<wpwrak>
xiangfu: heya ! i may have found another bug in usbboot ! :)
<wpwrak>
xiangfu: when i follow the jlime install instructions (on my x86-64), the flashing of u-boot returns an error, and the byte patterns usbboot shows are indeed different.
<wpwrak>
xiangfu: now, when i use usbboot in interactive mode instead with -c, everything works
<wpwrak>
zrafa: (ldd) okay. still have to try it. i solved the problem i was trying to debug with ldd already, so i didn't try to install ldd. it's on my mental to do list, though ;-)
<kyak>
zrafa: btw, it's Alt+q, as mentioned in gmu help ;)
<wpwrak>
kyak: i tried Alt-Enter and it exited
<wpwrak>
(of course, maybe it didn't really works and it just crashed. you never know with those computers :)
<wejp>
Gmu never crashes :P
<wpwrak>
maybe it's a solipsist. solipsists never die - it's just the universe that ceases to exist ;-)
<bartbes>
it has always been alt+enter afaik
<bartbes>
and it lists that in the 'status bar'
<zrafa>
kyak: ah.. well, it is alt+enter on jlime.. BTW, is not wejp the gmu author? he also said alt+enter
<wejp>
yes, i am. but as it caused so much confusion i changed it to alt+q in the new version
<wejp>
in the old one it is of course still alt+enter
<zrafa>
wejp: ( it caused so much confusion i changed) ;-)) a lot more of confusion
<wejp>
but you can of course map it to something else, if you want
<wejp>
ESC or whatever ;)
<zrafa>
wejp: yeah..
<zrafa>
that is great
<wejp>
it is a simple config file :)
<kristianpaul>
ESC !
<zrafa>
kristianpaul: "press any key to exit"
<zrafa>
:)
<wejp>
i thought about using escape, but then i thought it is so easily pressed by mistake..,
<kristianpaul>
lol
<kristianpaul>
thats true
<kristianpaul>
wejp: what is the easy way of increse font size?
<wpwrak>
ESC !!
<zrafa>
wejp: btw.. would you like to do a similar video player?.. I mean, similar to gmu but for playing videos on nn. I have the engine of a theora player.. which is not mine, it is a very stupid way to modify the theora example player which comes into theora libs.
<wejp>
kristianpaul, using another theme, there is one included with gmu with a much larger font for the text, you could try that
<kristianpaul>
wejp: there are more themes?
<zrafa>
wejp: it just needs a GUI. I added into jlime on qi that player with a simple fileselector.
<wejp>
zrafa, maybe, what i will probably not do is include video playback capability in Gmu, but another program with similar controls.. maybe
<kristianpaul>
may one simpler (no playlist, easy play of files)
<wejp>
kristianpaul, yes, and you can also create our own if you want to
<kristianpaul>
i always get confused to play a file is enter in mode and "a! in other
<zrafa>
wejp: yes, I am talking about another program, but similar look&feel
<kristianpaul>
ok wejp
<wejp>
oh, and if anybody of you wants to create another theme for Gmu, i would really appretiate it :)
<zrafa>
wpwrak: I added the touch /etc/shadow to the items to add into myrootfs (when some day some of us write it using jlime repo for nn) :)
<wpwrak>
zrafa: great, thanks  !
<kristianpaul>
AES will be nice for private chat, or voice..
<lekernel>
if you compare it to other S6 systems, yes, and we are working to lower it
<wpwrak>
yes, if you compare to other solutions that take an expensive approach :)
<lekernel>
there aren't others :)
<wpwrak>
but who said a "reconfigurable computer" needs an fpga ?
<wpwrak>
of, for that matter, a "multimedia computer" :)
<lekernel>
anyway, I can just try to put those keywords on the website and see if it improves traffic/sales/feedback
<roh>
lekernel: nah. its not low cost and calling it computer pulls the wrong crowd (clueless people)
<lekernel>
what do you compare it to, to say it's not low cost?
<lekernel>
and btw, making people give a shit is a lot harder than dealing with a few lusers, so if this tagline appeals to some of them but otherwise works, success!
<wpwrak>
settop boxes. also, "multimedia computer" suggests "reduced function". so, cheaper than a pc.
<wpwrak>
and pc here would mean the kind derived from a netbook, i.e., already quite cheap.
<wpwrak>
you seem to be very concerned about not drawing enough attention. why is this ?
<wpwrak>
everything is embedded :) also, embedded << pc. ergo "low-cost" would be something like the ben or cheaper
<lekernel>
there's also "reconfigurable", which is the important thing. today, reconfigurable computers are expensive
<lekernel>
why do I want to draw a lot of attention? because I want to live off this product so I don't have to work an uninteresting job
<lekernel>
simple :)
<lekernel>
(though doing marketing/advertisement is pretty boring as well... but it's a compromise)
<wpwrak>
i don't think "low-cost" describes it well unless you're very specific about what's inside. and not a lot of people will understand the price structure even then.
<wpwrak>
will people understand that "reconfigurable" means FPGA in this case ? my netbook is also reconfigurable - vi /etc/whatever ;-)
<wpwrak>
ah, the old money thing :)
<wpwrak>
but do you think the attention it's getting if falling behind your expectations right now, or is it more an abstract/future concern ?
