#milkymist IRC log for Thursday, 2011-03-31

CIA-43milkymist: Michael Walle master * rafcb5f6 / cores/uart/rtl/uart_transceiver.v : uart: begin sampling with falling edge - http://bit.ly/ec3I4Z07:00
CIA-43milkymist: Michael Walle master * r9ff8deb / cores/uart/rtl/uart.v : uart: fix thru bit initialization - http://bit.ly/gExhuO07:00
CIA-43milkymist: Michael Walle master * rbaf9c09 / (cores/uart/rtl/uart.v cores/uart/rtl/uart_transceiver.v):07:00
CIA-43milkymist: uart: add break support07:00
CIA-43milkymist: This patch adds UART BREAK support. A BREAK signal violates the framing by07:00
CIA-43milkymist: transmitting a logical 0 for a period that is at least longer than one07:00
CIA-43milkymist: frame. - http://bit.ly/g622RM07:00
wolfsprauloops :-)07:18
wolfspraulxiangfu: after reflashing your binaries, I could not find any patches installed?07:20
wolfspraulare they somewhere or just no patches?07:20
xiangfuwolfspraul: no patches.07:20
wolfspraulI reflashed rejon's m1 with your binaries, but now it's actually less usable than before :-) What is the easiest way for rejon to get a bunch of patches onto his m1 now?07:20
xiangfuI copy those patches to sd card.07:20
wolfspraulwhen you create binaries, can you include the full latest set of patches?07:21
xiangfucopy patches to sd card.07:21
wolfspraulwhere are the patches? (url)07:21
xiangfuwolfspraul: hmm... not easy. there only 4MB for OS. (flickernoise)07:21
wolfspraulok but there is another 24 mb or so data partition, no?07:22
xiangfuwolfspraul: yes. working on that now.07:22
xiangfuwolfspraul: two method for now. 1 make  mkyaffs2image works with milkymist one.07:23
xiangfu2. dump the data partition from one device07:23
xiangfuI have tried those 2 method. both not working. :(07:23
lekernelwolfspraul: worst case you can erase the flash partition and upload the patches with FTP07:24
xiangfuwolfspraul: copy a png file named "wallpaper.png" . when mm1 boot it's automatic log (just fyi)07:24
wolfspraulwe need dumbed down step-by-step instructions for Jon ;-)07:24
wolfspraulxiangfu: log?07:24
wolfsprauldon't understand07:25
wolfspraulit sounds like the easiest is if Jon takes a small FAT formatted memory card, and copies the patches from that URL to the root folder07:25
wolfspraulthat should work?07:25
xiangfuyes. correct.07:25
xiangfualso copy a wallpaper.png to this memory card.07:25
wolfspraulah ok07:26
rejonwolfspraul die!07:26
wolfspraulI see the 4 folders of .fnp patches at that url, so looks good07:26
rejoni am pretty dumb though07:26
wolfspraulrejon: sorry I just didn't prepare to that point, so we only found out there are no patches this morning :-)07:27
wolfspraulwhich is great, nothing better than get a bunch of shitty reality right in my face :-)07:27
wolfspraulrejon: try the memory card path07:27
wolfspraulhopefully it will work07:27
rejonis it on the wiki07:27
wolfspraulsome people have reported problems with memory cards07:27
rejoni can't keep up with you smart people right now07:27
wolfsprauljust copy files!07:28
wolfspraul1. fat formatted memory card07:28
wolfspraul2. copy files from https://github.com/lekernel/flickernoise/tree/master/patches07:28
wolfspraul3. reboot m107:28
wolfspraulthen either it shows up or not :-)07:28
lekernelhm, for copying files you'll need the shell for now07:28
wolfspraulhe can just leave the card in the m1, no?07:29
wolfspraulalways run the patches from there, why not?07:29
lekernelyeah, too07:29
wolfspraullekernel: different question. how do you feel about changing the power supply circuit of m1 to something like some Arduinos have, with a regulator that supports 6-20V or so, fairly robust... ?07:31
wolfspraulnot for rc3, but I'm wondering how you feel about it in general, say for rc4 or so07:31
lekernelwaste of time07:31
wolfspraulyou wouldn't need to spend any time on it. so that means you have no problem with it?07:32
rejonbetter than frying your mm107:33
rejonjust showed us embassy guy milkymist07:33
lekernelno... except that you'd have two different PSU07:33
lekernelso far we have exactly 0 fried M1s because of power supply problems, so....07:34
wolfspraulyes there are some issues in switching, but they are all minor and if I believe this is not good (robust enough), then I rather switch earlier than later.07:35
