<rjeffries_>
tuxbrain tuxbrain_away re WPAN shield for arduino: there are a few 802.15.4 boards already. e.g. Zigbee protocol uses same radio as 6LoWPAN so may not need to develop hardware
<wpwrak>
Jay7: i think that's the kind of test that probes certain cosmological constants :)
<wpwrak>
tuxbrain: so you're considering to fab the wpan stuff ?
<wolfspraul>
wpwrak: we are thinking how to speed up, and maybe with David's new UBB production experience I can outsource to David :-)
<wolfspraul>
just thinking right now
<wpwrak>
oh, i very much like the concept :)
<wolfspraul>
Adam is squeezed hard to make a great milkymist rc3 run
<wpwrak>
location independence is one of the great things open hardware should give us
<wolfspraul>
totally
<wolfspraul>
I only have 2 Milkymist One left
<wolfspraul>
so if another two daring pioneers come along, I'm out of stock
<wpwrak>
in fact, i hate it each time when someone says "wolfgang/sharism should do this or that"
<wpwrak>
(mm1) nice :)
<wolfspraul>
I think most people get the copyleft hardware concepts...
<wolfspraul>
it's not that hard, one just has to think about copyleft a little :-)
<wpwrak>
how many bens do we have to get rid of ?
<wpwrak>
(adam) would be good if he could at least check if there are any obvious production issues. also, if tuxbrain produces them, he'll need a way to flash the firmware
<wpwrak>
the dfu boot loader isn't ready yet. the only approach for flashing anything so far is via the atusb-ben cable
<wpwrak>
also, in the future, fab flashing will need that cable
<wpwrak>
if you want to outsource it to tuxbrain, would adam send his devices to spain ?
<wpwrak>
minimum requirement for production is the atusb-ben cables (well, one that works. two to be sure there's at least one)
<wpwrak>
production testing can also be bootstrapped from a production run, particularly if out aim is "prototype quality"
<wpwrak>
s/out/our/
<wolfspraul>
of course Adam can send anything to Spain, I need to send a few things anyway
<wolfspraul>
that's no problem at all
<wolfspraul>
just consider the two tightly linked :-)
<wpwrak>
which two ?
<wolfspraul>
Tuxbrain and Adam
<kyak>
viric: (memory usage) could be intersting, is it something different from ps aux?
<wpwrak>
tuxbrain: what i don't know yet is the kicad-to-smt process. adam should have that (from the mm1 jtag)
<wpwrak>
wolfspraul: ah, okay :)
<wolfspraul>
yes and he was writing some wiki pages about it, although they are wip
<kyak>
xiangfu: could we please rebase again, there was a compilation fix backported? :)
<wpwrak>
very good :)
<wpwrak>
tuxbrain's kicad-to-pcb was extremely painless. that's good. they politely ignored a few constraints but that didn't cause any real troubles.
<xiangfu>
kyak: yes. sure.
<rjeffries_>
David's experience with fisrt run of pcbs for UBB was fine. does he also have a goor vendor for assembly?
<tuxbrain>
wpwrak: what constrains?
<tuxbrain>
wpwrak wolfspraul: (at* prodution) well at least let's chek if is economically feasible, pcb part seems quite good quality/price ratio, now is time to see if smt is also that good, as you know I have NFI on such procces and what I can learn reading maybe is totally different from what the manufacturer expects :P
<tuxbrain>
rjeffries: yes I have a pair of vendors to ask with good reputation on quality, now let's check price :)
<rjeffries_>
nods
<wpwrak>
tuxbrain: the constraints on there to put the tabs (the stuff you had to remove). also, i don't know if they checked with you than tin/hasl is okay instead of enig finish
<wpwrak>
tuxbrain: for me, that's the kind of things to remember for occasions when it actually matters. with ubb, it wasn't an issue. if you were making 100kunits, for example the cleanup it would be.
<kyak>
xiangfu: btw. I tried to follow your instructions in wiki for rebasing, they don't quite work. I made some minor corrections. I followed the steps, but wasn't brave enought to push -f :)
<kyak>
xiangfu: there were no confilicts btw, i guess due to the fact that they have been alreadu resolved by previous rebases?