<wpwrak>
s/if/is/
<wpwrak>
(marketing boring) dunno ... learning how to seduce people to buy your stuff doesn't sound too bad to me :)
<wpwrak>
alright, but how many people know that ? e.g., i didn't. i knew that "reconfigurable hardware" -> FPGA
<lekernel>
mh, let's see
<wpwrak>
how about Open Hardware ? will people who would be attracted by FPGAs also want to look at anything that is "Open Hardware" ? if yes, you don't need the FPGA/reconfigurable in the headline
<lekernel>
maybe I can just put a meta keyword tag and see how much traffic/feedback that brings
<wpwrak>
of "Reconfigurable Open Hardware" if you want, but that already gets a bit long
<lekernel>
"open hardware" is more or less synonymous with arduino atm, and similarly trivial electronics
<wpwrak>
i don't think "lay out the bait and wait" is going to work
<wpwrak>
hmm, dunno. but i'm not unbiased enough to be a useful reference
<lekernel>
(marketing boring) let's say programming crazy things like mm soc or llhdl is more suited to my nerdy taste
<wpwrak>
also, if people's expectation of "open hardware" includes the arduino, that doesn't have to mean that it couldn't also include the MM1. "multimedia" or VJ would already contradict that it's something as simple as an arduino
<wpwrak>
maybe they'd expect a propeller. of course, if they know what the propeller is, they'd probably have a look, just out of curiosity if someone actually figured out something useful to do with it ;-)
<wpwrak>
back to the marketing strategy, i don't think you can expect people to just find you by accident. you need some multipliers, be it the classical press, the internetized para-press, or outright viral marketing
<lekernel>
yup, I know
<wpwrak>
for the classical press, you need a press release or you have to be where they're looking for news.
<lekernel>
but keywords work too, and are actually how google makes money :)
<wpwrak>
did you speak to any press at 27c3 ? would it even have been likely for them to notice you ? or were the infrastructure problems too overwhelming ?
<wpwrak>
you mentioned fosdem. will you give a talk there ?
<wpwrak>
how often do you idly google around for relatively general expressions ? :)
<wpwrak>
e.g., do you google for "reconfigurable computer" often ? or more like "xc6s45" ?
<lekernel>
putting those keywords and measuring the resulting traffic is an unbiased way to know this
<wpwrak>
i'm also not sure the end user focus is a very efficient approach. it may work, but you could also try to draw the attention of companies who would be willing to finance further work.
<lekernel>
mh?
<lekernel>
with NDA and such? :)
<wpwrak>
so it's not only places where the press goes but also places where people go who would be interested in novel technological ideas
<wpwrak>
you can make your requirements clear
<wpwrak>
i know you don't like to emphasize linux, but there are a number of linux events that do get a fair amount of attention. e.g., you could consider having a shared presentation with someone working on the kernel.
<wpwrak>
it would be even better if your story could include an element where software work influenced the hardware design, e.g., an optimization of some item or a new feature.
<wpwrak>
because that would be something truly unique. not just "linux ported to yet another cpu", but "cpu optimized on the fly for linux"
<wpwrak>
of course you probably have that already on the multimedia side. so perhaps you could just work that angle.
<zrafa>
lekernel: I know at least one guy in Argentina who would buy Milkymist machines :).. He (and friends) writes moldeo (moldeo.org) and they do business as well with graphics on real time for performances, etc..
<zrafa>
lekernel: sorry for the spanish web sites.. but you can see the videos at least :P
<zrafa>
ah.. they have english web sites as well.. cool
<kristianpaul>
Free CPU multimedia computer...
<kristianpaul>
reconfigurable CPU
<kristianpaul>
you cant reconfifure your cpu using vi wpwrak ! :-)
<kristianpaul>
"cpu optimized on the fly for linux" <-- awesome :-)
<lekernel>
maybe but it isn't. actually, the linux port pretty much sucks
<lekernel>
zrafa: but they're developing their own platform, aren't they? did they tell you they were interested?
<kristianpaul>
sure sure
<kristianpaul>
zrafa: moldeo, nice !
<kristianpaul>
music is very familiar :-)
<kristianpaul>
lekernel: now you can run mupdf on M1, are you planing use the computer it self for next presentations?
<zrafa>
lekernel: they did not tell me anything yet because I know them but they do not know me :) well, I know their work, and I met a couple of years ago them. But I just will let them know your work at least. They are really fans of all this kind of technology (video, sound, real time..) so I guess that they would like to meet some open hardware for that
<zrafa>
I met= I met them
<kristianpaul>
I dont know nobody  (may be interested in multimedia) yet in Colombia, but i was invited to a event in wich i can show M1 working all day !
<kristianpaul>
It is called LabSURlab
<wpwrak>
lekernel: (linux) yeah, probably needs to get a bit better for this. so, something to keep in mind.
<lekernel>
why do you insist that much on linux...
<kristianpaul>
Because we use it everyday?
<wpwrak>
lekernel: (linux) well, for one thing it's a fully featured operating system. if you want your SoC to be general-purpose, that's about the only sane choice you have
<wpwrak>
lekernel: (linux) second, i was specifically referring to PR opportunities. linux is a well-established channel for "open" stuff. so if you can present your work at a linux event, you may be able to reach a significant audience.
<wpwrak>
well, a major linux event
<kristianpaul>
linux.conf ?
<wpwrak>
linux.conf.au ? yes, that could be a good place
<wpwrak>
linux symposium is a bit too close, time-wise
<wpwrak>
linuxtag and linux kongress may be convenient in terms of travel
<lekernel>
fixing all the linux problems is probably way more painful than fixing the remaining flickernoise/rtems/whatever software i'm using atm and marketing that thing directly to VJs :)
<lekernel>
it's sooo easy to say "improve the linux port" :)
<wpwrak>
for a single-function device, that approach is fine. but do you really want a single-function device ? also, can you keep on maintaining it ? it does have usb host, so people will come with peripherals you didn't anticipate. do you tell all of them to just bugger off ?