xiangfuuploaded two wallpapers we talked about before.07:35
wolfspraullekernel: it sounds like you are not opposed to it, only that you feel 'waste of time'07:35
wolfspraulwhich I can actually understand from your perspective, don't get me wrong07:35
wolfspraulapart from this chat you will never hear about it anymore :-)07:35
wolfspraulbut there are stupid people like rejon and me in the world, we just plug in stuff and turn things on :-)07:36
wolfspraulunfortunately these stupid people have a lot of money and may very well buy m1, one day...07:36
wolfspraulthis is definitely not for rc3, I want to get rc3 done asap07:36
wolfspraulno worries07:36
rejonesp. if there is a camera power supply in that box07:37
wolfspraulrejon: and what did the embassy guy think?07:37
lekernelwell, also keep in mind that switching regulators are prone to EMI problems. we should go to CE/FCC tests again if you change it.07:38
wolfsprauldo they want to buy one to run in the lobby somewhere?07:38
wolfspraulor in their bar, if they have one :-)07:38
rejonthey love it, he kind of poked fun of the name slightly07:38
rejonwill try to support press stuff is their main help07:38
wolfspraulhe better not do that, we have Wikileaks documents about activities of embassy staff in Beijing07:38
rejonmeet n greet07:38
rejonyeah, 1st time to meet him07:38
rejonlike 6-7 ppl came07:39
rejonmostly firefox fans07:39
wolfspraulyou can always just tell them "BUY"07:39
rejonbloggers, etc07:39
wolfspraulI do support my customers07:39
wolfspraulit's ready for end users now07:39
wolfspraulI personally go there for updates, fixes, etc.07:39
rejonyeah, i did07:39
wolfspraulit's not a developer box07:39
wolfspraulok good07:39
rejonthey want to support anything we do in china07:39
rejonbut its pretty much limited07:39
wolfspraulemail me, and we are on track07:40
rejonyeah, i told them to buy mm107:40
wolfspraullekernel: yes sure, it's quite a bit of work.07:40
rejonand purify their internet too07:40
rejonhe will hook me up with others07:40
rejonspeculative shit07:40
wolfspraulbut without wanting to discredit your 'waste of time', the current power supply circuit is in no way 'end user ready'07:40
wolfspraulrejon is American, and I got americanized in over 10 years07:40
wolfspraulyou have to dumb it down to the point that you could never even believe07:40
rejonthanks for dumbing down for me07:41
wolfspraulyou cannot put a fork into the pre-made cake box, because the danger is too high that the customer would try to eat the fork and hurt themselves!07:41
wolfspraulthat kind of thing...07:41
rejonjust get sued07:41
wolfsprauloh sure07:42
lekernelhaving zener+polyswitch fuse will already reduce the risk of damage by a fair bit07:42
rejonfried mm1 is expensive for returns at this point07:42
wolfspraulyes, we do that in rc307:42
lekernelrejon: did you fry your board?07:42
rejonlekernel : i should try it07:42
rejonfreedom fries07:42
wolfsprauldon't worry, we will not bother you with this. I just wanted to see whether you had any strong reasons _AGAINST_ a more robust regulator.07:42
wolfspraulI understand where you are coming from.07:42
lekernelwolfspraul: you can try applying 20V to the protection circuit and see what happens07:42
wolfspraulwe had real WARS at Openmoko over this07:42
wolfspraulsome people believe in making more robust, other say that's butt stupid because it can never be totally robust07:43
lekernelthen decide whether it's worth it to work on that regulator that supports 20V07:43
wolfspraullike lightning strike or so :-)07:43
wolfspraulof course the two camps can fight forever because both have a lot of arguments07:43
wolfspraulwe don't need to repeat that07:43
wolfspraulI do understand you, but I may still go for some more robustness at some point.07:43
lekernelimo, with properly chosen zener/fuse, that circuit should protect the board at 20V07:43
wolfspraulwhen you don't notice it, and when it doesn't cause any delays :-)07:44
lekernelI'm not sure the regulator would give more robustness compared to zener+fuse07:44
wolfspraulok, even better07:44
wolfspraulI will look into that.07:44
lekerneljust ask Adam to supply 20V to the chosen zener+fuse (alone, not on the board) and measure the output voltage...07:44
wolfspraulok, will check07:46