<xiangfu>
yes. when some confilicts resolved . will not come up again.
<rjeffries_>
wpwrak i was wondering about that issue. with the pcd house that quoted UBB for me, we made quite a big deal re: requirement to carefully machine or laser cut UBB
<kyak>
xiangfu:Â Â maybe we could also do rebasing automatically in case there are no conflicts? i.e. daily?
<rjeffries_>
wpwrak I read your c.v. Mr DiffServ
<kyak>
and if there are conflicts, no rebasing
<xiangfu>
kyak: hmm...
<xiangfu>
rebase daily, scare me a little. I am not sure about this. I also worry about  broken the git history. lose some commits
<kyak>
xiangfu: ok, i don't insist. Maybe it's best done as manual job, i agree
<xiangfu>
kyak: maybe we should keep the 'data' folder to another repo.
<xiangfu>
most of the commit is about 'data' folder.
<wpwrak>
hm, very strange. in general, minicom ought to work (if you turn off hardware flow control and all that)
<xiangfu>
PC --- nanonote works fine
<wpwrak>
if you have a scope, it would be useful to look at the signals
<xiangfu>
PC ---- milkymist works fine
<xiangfu>
nanonteo --- milkymist . there are some problem.
<xiangfu>
wpwrak: oh, yes. don't have that. also don't know how to use that. :( sorry
<xiangfu>
and not easy to borrow that for me.
<wpwrak>
xiangfu: if you say "then nanonote became very un-stable. very easy reboot.", what does that mean ?
<wpwrak>
xiangfu: (scope) oh, you really ought to get one. rigol makes nice scopes. and they're not very expensive.
<xiangfu>
wpwrak: 1. run "stty -F /dev/ttyS0 raw 115200; while : ; do cat /dev/ttyS0 ; done &Â Â cat > /dev/ttyS0"
<xiangfu>
then if I input slowly. everything works fine.
<xiangfu>
if I press 'enter' five times and a little faster. the nanonote have kernel panic. then reboot
<wpwrak>
wow
<xiangfu>
wpwrak: (rigol) search the price now :)
<wpwrak>
(kernel panic) do you know where it panics ?
<wpwrak>
(rigol) xxxxC. xxxxCD. xxxxE, and xxxxD are good
<wpwrak>
(rigol) xxxxCA and xxxxB are less fun ;)
<wpwrak>
(rigol) xxxxC/CD/D/E have a lot of memory
<wpwrak>
oh, and with xxxx i really mean 1xxx. there's also a 6xxx (?) series. they're top-end and expensive. they may be very good, but i don't have any real-life information about them.
<kyak>
xiangfu: thanks :) and yeah, it's a good idea to move data/ somewhere...
<xiangfu>
wpwrak: (kernel panic) not record that, I will try to catch the message next time.
<kyak>
xiangfu: hm, i noticed there are problems building plplot on buildhost.. Strange, it built fine for me
<xiangfu>
kyak: (data folder). what do you think. create a individual repo or add it to openwrt-package.git?
<kyak>
xiangfu: individual repo sounds better, because openwrt-packages will only show up under the "feeds/"?
<kyak>
xiangfu: but i'm not sure how we can have files from two git repos in one working directory?
<xiangfu>
kyak: (show up in feeds/) we need create symlins anyway. ln -s data/qi_lb60/files or ln -s feeds/qipackage/nanonte-openwrt-files/data/file
<xiangfu>
cp data/qi_lb60/config.full_system or cp feeds/qipakcages/nanonte-openwrt-files/data/qi_lb60/config.full_system
<xiangfu>
that mean . 1. needs run make package/symlinks 2. setup the symlinks to files/Â Â 3. copy the config.full_system 4 run make pakcage/symlinks again.
<xiangfu>
the individual repo: for me it needs git clone twice. 1. openwrt-xburst.git 2. nanonote-openwrt-files.git
<xiangfu>
the openwrt-package.git repo: it just not a package.
<wolfspraul>
xiangfu: keep data/ in a separate repo sounds right to me
<xiangfu>
wolfspraul: kyak  ok. I will create a new repo. name: openwrt-nanonote-data.