<wpwrak>
i can see rtems as an intermediate step. sort of a proof of concept. make the device work, see what it can do. but then, take the essence of what you actually need and let others worry about the 99% that's the boring rest.
<lekernel>
a single function device is better than no device at all :)
<lekernel>
and if I used linux, there would be no device at all NOW
<lekernel>
because I would have been so slowed down by its issues that I couldn't do anything else
<wpwrak>
as i said, intermediate step. dunno where you'd be with a 100% linux approach. maybe you would have something usable now, too. maybe not.
<lekernel>
definitely not
<lekernel>
linux is horrible
<lekernel>
especially if I used X as you recommend :)
<wpwrak>
there's a ton of things you can do without x :) so if you insist of scribbling directly into your frame buffer or whatever, linux won't stand in your way :)
<wpwrak>
of course, if you do have X, there's a gazillion of applications that you don't have to worry about anymore. they just run. even if perhaps slowly.
<lekernel>
yeah, after you have fixed all the related gcc, libc, and autotools problems - and there are tons of them
<lekernel>
plus a few random instability bugs in the kernel, every now and then...
<lekernel>
just to spice up the whole mess
<wpwrak>
well, you yourself said that these things get better if you're on a GNU system. does Linux or, like rms pronounces it, GNU/Linux count ? :)
<wpwrak>
kernel instability is bad, agreed
<lekernel>
not on LM32
<zrafa>
lekernel: what do you use as compiler for development? and What do you use as OS on your desktop for development?
<kristianpaul>
no no dont ask that !
<kristianpaul>
;)
<zrafa>
kristianpaul: I can change the question: which environment is used to build the software for Milkymist? (if there is such environment or building process)
<kristianpaul>
zrafa: boostraping right :-)
<kristianpaul>
bootstraping*
<lekernel>
the only gnu component is gcc. but the fact that's it's the only available compiler for some architectures like lm32 is just sad
<kristianpaul>
is linux a gnu project too?
<lekernel>
no
<larsc>
hurd is the gnu kernel
<lekernel>
the gnu equivalent of linux is the hurd :)
<lekernel>
haha
<lekernel>
which says a lot about GNU :)
<lekernel>
at least the linux kernel works :)
<kristianpaul>
ok, how receptible are llvm people to support your work?
<lekernel>
they don't care
<lekernel>
they think apple is cool
<kristianpaul>
:p
<zrafa>
kristianpaul: I wanted to ask because I would guess that it is not the same to build software with gcc if you run it on different OS. For example, if you build software with gcc on a propietary software OS, it could give a different output than if you build your software with gcc on an open source OS.
<larsc>
they are paid by apple
<kristianpaul>
good point
<lekernel>
zrafa: huh?
<wpwrak>
zrafa: heard of the M$ conspiracy, too ? ;-)
<zrafa>
wpwrak: no, I have not. But it could be possible to do interesting things inside a propietary software running binaries (like gcc)
<zrafa>
interesting things: no for us surely
<lekernel>
zrafa: wtf?
<wpwrak>
zrafa: i'm sure the M$ anti-milkymist division is working night and day on this in their secret underground base in antarctica. stuxnet was just a first test balloon.
<kristianpaul>
?
<wpwrak>
kristianpaul: guess he's still a bit jet-lagged ;-)
<zrafa>
wpwrak: haha.. well, anyway milkymist software is not built using gcc on M$ anyway, RIGHT?
<kristianpaul>
no no
<kristianpaul>
thats scary zrafa !
<lekernel>
well you could do it quite easily
<kristianpaul>
xD
<wpwrak>
zrafa: you were right ! he _is_ in league with satan ! :)
<lekernel>
oh well
<lekernel>
at least the windows people won't come up with ideas like using X11 :)
<zrafa>
I could do another users question: if I buy a Milkymist machine, is the software inside open source software built on a propietary OS?
<wpwrak>
windows ... when you're tired of choosing the lesser evil
<zrafa>
user question=newbie Milkymist user question
<kristianpaul>
anyway, M1 is out,  thats good, more people will look at it, thats needed *reflesh* for the project
<lekernel>
zrafa: no one would ask that question... anyway, there are people building it on mac os, but the binary images I release are built on an open source OS...
<lekernel>
running on proprietary hardware however :)
<zrafa>
lekernel: that is cool (built on an open source OS).. propietary hardware.. well, no many choices I guess
<zrafa>
lekernel: now we have a goal to port linux to milkymist! ;)
<zrafa>
lekernel: open hardware to build the milkymist software on an open source OS ;)
<lekernel>
people having this goal aren't that hard to find
<lekernel>
people doing it after they endured a couple of gcc segfaults and similar problems are a bit more difficult to come by
<wpwrak>
is the gcc port being maintained ? i.e., if you hit a compiler bug, is there someone who will fix it ?
<lekernel>
and there are _way_ more simple software tasks in this project, but it's rtems or vj stuff
<lekernel>
wpwrak: even when I send a patch, and even for simple things, they would question it
<kristianpaul>
wpwrak: (gcc port being maintained) i dunno
<wpwrak>
lekernel: the patch or the underlying problem ?
<kristianpaul>
lekernel: sure they question you're not gcc developer :-)
<lekernel>
more than a month ago I sent a patch so that newlib based compilers would support the hardware divider in the libc instead of using the software emulation everywhere, which is totally stupid
<kristianpaul>
thats normal
<lekernel>
it's still not merged, and all they say is "there's no valid reason for doing this"
<lekernel>
I don't hate the GNUtards for no reason...
<lekernel>
actually it seems some people seem turned down from this architecture (partly) because of the issues I'm describing
<wpwrak>
lekernel: well, if you think your upstream isn't cooperating enough, you could also ask that person in private communication what the problem is. if you're indeed the only lm32-*, your own QA should have a lot of weight. if there are others, you may want to find out about them.
