<kristianpaul>
wpwrak: When you sample something from the usrp, does it handles the i/q data as a complex variable? or just a 16bit integer interleaved?
<kristianpaul>
having problems understandig/interpreting signed number representations in C
<wpwrak>
it sends pairs of q/i integers. each pair is a single complex number ;-)
<kristianpaul>
yeah sure.
<kristianpaul>
complex^
<kristianpaul>
I'm just debugging something here so i can have this bit2bloatbyte util done, and _finally_ send 16 bit i/q samples for processing :/
<wpwrak>
wow :) shouldn't this be a ~10 minutes task ? :)
<kristianpaul>
:(
<kristianpaul>
wpwrak: I had to learn C in the process tought :(
<kristianpaul>
feel bad about that
<wpwrak>
talk about a nice learning curve ;-)
<akiwiguy>
hello
<kristianpaul>
hi akiwiguy
<kristianpaul>
wpwrak: yeah, i learn a lot the last weeks :-)
<kristianpaul>
s/learn/learnt
<wpwrak>
kristianpaul: i can imagine ;-)
<rjeffries>
as a point of reference and yes I am aware, this example is not free...
<rjeffries>
LG Optimus V is sold by Virgin Mobile for $150 USD no contract required
<rjeffries>
is a decent 3.2 inch Android gadget, good specs. best of all they offer a reasonable service plan for$25 a month
<rjeffries>
inckudes unkimited data and 300 voice minutes. naturally has camera and wifi
<rjeffries>
someone will root it soon one assumes
<larsc>
???
<akiwiguy>
our Ben has no camera, no touchscreen, no built-in wifi, slower processor, less ram, etc. and we are selling it for $99usd, $51usd cheaper than that phone. guess what one most people will go for.
<larsc>
there wouldn't be enough nanonotes to supply "most people" anyway ;)
<wolfspraul>
akiwiguy: ok, go on. what's your conclusion from this analysis?
<akiwiguy>
well, its not _free_, so i still want a Ben.
<wolfspraul>
akiwiguy: btw, are you aware of the Milkymist/Milkymist One project?
<akiwiguy>
yes I am
<akiwiguy>
i am a lurker on the mailing list, i hear everything
<larsc>
imo the nanonote is not a consumer product
<wolfspraul>
he, good [hear everything]
<wolfspraul>
larsc: I would agree, although we could probably focus for example on a pure dictionary image now that would make it a 'consumer product', however that is defined
<wolfspraul>
the 02-23 image is by far not competitive with electronic dictionaries, but if we would focus on that for a few months, we probably could pull off an image that would be competitive
<wolfspraul>
and then we could claim 'consumer product' status, because several tens of millions of electronic dictionaries are selling each year
<wolfspraul>
but I'm not into claiming ownership of certain categories, so at least for myself, I won't prioritize like that and instead I will continue to prioritize what makes the Ben more fun and interesting to me, and watching it grow over the long term.
<wolfspraul>
as of today, even with 02-23 image, the Ben is definitely not a 'consumer product' (although I am not 100% sure on the definition of that, let's say something that can sell in 1 million or more units)
<akiwiguy>
hmm, is vim included in the latest image?
<wolfspraul>
if not that would be a big oversight - a 'dedicated vi device' is how we got our first press coverage :-)
<akiwiguy>
O_o
<akiwiguy>
wow.
<akiwiguy>
when I get my Ben, i'm installing the todo.txt cli, and copying some music to it.
<akiwiguy>
I will be my music player, a replacement for my old Palm III, and my note taking device for school.
<akiwiguy>
And, if I manage to get a wifi card, my IRC client.
<akiwiguy>
and i'll disable gmenu2x as well
<akiwiguy>
because it looks horrible and clunky from what i can see
<wolfspraul>
akiwiguy: one by one, keep us posted about your Ben experience, it sounds like you are on a good path.
<wolfspraul>
Palm III, eh?
<wolfspraul>
man this is long ago
<akiwiguy>
Yep.
<wolfspraul>
is it still working?
<akiwiguy>
nope.
<akiwiguy>
stopped working a couple of months ago
<wolfspraul>
not bad
<akiwiguy>
Brought it second-hand off TradeMe.
<wolfspraul>
I developed (proprietary) software that was shipped on most Palms for a number of years
<akiwiguy>
(the New Zealand equivalent of ebay, if you're wondering)
<wolfspraul>
I forgot whether we already worked on the Palm III, that's really a long time ago
<wolfspraul>
Palm V we did
<akiwiguy>
oo
<akiwiguy>
cool.
<wolfspraul>
when it stopped working a few months ago, what broke/failed?
<wolfspraul>
was it sudden failure, or slowly?
<akiwiguy>
i accidentally shorted the battery contacts
<akiwiguy>
thats what blew it up
<wolfspraul>
oh
<wolfspraul>
:-)
<wolfspraul>
that's a sad end
<akiwiguy>
yeah..