<kyak>
xiangfu: how it will work with individual repo/
<kyak>
?
<xiangfu>
kyak: manually . git clone twice.  run 'ln -s ' manually
<kyak>
i mean, we will git clone in openwrt-xburst/openwrt-nanonote-data and then ln -s openwrt-nanonote-data data?
<kyak>
hm....
<kyak>
so it means cloning three times
<kyak>
counting the openwrt-packages
<kyak>
can we setup git to clone remote automatically
<kyak>
?
<kyak>
so we only have to update openwrt-xburst git, and then it wil update openwrt-nanonote-data git automatically?
<xiangfu>
(write a script file :)
<xiangfu>
the 'git update'' can be 'git pull' 'git fetch' ...
<xiangfu>
I dont' know how to hook git update
<kyak>
maybe the idea to put data/ in openwrt-pacakges is not so bad after all
<kyak>
then we only hve to install one symlink
<xiangfu>
I even found we can create a 'Makefile' under data/
<xiangfu>
like :
<xiangfu>
all:
<kyak>
he, that's a great idea :P_
<xiangfu>
cd ../../../; ln -s  feeds/qipackage/data/qi_lb60/files)
<xiangfu>
wolfspraul: kyak I think put it to openwrt-packages.git is better for me. (not makefile, just mv the data/ to openwrt-package.git/nanonote-files/)
<xiangfu>
1. files/* : those are files that overwrite the openwrt default files.
<kyak>
i would agree, since there is no good way to handle another git repo from within the git repo
<xiangfu>
we can create a package for 'data/' then we can using 'opkg' update those overwrite files
<xiangfu>
2. keeps config.* and build reflash_ben.sh scripts file in nanonote is not a bad idea
<xMff>
for reflashing the nanonote itself?
<xiangfu>
xMff: no. this reflash_ben.sh is running in host . since there is no usb-host in nanonte. we can not run 'usbboot' inside nanonote.
<xiangfu>
xMff: ( is there is a usb-host. we can using one nanonote flash another nanonote. :P )
<xMff>
xiangfu: I just wondered about "keep reflash_ben.sh in nanonote"
<Jay7>
it should be possible by nandwrite/mtdwrite
<xiangfu>
Jay7: there is one command 'mtd.nn' can flash kernel and rootfs but not u-boot
<Jay7>
is u-boot stored somewhere out of NAND?
<kyak>
xiangfu: have you checked if kernel/rootfs flashing is faster or slower from Ben or from PC?
<xiangfu>
Jay7: no. when JZ4720 boot. it load first 8KB of nand. but it assume that the nand is 2KB. so when we flash the u-boot we have to write 2KBdata-2KBecc. PAD to 4KB; ...
<xiangfu>
xMff: sorry for confuse
<xiangfu>
Jay7: do you know is there is  a software can calculate the ECC just like soc does?
<xiangfu>
kyak: reflash rootfs is much faster
<Jay7>
xiangfu: hm.. no.. can you ask Ingenic? :)
<Jay7>
but u-boot updating is relatively rare operation
<xiangfu>
kyak: ok. let's move the data to openwrt-package.git
<viric>
kyak: I've just built a kernel for the nanonote with the proper options
<kyak>
viric: yeah, the # CONFIG_PROC_PAGE_MONITOR is not set
<kyak>
i will have a look at memstat later
<kyak>
viric: does CONFIG_PROC_PAGE_MONITOR add any overhead?
<kyak>
to the running kernel
<viric>
kyak: in the linux world, unfortunately, is very hard to tell
<kyak>
heh, ok
<viric>
kyak: did you try memstat on a more powerful linux than the nanonote?
<kyak>
viric: hm nope, should i? I thought we wanted to compare glibc/uclibc?
<viric>
kyak: well, that's one thing
<viric>
but I wrote it also to help me in the PC :)
<kyak>
viric: hm, where do i download the source?
<kyak>
the web interface doesn't look trivial :)
<viric>
:)
<viric>
login anonymously
<viric>
last trunk revision -> zip archive
<viric>
I should put some link...