<lekernel>
a lot of lm32 ports are dead, including the rtems one... when porting to milkymist, I upgraded the interrupt API of the generic architecture code to the latest rtems interrupt system (the previous one was deprecated)
<lekernel>
this broke the initial port to the original LM32 system (EVR from lattice)
<lekernel>
I contacted the maintainer about that, and he didn't care... his lm32 port was "nothing really serious" :)
<wpwrak>
lekernel: ah, never break the origin ;-)
<lekernel>
well at least the rtems code is manageable. I can fix it when anything breaks without spending inordinate amounts of time
<kristianpaul>
how long are you planning to support all this fixes?
<lekernel>
?
<wpwrak>
maybe the gcc lm32 maintainer is also not quite aware of what ports are around and in what state they are. maybe what's missing is a place where the various people interested in lm32 communicate ? a bug tracker is a horrible means of communication.
<wpwrak>
kristianpaul: seems that he doesn't have much of a choice
<kristianpaul>
wpwrak: i know :(
<wpwrak>
kristianpaul: that is, before solving the upstream issue ...
<wpwrak>
a second-order question would also be how to get gcc maintainers to play a more active role in supporting lm32 and specifically lm32-mm1.
<lekernel>
can we replace gcc with llvm? :)
<lekernel>
if we try hard to get compiler developers to jump into this project, at least target the good programmers *g*
<wpwrak>
dig yourself even deeper into a niche ? :)
<wpwrak>
if auto* is bad now, just imagine how much fun it would be with llvm
<lekernel>
at least now I understand better why companies like ARM crack down on open source instruction-compatible processors
<lekernel>
they must have thrown a crazy sum of money to pay redhat, codesourcery, etc. to get them to fix the gnu stuff and others...
<wpwrak>
;-)
<wpwrak>
maybe they also enjoy all the pain you have to go through if you're off the beaten path
<lekernel>
yeah, this changed a lot my view on free software and GNU
<wpwrak>
well, it's always the question of getting enough of the right people interested enough
<wpwrak>
sometimes it's easy, sometimes it's hard
<wpwrak>
the more niche you make mm1, the harder it will get. you'll gain more control, but also a lot more responsibility.
<lekernel>
otoh i'm not that impressed by embedded linux
<lekernel>
the only successful use I've seen of it is openwrt routers
<wpwrak>
now that comes as a total surprise ;-))
<lekernel>
openmoko, olpc, and, i'm sorry to say, the nanonote all suffer from serious software issues
<wpwrak>
openwrt are among the smallest, yes. well, if you ignore uclinux and such
<lekernel>
the only exception is android, but they are only using the kernel...
<wpwrak>
imho, nanonote suffers from a misguided distribution strategy
<wpwrak>
virtually all the sofgtware development effort seems to go into porting things that already run on the device
<wpwrak>
just not with openwrt
<lekernel>
sometimes I think I should have just written an ARM clone and throw their C&D letters directly into the trash
<wpwrak>
the question that i ask myself is, what would people like kyak, david kuehling, and xiangfu do if the "standard" distribution was jlime ? if the answer is "nothing" or "leave", then that's okay. if the answer is "do something new and great", then the focus on openwrt is costing us.
<lekernel>
I don't think they have any legal basis, at least in Europe, it's mere intimidation
<wpwrak>
lekernel: yeah, arm should be old enough that the first stuff should be free from patents now :)
<wpwrak>
lekernel: remember "look and feel" ? ;-)
<wpwrak>
lekernel: well, if you feel bored ... the weekend is still young, and you have an fpga all to yourself ... ;-))
<wpwrak>
lekernel: on the legal side, the problem may not be the instruction set per se, but anything else they may be able to dig out when they choose to attack you.
<lekernel>
mh?
<wpwrak>
lekernel: you may use some implementation technique they've patented more recently
<lekernel>
btw, Verilog is software, no?
<wpwrak>
lekernel: something that's not essential for implementing the instruction set, but just something you happen to use. or that a judge can be convinced to see that way.
<wpwrak>
lekernel: (verilog = sw) not sure what the legal view of that is
<wpwrak>
lekernel: if i was them, with a multi-million interest in crushing that little upstart geek, i'd spend a bit of money to try to get the courts to prove my view - or to enjoy watching him run out of funds, whatever happens first
<urandom__>
wpwrak is there still a reason for openwrt on nanonote? (other than booting a bit faster and having multiple distros is better than having one)
<wpwrak>
lekernel: of course, you never know where they may draw the line
<wpwrak>
urandom__: may reasons. 1) wolfgang had a bad experience with OE and doesn't want to touch it again. 2) so much effort has been put into openwrt, it's hard to change. 3) openwrt should consume a little less memory than jlime.
<wpwrak>
urandom__: of course, regarding 3), one could argue that, instead of building a shrine around the lack of memory, one could try to move forward and make a model with more memory.
<wpwrak>
urandom__: alas, that may be easier said than done
<urandom__>
i really cant imagine the huge waste of human labor done cause of tousands of linux distros with basically the same goals, fucking stupid
<wpwrak>
well, some have different goals and approaches
<wpwrak>
so a bit of diversity is good
<wpwrak>
urandom__: ah, and 4) jlime-nanonote isn't really supported at the moment. that's of course also a problem.
<urandom__>
i would prefer to see real diversity, people that try something new
<urandom__>
we just have posix, tons of tons of unix clones, thats the whole world of free operating systems, other story but same problem
<wpwrak>
urandom__: well, where would you want to try something new ? in the kernel ? in user space ? everywhere at the same time ?