wolfspraulrejon: you should get a projector like the one I have07:54
wolfspraulhow did you like it? in terms of effectiveness of demoing I mean...07:55
rejonwolfspraul great07:55
wolfspraulif I take a small power strip with me, all I need is one power plug, and not too bright room/environment, and I'm good to go07:55
rejonyeah, i'll look into07:55
rejonyep, great, all fits in that case07:55
wolfspraulat night will never be a problem07:55
wolfspraulof course it's not as bright as a real big projector07:56
wolfspraulbut if one of those is there, I can still use that07:56
wolfsprauland if not, that little thing is definitely better than a monitor, especially one I would have to lug around with me :-)07:56
wolfspraulI need to find some sort of micro power strip07:56
wolfspraulwill keep my eyes open07:57
wolfspraulwith projector, camera, m1 I already need 3 now07:57
lekernelrejon: ?08:02
rejonoh, sorry, dude from usgovt08:09
xiangfuthe "mkyaffs2image.c" in yaffs. code style is a little strange.09:21
wolfspraulrejon: cool, this is what I need for my m1 kit http://www.amazon.com/Monster-MP-OTG400-BK-Outlets/dp/B000F9YN2M10:08
rejonyeah, i need one of those10:35
rejonlekernel what issue tracker is being used on milkymist?10:40
rejonseems like code is a lot of places10:40
rejonhard to find on the site10:40
lekernelall code is on github, I don't know where you managed to find "lots of places"10:41
rejonlekernel ook, cool github is good10:41
lekernelexcept urjtag and qemu which are in mwalle's repos and/or upstream10:41
wpwrakwolfspraul: (power supply) if you can find a suitable up/down converter, you may even be able to use the existing supply (if it can deliver a bit more current than actually needed)12:32
wpwraklekernel: it's not only protection but also makes it easier to replace the power supply, e.g., if lost, forgotten, or broken12:33
lekernelsuitable power supplies are only a few dollars at digikey12:34
lekernel(and we can sell replacements)12:35
wpwraklekernel: having them shipped around the world can be crazily expensive. plus, what do you do if you need one on short notice ? e.g., because you've arrived at the club but your power supply is at the hotel room / at home ?12:36
wpwraklekernel: also, some places have local certification rules. so the one from digi-key might get confiscated by customs. (though in argentina, i only had them try to do this if the power supply was accompanied by something else)12:37
CIA-43flickernoise: Sebastien Bourdeauducq master * r403acd1 / (src/filedialog.c src/flash.c): File dialog box: add extension on save - http://bit.ly/f7rLxQ12:57
Fallenouxiangfu: indeed I think you cannot start an executable file from RTEMS shell13:04
Fallenouxiangfu: you can just run "commands"13:04
Fallenouso you have to put your program as a command13:04
xiangfuFallenou: it's not support on rtems. or only milkymist ?13:04
Fallenouon RTEMS AFAIK13:04
xiangfuFallenou: thanks13:05
FallenouRTEMS is like a library, that you link to your application13:05
Fallenouit's not like a real OS that loads programs13:05
Fallenouyou cannot load programs in RTEMS, it's the same in the shell13:05
Fallenoubut you can start "threads"13:06
Fallenouso if your program is already built-in13:06
Fallenouyou can start it, if it's a thread13:06
Fallenou(or a command, in rtems shell)13:06
Fallenouthere is no per-process memory mapping, no MMU support in rtems (and no mmu in lm32), no process support in rtems13:07
xiangfuFallenou: oh. yes. it's libs. not real os.13:07
Fallenourtems is basically a big library that does scheduling  among threads and interrupt management13:07
xiangfuFallenou: thanks again.13:12
Fallenouxiangfu: there has been a thread about dinamically loading executable in rtems (started by kristianpaul for his lua interpreter)13:13
Fallenouit could be done, using this http://www.slac.stanford.edu/~strauman/rtems/cexp/index.html13:13
Fallenoubut it's not easy I guess :)13:13
Fallenouand it's only supported on a few platforms13:13
Fallenouyou would have to port cexp on lm3213:14
Fallenouand play with gcc bugs13:14
lekernelthere doesn't seem to be GCC bugs affecting CExp if you stick to C (no C++)13:14
Fallenouxiangfu: http://www.rtems.org/pipermail/rtems-users/2011-February/007993.html13:15
lekernelthe lm32 c++ compiler is totally broken anyway13:15