<viric>
link written
<kyak>
ok, it's running
<kyak>
now i need to read to know the meaning of columns :)
<kyak>
95008Â Â 73866Â Â 218212Â Â 114084Â Â 89203Â Â 12700Â Â TOTAL
<kyak>
running the openwrt build at the moment..
<viric>
'-h' tells you more
<viric>
Some totals (sums of the columns) are not helpful, like the sum of RSS
<viric>
but they are there. Maybe I should take them away.
<kyak>
heh, do you think "Sum of virtual size of anonymous pages" is any more helpful? :) I have to google every word
<kyak>
the comparison between processes is interesting
<kyak>
somehow ircd process is "lighter" than eggdrop
<viric>
:D
<kristianpaul>
wpwrak: about content i think the is a small issues about a missing \n but i dont see what it should reboot the ben
<kristianpaul>
I'll take a look tomorrow that i'm at home with the scope :-)
<viric>
kristianpaul: you have to go *home* for a oscilloscope?
<viric>
what a world... Most people I know has to go *to the office* for an oscilloscope :)
<kristianpaul>
viric: yeap, at work/office there is _nothing_, well there a soldering station  wich i'm not allowed to use and a multimtere somwhere ;-)
<wpwrak>
kristianpaul: is that ben-mm1 conflict xiangfu reported also something you're experiencing ?
<kristianpaul>
wpwrak: not that i can remenber
<wpwrak>
kristianpaul: okay, let's see what xiangfu finds. it's a weird one for sure
<LunohoD>
do you participate in GSoC?
<kristianpaul>
who?
<LunohoD>
qi-hardware
<LunohoD>
the nanonote community
<kristianpaul>
well.. you should back that question to all the software projects around it
<kristianpaul>
openwrt, linux, oe....
<viric>
who is here working in the mylkimist?
<kristianpaul>
working how?
<viric>
who writes the verilog for the spartan6 there?
<wolfspraul>
viric: the core milkymist soc channel on freenode is #milkymist
<viric>
ahh
<kristianpaul>
:-)
<wolfspraul>
of course you are welcome to ask milkymist related questions here as well, since we (as copyleft hardware manufacturers) manufacture this thing
<viric>
I thought someone was here too.
<wolfspraul>
yes, there is a big overlap
<kristianpaul>
i'm also larsc
<wolfspraul>
but qi-hardware is a copyleft hardware project
<viric>
kristianpaul: you are larsc?
<wolfspraul>
wherease milkymist is an SoC IC design project
<kristianpaul>
also tuxbrain
<wolfspraul>
milkymist one is a vj station built around the milkymist soc running in an fpga
<viric>
I know I know
<wolfspraul>
so it's all connected together :-)
<kristianpaul>
viric: i'm not larsc
<viric>
don't they plan to have a general purpose CPU in there too?
<kristianpaul>
sorry i mean, larsc and me are also in the #milkymist
<viric>
that could run linux, I mean
<kristianpaul>
that can be an addon
<wolfspraul>
you can already boot Linux on Milkymist One
<kristianpaul>
no need to change cpu
<viric>
wolfspraul: but that's an ingenic processor running it
<wolfspraul>
no
<viric>
ah no?
<wolfspraul>
Milkymist is an SoC, it uses a free LatticeMico32 core for 'general purpose' instructions
<viric>
aaaah ok
<kristianpaul>
navre core for USB stack
<kristianpaul>
and more of free space on fpga for your own stuff if needed ;-)
<wolfspraul>
though not all (by far not) peripherals of the milkymist one are supported in Linux right now
<wolfspraul>
also the Milkymist SoC has no MMU right now, which is probably a good thing to have on a Linux system
<wolfspraul>
but Linux boots
<kristianpaul>
uclinux
<viric>
wolfspraul: I've never used a linux without mmu
<viric>
cat ~/uImage > mtdblock1
<viric>
Is this a good idea?
<viric>
I hope so... let's see.
<viric>
ERROR: can't get kernel image!
<viric>
grrrr
<viric>
failed.
<kristianpaul>
viric: you can use rtems too
<viric>
usbboot, I've to use :)
<kristianpaul>
i'm confused are you talking about the nanonote or milkymist one?