<lekernel>
actually I think haiku or aros could look nice on the NN
<urandom__>
well there are some interesting os projects lekernel but they dont get much attention
<lekernel>
I tried haiku and aros some time ago and they actually work
<lekernel>
and they're much lighter than a GNU/X/Linux setup with the same features
<lekernel>
the biggest problem with aros was instability
<wpwrak>
lekernel: yeah, there are tons of experimental/niche/research systems out there. i mean, it's okay as a hobby. you can now have your own kernel and your own whatever. in your parent's generation, they had their own garden. they didn't expect the whole world to come there visiting either. so it's cool.
<lekernel>
pff, stupid
<lekernel>
so no one should innovate in operating systems, ever
<lekernel>
just use the same good ol' crap
<wpwrak>
lekernel: no, that's not what i'm saying. linux is actually quite open to innovation.
<lekernel>
sure. let's bother Linus again about switching to a microkernel approach, for example :)
<wpwrak>
lekernel: also, if you have a more extreme point to make, perhaps you really do need to create your own universe.
<wpwrak>
lekernel: but it's unlikely that you will find a large following, become the next linux or whatever.
<wpwrak>
lekernel: well, if "microkernel" is all you can think of when it comes to innovation in the kernel, ... ;-)
<wpwrak>
lekernel: interesting things in linux: RCU, the traffic control subsystem, /sys and udev, block scheduling algorithms, cpu scheduling algorithms,
<urandom__>
microkernels are awesome, you can updating parts of your kernel without rebooting, or when a part crashes it doesnt do harm to your whole system
<wpwrak>
lekernel: the network driver architecture, and so on
<wpwrak>
lekernel: with loadable modules, you can update part of your kernel without rebooting
<urandom__>
you mean urandom__
<larsc>
wpwrak: and there is ksplice
<wpwrak>
lekernel: there are also more or less developed process state dumpers that would allow you to pretty much pick up where you left off after a reboot. i think their lack of development is a testament to their limited real-life utility. (and yes, those things can go pretty far. i did some work there myself, although more because of intellectual curiosity)
<wpwrak>
(urandom) oh, sorry :)
<wpwrak>
ksplice ... scary :)
<urandom__>
speaking of interesting os konzepts, http://code.google.com/p/loper/ , nothing to see yet but has some interesting thoughts and links
<wpwrak>
there was also significant work on the file system side: ext3/4, reiserfs, fuse
<wpwrak>
if you're looking for things that go a bit beyond the practical, i can suggest a look at my umlsim and tcpcp. all abandoned by now, but they were fun at their time :)
<lekernel>
sure, but no fundamental OS concepts
<roh>
how much work is something like the lm32 cpu?
<wpwrak>
memory mamnagement and scheduler also saw their changes
<urandom__>
wpwrak do you think people will still use linux in 50 years? still will use C and C++ for programming , that would be very sad
<roh>
or rather.. implementing something like arm in vhdl?
<wpwrak>
rcu is pretty fundamental
<wpwrak>
roh: verilog :)
<wpwrak>
urandom__: why would that be sad ?
<wpwrak>
urandom__: C is a very nice language
<roh>
or verilog.. dont care.. i just have no clue how much time it needs a developer to 'implement a cpu down'
<lekernel>
for a simple risc like arm7 or lm32 probably a month or two
<wpwrak>
urandom__: can't predict what will happen with linux in such a large time frame. it could collapse under its own weight. but maybe not.
<wpwrak>
roh: s/month/year/Â Â for the rest of us :)
<roh>
lekernel: i know a very good hardcore arm developer and some guys interrested (and imho qualified) for cpu design.. so i just do the mind-experiment what it would take to develop a verilog cpu which is like 'arm-esque' but not patent encumbered or anyhow compatible and doing a linux port for it. like... using all the good ideas which companys often can't because of 'fear' about new platforms
<urandom__>
wpwrak collapse under its own weight, that the point, that why we better look at microkernels for example
<lekernel>
then if you want fancy things like out of order execution, jump predictors, superscalar execution, etc. it can become really complex
<roh>
lekernel: and how much work is stuff like an mmu which linux would need to run properly?
<wpwrak>
urandom__: it's not a monolithic vs. microkernel issue but a question of managing the development process. you have exactly the same issues with a microkernel.
<lekernel>
not that much for the simplest mmu
<roh>
linux iss still very slim for what it can do since most code isnt used all the time or at once. in the end the codebase has stuff in it which will never exist on all systems (or can't)
<lekernel>
i.e. with a fully onchip-SRAM-based TLB with all TLB misses to system memory handled in software
<roh>
lekernel: means given one has a 'known good' cpu design its like 3-4 month doing the verilog for a cpu which linux could run halfway decent fast on? (given that somebody does the port)
<wpwrak>
didn't old PPC do it like this ?
<lekernel>
yeah, maybe
<lekernel>
but if you want to try out "good ideas" it takes time
<urandom__>
cant help but constantly think about lisp maschines when he hears "FPGA"
<lekernel>
also you're also subject to Murphy's law and in case of really bad luck that intermittent instability issue that takes you 6 months to fix
<roh>
lekernel: sure. i an not saying one needs to start such a project now
<roh>
i just realize lately that there seems to be demand for a not completely low end foss cpu design which is fun even when running a linux on.
<lekernel>
why not start yet one more effort?
<lekernel>
there's lm32 which is almost good enough for that (not counting the gnu problems)
<lekernel>
and tons of failed projects
<lekernel>
-not
<roh>
if you thinkg lm32 can be fixed easier than that.. sure.. that would be the way
<lekernel>
just find someone working at redhat or codesourcery who believes in that... and you could probably get most of the gnu issues fixed
<roh>
my thinking was in the line of 'adapt a basic concept which already has a well working port' and redo all the non-freely useable parts..