Fallenouthis thread13:15
xiangfuwow. seems this task is very hard for me :)13:15
Fallenouso yes13:15
lekernelxiangfu: don't do it. it's not worth it13:15
Fallenouthe best is really doing a shell command :)13:15
xiangfuyes. sure. understand.13:16
lekernelif you want dynamic loading of code, use linux :)13:16
CIA-43flickernoise: Sebastien Bourdeauducq master * r35d905e / (src/sysconfig.c src/sysconfig.h src/sysettings.c): System settings: add language, resolution and wallpaper - http://bit.ly/hQKDcS14:23
CIA-43mtk: Sebastien Bourdeauducq master * re5e25e0 / (include/mtklib.h lib/background.c): New wallpaper API - http://bit.ly/g7eQWB14:54
CIA-43flickernoise: Sebastien Bourdeauducq master * rf006dce / (src/main.c src/sysconfig.c src/sysconfig.h): Wallpaper selection - http://bit.ly/g0hBPm14:54
CIA-43mtk: Sebastien Bourdeauducq master * r9a0ef32 / (lib/edit.c lib/entry.c): CtrlA selects all text - http://bit.ly/fLnK8215:13
carlobarhi, im tryng to run the lm32 processor and uclinux on a FPGA spartan 3. i know that the milkymist project is using RTEMS, but some time ago used uclinux... anyone know of a site to find uclinux developed by milkymist?15:27
kristianpaul(if you want dynamic loading of code, use linux :)) i agree15:57
kristianpaulhi carlobar :-)15:57
carlobarhi  kristianpaul15:58
carlobarthese linux need MMU, right?16:03
kristianpaullinux does16:04
kristianpauluclix not16:05
carlobari need uclinux, have you implemented it with the lm32 processor?16:09
kristianpauli think lekernel pointed to it already16:12
kristianpaulpointed you*16:12
kristianpaulcarlobar: what you need it for?16:13
kristianpauli will siguest not to try linux for a real-word app, unless you are willing to develop missing drivers16:13
kristianpaulrtems is the best suported so far :-)16:13
carlobar:), i just need to execute some testbenches and made some performance measurements... but it has to be uclinux, because i need to compare it with the leon3 processor runing on uclinux16:15
kristianpauli see16:25
kristianpauli dunno if lars_ may have an already init to share? :-)16:25
kristianpaulbut worth to ask,16:25
lars_lm has a example rootfs16:36
kristianpaulcarlobar: ^16:40
carlobarhi lars_,  can you send me the example rootfs?16:47
carlobarthe rootfs have to be coupled with a kernel and a bootloader? im a little confused..16:49
lars_carlobar: http://www.theobroma-systems.com/assets/downloads/mico32/lm32linux-20080206bin.tar.gz16:50
kristianpaulbroma :p16:56
carlobarthank you, i had seen it before (not in detail), but i read that it had problems: http://mailman.uclinux.org/pipermail/uclinux-dev/2010-February/002010.html16:58
carlobarso i was searchig a version that had corrected t16:59
carlobarthese errors17:00
lars_well its the only prebuild image afaik17:01
lars_and it will work17:03
carlobarok, im going to start to study it.. if i have problems i will be back seeking for help :).. thank you17:05
mwallelekernel: maybe you have a better idea, if the break for the gdb stub is received just after a uart tx character is enqueued, there is a problem with the irq and pending state of the (shared) uart22:50
mwalleatm im doing, blocking on "uart tx in progress" bit22:51
mwalle(2) reading uart tx IP bit22:51
mwalle(3) ack uart tx IP bit22:51
mwalle(4) gdb stub takes over uart22:52
mwalle(5) after gdb packet processing, eg leaving the stub, posting an uart tx irq it bit was set in (2)22:52
lekernelsounds good22:53
mwalleunfortunately this needs some more registers within the uart core and more resources on gdbstub :)22:53
lekernelis there a problem with doing that?22:53
lekernelwell, I guess it's only a dozen more LUTs or so22:54
lekernelthe rest of the design uses 18K :)22:54
mwalleotht i could defer the break until the tx transaction is finished and do a break after that22:54
mwalle6k is really small *fg*22:55
lekernelyeah or defer the break22:55
lekernelboth solutions seem good to me22:55
mwallemh ok, let me sleep on it ;)22:56
mwallehttp://sourceware.org/gdb/current/onlinedocs/gdb/Memory-Map-Format.html#Memory-Map-Format << that sucks!22:56
lekernellol, very useless utilization of xml indeed23:01
lekernelfor extra crapware points they should have used a java parser23:01
--- Fri Apr 1 201100:00

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