<akiwiguy>
if i didnt do that, it would still be in use heaps today
<akiwiguy>
i used it for my to-do list, and tetris, mainly
<wolfspraul>
the Ben will not be able to match it in usability right now, I think
<wolfspraul>
the way Palm implemented the super-fast bootup time by keeping part of the memory powered in 'off' mode was great.
<wolfspraul>
that truly 'instant' on alone is a big win for the Palm, compared to Ben
<wolfspraul>
then they used execute-in-place from the flash, etc.
<akiwiguy>
just keep it in suspend mode
<wolfspraul>
can't easily beat that system
<wolfspraul>
ok I talk about what we really have today, say with the 02-23 image, not what is theoretically possible
<wolfspraul>
suspend works, but
<akiwiguy>
it crashes after a few hours
<wolfspraul>
a) it is not well integrated into the image, I think you need to manually enter it on the command line
<wolfspraul>
b) correct. I hear that it crashes after x minutes/hours
<wolfspraul>
c) we don't know (or maybe just me) how fast the battery will run down in suspend mode
<akiwiguy>
how long does the Ben take to boot? ~20 seconds?
<wolfspraul>
bottom line: in today's reality, suspend doesn't work for an 'end user', I think most of Ben users will always do a full boot/shutdown.
<wolfspraul>
yes, about 20 seconds
<akiwiguy>
well, thats acceptable
<akiwiguy>
i dont mind waiting for it to boot up
<wolfspraul>
maybe we can bring that down, Lars did some great improvements recently, but we will most likely eat them up by mounting more of the NAND space
<wolfspraul>
then we can look at some other big parts of the equation, like openwrt boot scripts, even some slowness in u-boot or very early power-on
<wolfspraul>
the proprietary dictionary software we are replacing boots into a fully usable state in about 4 seconds
<akiwiguy>
why does mounting more of the NAND slow bootup?
<akiwiguy>
does it force an fsck or something?
<wolfspraul>
because of filesystem/nand behavior
<wolfspraul>
it's all just a matter of optimizations, no theoretical limits
<wolfspraul>
not that many developers are on similar embedded systems and care about nand/ecc/filesystem speed optimizations, or they are not willing to do ugly hacks to ship a particular product, which is what will typically happen at a 'normal' manufacturer
<wolfspraul>
devices on the market have 16 gb, 32 gb and more, bootup faster, and somehow make the space accessible. don't know how they achieve this in detail, but they do.
<akiwiguy>
an iPad with 16gb of flash boots up in ~10 seconds to a fully usable state
<wolfspraul>
I can imagine.
<kristianpaul>
still having a booting palm m101
<akiwiguy>
Right, I've got to go. Bye everyone
<kristianpaul>
begin to like more pointers, oh memory ! :-)
<wpwrak>
akiwiguy: (kill gmenu2x) sounds like a good move ;-)
<wpwrak>
akiwiguy: (love the ben despite its shortcomings) even better :)
<wpwrak>
in my view, then ben is a foot in the door. its successor could be an open door. the next device after than could be entry into whatever is behind the door. and maybe the 4th would actually take possession of the place.
<kyak>
wpwrak: hm, sounds like a four-step world plan! :)
<kyak>
for sftp, you could install openssh-sftp-server
<dvdk>
maybe openssh-sftp-server should be included by default?  using gnome vfs for mounting is just very practical
<dvdk>
ah, ok.  have to put authorized_keys to /etc/dropbear
<akiwidroid>
Good.night.all
<dvdk>
kyak: cool, works.  thanks
<kyak>
dvdk: my guess is that openssh-sftp-server would install only one binary sftp-server, which will be used by dropbear to server sftp requests
<dvdk>
kyak: yes, certainly.  rsync does it that way, too.
<kyak>
btw, i prefer typing scp commands on Ben, rather then on PC :)
<kyak>
of course, in a ssh session
<kyak>
(including openssh-sftp-server by default) - why not?
<kyak>
i think we also need to include coreutils-sleep (busybox's sleep only knows integer sleep values) and use more functional procps (replacing busybox's crippled ps, top etc)
<kyak>
i already included gnu tar some ago, cause busybox's tar fails on some files
<kyak>
and really one thing that bugs me in ash is a lack of reverse-i-search (ctrl+r)
<kyak>
but bash consumes at least twice as much memory as ash, so there is no option to replace ash :)
<viric>
how much is that?
<viric>
here I run bash with glibc, and it takes a VSIZE of ~4MB :)
<viric>
I wondered how much was due to glibc
<Jay7>
kyak: did you compared dash/zsh?
<viric>
kyak: and the reverse search is up to readline, I think. Not bash.
<kyak>
viric: RSS is 512/984 (ash/bash), VSZ is 1636/2036
<viric>
ok, bash+glibc doubles that
<kyak>
viric: how to enable it in ash then?
<viric>
kyak: is ash linked with gnu readline?
<kyak>
viric: ash is a symlinks to /bin/busybox, and there is no link to readline
<viric>
ahh
<dvdk>
cool, mplayer is now drawing y,u,v to physical memory.  now i only need to tell the IPU to blit that stuff to the screen.