<roh>
which would break abi, api, everything. meaning resuling in a cloned port modified beyond any compatibility
<wpwrak>
lekernel: roh was just echoing your own thoughts from a while back :-)
<wpwrak>
roh: the amount of breakage you environ sounds a little scary, though
<roh>
e.g. i know that gcc and linux uses some feats of cpus heavily and depends on them being fast, while some arent really that important (while they are still in the cpu and complex). so use the most simplest approach and only do the parts which are important for linux
<wpwrak>
hmm. seems that i have an AC leakage problem in my cnc setup. sigh. will technology ever stop being a nuisance ?
<roh>
wpwrak: sure. yet the result would solve the patent problems, and the 'not even en par even with arm9' issues
<larsc>
roh: exaclty my thoughts as well
<roh>
wpwrak: opto-isolation?
<wpwrak>
roh: laptop, usb hub, and mill each have their own supposedly isolated supply. at least one of them is leaking. i'll explore when the current job is done.
<roh>
sorry for being a bit naive there.. just doing this mind experiment since the current soc, cpu etc situation with the industry is annoying ;)
<roh>
doesnt like the industries non-reuse-of-good-ideas development
<wpwrak>
well, patents help innovation, don't they ? ;-)
<roh>
wpwrak: do the steppercontrollers have isolation from the pc stuff?
<roh>
wpwrak: *kick* ;)
<wpwrak>
roh: the mill is a closed device, like a printer. no idea what isolation is inside. the interface is rs-232. DB25 connector, if you can imagine that :)
<wpwrak>
(patents) instead of standing on the shoulders of giants, we nowadays merely dance around them, trying not to step on their toes
<roh>
wpwrak: sure.
<wpwrak>
grmbl. the ac leak is in the mill's supply. 19 V, 2.1 A. this will be fun to find.
<wpwrak>
ah no. common laptop voltage, it seems. good :)
<wpwrak>
alright. no more milling until monday then.
<kristianpaul>
wpwrak: taling about GSM Networks, in Argentina, autotirezed freqs are  PCS1900 and GMS850? or is following europe regulations maybe?..
<kristianpaul>
s/autotirezed/authorized
<wpwrak>
the last job produced a funny result. i did two boards at a time and forgot to check the depth settings. the 2nd basically doubled the depth, and milled right through the pcb acting as a support and into the table. amazingly, nothing broke.
<wpwrak>
kyak: does openwrt let you set options depending on other packages ? e.g., if readline is enabled, then use --with-readline in bc, otherwise don't
<bartbes>
hmm I guess it should be possible
<xMff>
you can try this: CONFIGURE_ARGS += $(if $(CONFIG_PACKAGE_libreadline),--with-readline)
<xMff>
you'll need a conditional dependency as well then,  DEPENDS:=+CONFIG_PACKAGE_libreadline:libreadline
<bartbes>
why the change in syntax?
<xMff>
hm?
<xMff>
its not change, just an extension
<xMff>
+foo:bar  -> only depend on bar if foo is enabled
<xMff>
thats for opkg later on
<bartbes>
oh
<bartbes>
that's passed on straight to opkg?
<xMff>
no
<bartbes>
well then, why not use the if you used for CONFIGURE_ARGS for DEPENDS as well?
<xMff>
not straight, it gets translated to the appropriate metadata then
<bartbes>
or the other way around
<xMff>
because the foo:bar syntax has another effect
<xMff>
it ensures that bar is a build time dependency
<kyak>
wpwrak: just make it depend on libreadline, why all these difficulties?
<xMff>
well for this particular case it makes no sense
<wpwrak>
kyak: if you do it cleanly, upstream may accept it :)
<xMff>
you usually do it when packaging something with plugins, where foobar-mod-mysql should enable mysql or similar
<xMff>
but only when enabled
<bartbes>
aaaanyway
<wpwrak>
xMff: great, thanks !
<bartbes>
has anybody managed to compile gettext-full?
<kyak>
wpwrak: don't really care. they can go on ripping off things
<kyak>
bartbes: there is gettext in openwrt-packages
<xMff>
they = me
<bartbes>
kyak: gettext-full
<bartbes>
gettext is pretty much a dummy
<kyak>
bartbes: it IS gettext-full
<bartbes>
I mean the package
<bartbes>
not the lib..
<bartbes>
the package gettext-full doesn't compile
<xMff>
not true, its ony my list, just haven't gotten around to test it
<bartbes>
I just updated
<bartbes>
...
<bartbes>
sigh
<bartbes>
gettext is a *stub*
<bartbes>
it doesn't even have a lib
<bartbes>
(or well, it doesn't install one)
<kyak>
well, yeah.. to bad, because we haven't been able to make a normal build since the beginning of December
<kyak>
first because of gettext, then libiconv and gettext
<wpwrak>
xMff: would you happen to be kyak's openwrt upstream ?
<bartbes>
gettext-full doesn't compile against libiconv-full even
<xMff>
yes, someone was on an update spree without fixing the usual fallout
<wpwrak>
xMff: (upstream) kewl ! we found you after all ! ;-)
<wpwrak>
xMff: (update spree) you're talking about the libiconv issue ?
<kyak>
:) time to sleep..
<wolfspraul>
n8
<xMff>
the issue for the whole iconv / intl issue is a bit.. complex
<xMff>
erm *reason for
<xMff>
it all begun with an attempt to update libtool...