<viric>
now the nanonote
<kristianpaul>
it seems :-)
<viric>
flashing with usbboot, it works.
<viric>
Why "cat uImage > /dev/mtdblock1" does not work?
<viric>
kyak: I just got memstat running in the nanonote finally (with the proper kernel)
<viric>
for what I see... it looks as if, of 13MB used, only 4800KB are from user mode processes.
<kyak>
so.. 8 Mb is kenrel?
<viric>
I've always found hard to understand, where the MB go
<viric>
The kernel is taking at least 2MB (Slab+KernelStack)
<viric>
'free' is not counting the 'Cached' shown at /proc/meminfo... weird.
<viric>
Ah, busybox 'free' only takes into account Buffers, not 'Cached'
<viric>
so we have 3.4MB Cached...
<viric>
It's getting close to the 8MB..
<viric>
kyak: the total of PSS should give you the total amount of RAM used by processes
<viric>
kyak: uhm counting that 'memstat' is counted twice :)
<kyak>
twice, huh?
<viric>
well, it's as PID 'self' and as the PID number
<viric>
I'll fix it
<kyak>
heh :)
<kyak>
viric: so again, PSS is memory occupied by running processes, who uses the rest of the memory?
<viric>
kyak: there is the kernel code, the kernel slabs (similar to its 'heap', if I understand correctly), the filesystem cache, kernel buffers, kernel stack, ...
<viric>
I'm not an expert.. I'm just on the way to understand.
<kyak>
ok, so the reset is "the kernel"
<kyak>
*th rest
<viric>
what are your numbers?
<viric>
(I fixed the memstat meanwhile)
<tuxbrain>
never a video of a bliniking led has been so elaborated
<tuxbrain>
I have just finished the edition of the first ArduNote video... now is rendering and I hope this night can upload and publish the Howto post....
<viric>
Does anybody know why copying an uImage to "/dev/mtdblock1" fails to write a good kernel?
<viric>
the same uImage written with usbboot works fine
<viric>
I can't say I understand the flash devices much
<wpwrak>
viric: what program did you use to write ?
<wpwrak>
viric: and did you erase the area first ?
<viric>
wpwrak: cat uImage > /dev/mtdblock1
<viric>
:)
<viric>
kyak: could you run memstat in the nanonote?
<kyak>
viric: i'm on it
<kyak>
recompiling the kernel
<viric>
ah grat!
<viric>
great
<kyak>
it would have been faster
<kyak>
but i accidentaly erased the rootfs ;)
<viric>
uf
<kyak>
(so i have to rebuild the image, too)
<viric>
how come? :)
<kyak>
this is a strange story :)
<viric>
did you use usbboot?
<viric>
or you tried that "cat uImage > /dev/mtdblock1" ?
<wpwrak>
viric: hmm, that may work, but it's probably better to use nandwrite
<wpwrak>
viric: did you erase the nand before writing ?
<jcom>
the Pandora is a nice device, but having to pay in advance without knowing when you are going to receive your unit is discouraging
<wejp>
yeah
<mth>
I'm still waiting for mine, ordered 2.5 years ago
<mth>
there still isn't a device that has the same feature set though, so I think it will still be nice to get it
<wejp>
yes, it is
<jcom>
well, some people thinks that the next PSP will be much better
<wejp>
of course smartphones are getting faster all the time, so they can do lots of things now which they couldn't do two years ago, so at least partly they can replace a device like the pandora
<wejp>
not for serious gaming of course
<mth>
PSP might be nice if it wasn't made by Sony
<wejp>
yeah
<wejp>
even if the next psp would be a lot more powerful i wouldn't buy it
<mth>
the Nokia N900 is probably the closest thing to the Pandora
<mth>
very similar internally
<mth>
but no game controls
<wejp>
yep
<viric>
well, it's not all about the power...
<viric>
a nanonote with more ram can beat many things
<viric>
and a pointer device
<jcom>
going back to my question, does anybody have a copy of the reflash_ben.sh script?
<jcom>
the current link ends in the xburst-tools git page