<viric>
busybox has its own short readline implementation
<kyak>
Jay7: nope. Don't what to use "another shell"
<viric>
kyak: maybe you can enable a better readline in busybox
<kyak>
viric: i noticed that ash respects soem readline settings, so yes it does
<viric>
kyak: I wonder how much of the bash vsize is 'anonymous memory'
<viric>
kyak: so, code pages not counted.
<kyak>
viric: good idea. maybe ash can use fully functional readline
<kyak>
viric: how do i check?
<viric>
in the busybox build...
<kyak>
i mean, the amount of "anonymous memory" :)
<viric>
hm
<viric>
/proc/pid/maps has something about that
<viric>
but it requires some processing I think
<viric>
I once had a script that processed those
<kyak>
viric: why don't you use uclibc anyway?
<viric>
I have some pieces that don't build fine with uclibc, iirc.
<viric>
glibc works straight away, and I did not get short on memory for what I run in the ben, still
<kyak>
i see!
<kyak>
dvdk: have you commited the patches by xMff already? i'm running a clean build after today's merging with upstream backfire, and it fails on gforth -\
<dvdk>
build for me last time i checked
<viric>
kyak: I'm one of those not running the openwrt :)
<dvdk>
think i committed.  only minimal patch anyways
<dvdk>
can you send the log?
<kyak>
viric: sure, i rememeber :)
<kyak>
dvdk: one moment
<kyak>
dvdk: yes, it built fine last time i checked, too. But we had a big update today due to merge of latest backfire into our branch
<viric>
anyone good in gcc?
<viric>
I want to know in gcc what can be setting the preprocessor variable LARGEFILE64_SOURCE
<dvdk>
hmm, trying to update myself.  good opportunity to have a long break (during recompile :)
<dvdk>
binary compatability still in place?
<kyak>
yep
<dvdk>
good
<kyak>
dvdk: very strange. It failed on "make". Then i ran make package/gforth/{clean,compile} V=99 > ~/
<kyak>
tmp/gforth.log 2>&1
<kyak>
to reproduce the problem. And it build fine!
<dvdk>
maybe an unset dependency causes the problem, so at the second attempt the lib was in place?
<dvdk>
do i need to make clean?  or just make world?
<kyak>
i'm running make in a single thread, so all dependecies should be satisfied one by one...
<kyak>
i built is from scratch, just in case
<dvdk>
only if DEPENDS line correctly set in gforth/Makefile
<wpwrak>
kristianpaul: they already arrived and are working on getting back connectivity, which is not so easy in the distant deserts beyond the pampa :)
<wpwrak>
kristianpaul: we have a long weekend here, so he may not show up before wednesday or so
<kristianpaul>
wpwrak: ah, good to know :-)
<kyak>
/bin/sh ../libtool --tag CC --tag=CC --mode=link mipsel-openwrt-linux-uclibc-gcc  -g -Os  -o libmath.la  stubs.lo  -lm
<kyak>
no fPIC there
<xMff>
kyak: exactly
<kyak>
it is passed down?
<kyak>
*is not
<xMff>
well all stuff before and after has it
<xMff>
the stub.o compile command is also way shorter than the rest
<qi-bot>
[commit] Cristian Paul Peñaranda Rojas: New bit2bloatbyte util, so now data can be off-line processed as 16bi tI/Q http://qi-hw.com/p/ben-gps-sdr/ac28678
<qi-bot>
[commit] Cristian Paul Peñaranda Rojas: Some comments and explanations about bit2bloat tool added to notes http://qi-hw.com/p/ben-gps-sdr/857135a
<qi-bot>
[commit] Cristian Paul Peñaranda Rojas: New bit2bloatbyte util, so now data can be off-line processed as 16bi tI/Q http://qi-hw.com/p/ben-gps-sdr/ac28678
<qi-bot>
[commit] Cristian Paul Peñaranda Rojas: Some comments and explanations about bit2bloat tool added to notes http://qi-hw.com/p/ben-gps-sdr/857135a
<kristianpaul>
hmm this bot should distinguish between branches
<wpwrak>
kristianpaul: (spikes) they shouldn't be there, but i don't know yet if they cause any real harm. i don't like them. will try to see if i can shrink them a little.
<wpwrak>
kristianpaul: (shrink) with that extra cap.
<wpwrak>
kristianpaul: some of them may also exist due to soldering problems. things like vias that don't connect properly or flux residues
<kristianpaul>
wpwrak: also how many reworks per board counts i think
<wpwrak>
kristianpaul: yeah, that too :) well, rework in this case usually meant to connect a via or to remove a solder bridge, so things converged
<kristianpaul>
start reading about DMA
<akiwiguy>
Hello.
<kristianpaul>
hi
<qi-bot>
[commit] Werner Almesberger: usrp/sps/: use PATH instead of hard-coding relative paths in invocations http://qi-hw.com/p/ben-wpan/cb3e80f