<wpwrak>
xMff: yeah, i saw the argument go back and forth, without getting the impression that both sides understood each other fully :)
<xMff>
I noticed that there is a need for switchable intl and iconv implementations
<xMff>
however I've not yet worked out a way to support that reliably
<xMff>
my goal is to couple that to the existing native language support global build option we already have
<xMff>
which is right now nothing more than "--disable-nls" for configure
<xMff>
which is ignored by ~90% of the packages anyway these days
<xMff>
the reasoning for stripping intl and gettext was that the liberal use of autoreconf brought those dependencies in for many packages where they didn't exist previously
<wpwrak>
if i understand things right, then some packages will not build at all without the *-full versions ?
<wpwrak>
(i don't even know what autoreconf is ... :)
<xMff>
most should build now, the gettext is still stubbed (no *.mo handling etc. at all) and the mini iconv now supports the same range of charsets the full iconv does
<wpwrak>
so kyak's worries should be over soon
<xMff>
"full iconv" was - despite its name - also stripped
<wpwrak>
heh :)
<bartbes>
yeah, it misses iconv_open, iconv and iconv_close
<bartbes>
and I would say those are pretty.. critical
<xMff>
bartbes: that is gnu madness, somehow they use libiconv_open, libiconv and libiconv_close and then alias those with #define
<wpwrak>
btw, how is the effort around the ben perceived by the openwrt "core" team or whatever you have ? is making it a more general distribution considered a good direction or are there concerns, e.g., because it may overload the configuration process ?
<xMff>
bartbes: if a wrong iconv header is picked up it falls apart
<bartbes>
from what I've heard they're mostly ignoring 'us'
<wpwrak>
bartbes: that's what kyak says. let's see if there are more sides to this story :)
<xMff>
well I assumed larsc and/or mirko are the openwrt liaison here
<xMff>
we're aware of the nanonote developments but I'm not personally involved with it (yet)
<wpwrak>
xMff: not sure how active they are in that role. didn't actually know larsc was an openwrt liaison. he's a man of many talents :)
<kristianpaul>
Is kind of worryingng read this " the other hand side we already have enough overriden packages, and perhaps more to come (like binutils, for instance https://dev.openwrt.org/ticket/8603).";
<kristianpaul>
just saying
<wpwrak>
xMff: the people i see most active around here are kyak, david kuehling, and xiangfu
<xMff>
okay
<xMff>
so I smell a communication issue ;)
<bartbes>
actually I remember larsc helping me get a patch upstream
<bartbes>
I wonder what it was for..
<bartbes>
oh I remember, ncurses
<xMff>
tbh, the openwrt is, probably like any project out there, understaffed, especially with regard to maintaining the packages feed. I spent a few weeks cleanung up libtool fallout in packages I don't even know what they're good for
<bartbes>
I fixed most of the efl packages
<bartbes>
and I got them running
<bartbes>
but gettext is holding me back from compiling enlightenment itself
<bartbes>
and that sucks
<wolfspraul>
xMff: in which way do you want help to reach you? write up tickets? send patches?
<xMff>
what is needed to accomodate the NanoNote / xburst needs are some changes in the build infrastructure
<xMff>
wolfspraul: definitely patches
<xMff>
wolfspraul: ven maintainership
<xMff>
*even
<wolfspraul>
if it's clear what you want, and you or other openwrt folks have consensus that that's what's good for openwrt, it should be possible to organize help for you
<xMff>
right now we need some way to solve the gettext and intl issues
<wolfspraul>
what slows things down sometimes seems to be how to resolve conflicting priorities
<wolfspraul>
xMff: well there are several people in this channel who want to help with that, the question is how, and whether their help is appreciated on your end, or causes more problems to you
<xMff>
personally I wonder why we maintaing the desktop feeds in openwrt at all
<wolfspraul>
you have a much broader view, we are very NanoNote specific here, so I assume what we think is 'great' others may very well find quite disturbing :-)
<xMff>
so it might sense to give up openwrt mini versions of packages that are only used on "capable" targets anyway
<wolfspraul>
so for example this latest bc thing, a configuration parameter that will enable line editing and edit history
<xMff>
*make sense
<wolfspraul>
it's a very easy thing to enable or disable
<wolfspraul>
but what are the priorities for people? who decides?
<wolfspraul>
enabling it will make the binary bigger (I would think), yet make the app more usable (to a very large degree some people would say)
<xMff>
the one who does the work
<wolfspraul>
ok but at some point someone decided to leave this disabled for bc
<wolfspraul>
that may have been intentional, or someone just disabled everything that was not very clear at that moment
<xMff>
yes, our agenda is to throw everything out that is not really required
<wolfspraul>
sure, but maybe the person deciding what is not required is not actually using the app?
<xMff>
a majority of openwrt users is wroking with free space in the 900k-1400k range
<wolfspraul>
sure sure
<wpwrak>
wolfspraul: regarding bc, i think enabling readline iff readline is already enabled in the system is likely to be considered the right choice. not sure how to handle this with packages, though.
<wolfspraul>
I totally understand that, and I very much like this philosophy in openwrt.
<wolfspraul>
it's a great value that was built up over time.
<xMff>
I just try to explain the perspective
<wolfspraul>
of course, fully understood, and appreciated
<wpwrak>
xMff: so where does the stuff you're throwing out go ?
<wolfspraul>
I try to understand the process
<wolfspraul>
we run into a little app like bc
<wolfspraul>
do you know bc?
<wolfspraul>
it's a simple calculator
<xMff>
yes, I know bc.
<wolfspraul>
we find some config options to be really bad (let's just assume this now, I don't mean this particular line editing thing)
<wolfspraul>
we can always make a new package, but that increases maintenance overhead, and we would rather help working towards a stronger openwrt in general
<wolfspraul>
or write up a ticket?
