cc17:04
lekernellars_: no, but I also want to make people aware of GCC alternatives :)17:04
mwalleor ignoring the mail entirely, this is so frustrating :(17:04
lekernelthe fact that fedora screwed up their package is just unfortunate17:05
lekernelah, yes... that works17:05
GitHub17[milkymist] sbourdeauducq pushed 1 new commit to master: http://git.io/BCTe0Q17:06
GitHub17[milkymist/master] tools: use the CC variable - Sebastien Bourdeauducq17:06
wpwrakbtw, if you really want to do fancy things with variables, there's also the $(origin VAR) function in GNU make. but luckily, we don't even need that here :)17:07
wpwrakbug 46898: so C is okay, C++ isn't ?17:12
lekernelno, neither is17:12
lekernelC fails later on17:12
lekernelC++ is also broken on 4.5.x by the way, you have bugs like one method being called in place of one another17:13
wpwrakmaybe you need to restart that bug report. it goes back and forth a lot. not sure the intended recipient can make sense of it this way.17:13
mwallewpwrak: mh there is no lm32 maintainer anymore17:14
wpwrakyou could have stopped the sentence after "broken" ;-)17:14
mwalleso i guess nobody cares :)17:14
mwallewell there is one but thats a lattice guy17:14
mwallewhich wont even reply to mails17:15
wpwrakmwalle: now there's your chance to rise to fame :)17:15
mwallepft :b17:15
lekernelthe LM32 LLVM backend seems to have made some progress, been a while since I checked it out17:16
lekerneland such problems seem more unlikely to pop up in LLVM, which has a much clearer interface between the language frontend and the CPU backend17:18
lars_mwalle: hm, v3.1 broken again17:24
lars_mwalle: what were the symtons you had on 3.0 again?17:30
mwallelars_: null pointer after scheduling, iirc17:51
lars_hm, yes looks like i have the problem now17:52
mwallesorry dont remember the struct atm17:53
mwalleoffset was 30 or sth like that17:53
mwalleshould now generate a nice bus error ;)17:53
lars_i have offset 7et ffc17:53
lars_offset ffc17:53
mwalleffc?17:54
lars_it looks a bit as if register get messed up during process switching17:54
lars_or maybe not. the corruption always happens after sys_newfstatat17:57
mwallethe fault i had, seemed to be non deteministic17:58
lars_one the other hand it seems to be r11 which is corrupted, which is a bit suspicious18:01
lars_this is the code http://pastebin.com/V8hpVx7B r11 gets corrupted  inside the loop18:19
mwallebetween context switches?18:31
lars_hm... this could have been easier if i've noticed right from the start that it is not sys_newfstatat but sys_mmap :/19:03
lars_some usespace app seems to expect mmap not to fail19:05
wpwrakand the kernel incorrectly fails to expect that some user space applications incorrectly expect mmap not to fail ?20:32
lars_more or less ;)20:42
--- Mon Nov 14 201100:00

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