<wolfspraul>
some people seem to like tickets a lot
<xMff>
I think we're not at the point where tickets would really help
<xMff>
some fundamental directions should be worked out first unless we want to discuss architectural stuff in the tracker
<wolfspraul>
how do you find the maintainer of a package?
<xMff>
easy: is there is MAINTAINER variable, you know the maintainer. If there is none, the package is unmaintained (kept compilable only)
<wolfspraul>
I think it's hard to discuss some specific thing about some package if the person to talk to us not using that software himself.
<wolfspraul>
how do you avoid this?
<xMff>
for new package submissions we urge the authors to do maintainership
<wolfspraul>
ok thanks that's helpful!
<xMff>
for others the original packagers might be long gone
<wolfspraul>
from 132 packges in openwrt-package, 3 have a MAINTAINER entry :-)
<wolfspraul>
so that gives me 129 little todo items to work on...
<wpwrak>
better than nothing ;-)
<wolfspraul>
xMff: if someone runs into a package without MAINTAINER entry, does this mean he can just adopt this package and fix/improve it in the way he wants?
<xMff>
imho, all graphic software should be kept in a spearate feed
<wolfspraul>
if so, where/whom would one to contact to put onself in as the MAINTAINER
<xMff>
wolfspraul: yes. it is a heavily used package, there might a clash of interests
<wolfspraul>
would one have to...
<xMff>
hm, I'm lacking words today
<wolfspraul>
sure, understand. but I first try to understand the general principle.
<xMff>
okay
<wolfspraul>
I mean that's a pretty clear concept
<xMff>
so right now we're looking out for maintaners
<wolfspraul>
no MAINTAINER, and you think you can improve something - ask to be put in as MAINTAINER, and then submit the patch right away, and it will be applied
<xMff>
basically anyone can ask for maintainership on a specific package, usually there is also the usualy "send a few patches, show interest etc."
<xMff>
to see the quality of contributions
<wolfspraul>
bc has no MAINTAINER entry
<xMff>
yes, so you could ask to maintain it
<wolfspraul>
he :-)
<wolfspraul>
I just looked at the 978 OpenWrt packages I have on my system
<wolfspraul>
71 of them have a MAINTAINER field in the Makefile
<wolfspraul>
that means 907 don't
<xMff>
yes, those accumulated over the years
<xMff>
we also have no overview over what is actually used and what not
<xMff>
I suppose we could check the repository access logs
<wolfspraul>
alright. so what I've learnt now: MAINTAINER field is significant.
<wolfspraul>
if it's not there, just assume you are the only one and request that your patch gets applied
<xMff>
yes
<wolfspraul>
in general - submit patches to mailing list (for packages with and without MAINTAINER field)
<xMff>
yes
<wolfspraul>
if a patch is sent to the mailing list, and nobody replies, what to do then?
<xMff>
keep pinging
<xMff>
don't overestimate the manpower
<wolfspraul>
I don't, I just try to understand in which way contributions are welcome.
<wolfspraul>
so it helps the overall project
<wolfspraul>
otherwise I can just say give me write access everywhere and all my problems are gone.
<wolfspraul>
but then maybe others will have new problems quite soon :-)
<xMff>
yes
<wolfspraul>
xMff: keep pinging how? also on the mailing list?
<xMff>
yes
<wolfspraul>
alright :-)
<xMff>
or maybe drop by in #openwrt or #openwrt-devel
<wolfspraul>
are you hanging out here in irc sometimes?
<wolfspraul>
yes I think that's already happening too
<xMff>
well I used to hang around here until a few months ago
<xMff>
then got distracted with other stuff
<wolfspraul>
no problem it's great you are back now :-)
<wolfspraul>
I have a clear goal already - MAINTAINER field in the missing 129 packages in openwrt-packages :-)
<wolfspraul>
3 out of 132, that's even worse than the 71 out of 978 in the rest...
<wolfspraul>
so let's try with the bc patch whether enabling line editing is welcome or not
<xMff>
there are a couple of facilities in the buildroot that aid with packaging multiple variants of the same package
<xMff>
we currently use that for hostapd for example
<wolfspraul>
the only thing I'm worried about is if there are people who just say 'no' in principle, to every feature they don't understand, even if they don't use the app.
<xMff>
something like that would be needed for packages where small and full featured versions make sense
<xMff>
others don't make sense to be stripped down at all, like the xorg stuff
<wolfspraul>
not that I would complain about that kind of attitude either, but it would cause a problem for us, because in many cases we have someone who is really passionate about a particular app, and then emotionally feels like "if abx is not enabled that's really really stupid" etc. etc.
<wpwrak>
wolfspraul: i wonder if this may be a moment to consider the general strategy. if openwrt upstream prefers most of the little nanonote things to be kept separate, the maintenance burden remains on the side of the qi-hw community. for potentially quite a lot of packages. is this a desirable outcome ?
<wpwrak>
(i'm assuming here that existing communication problems can be solved, etc.)
<wolfspraul>
and that may clash with someone who has to take care of 500 packages and is just pissed about all this small feature fixes
<xMff>
wolfspraul: thats definitely not the case
<wolfspraul>
ok great
<wolfspraul>
well I learnt a lot - thanks! let's see how it goes :-)
<xMff>
there are three types of things from my pov:
<xMff>
- packages that are only relevant for capable targets - those can be kept in a spearate feed, lowering the merge overhead for everyone
<xMff>
- packages that have users on small targets - we need build variants here to be able to turn of extra features
<xMff>
- stuff that needs support in the build infrastructure, this one must be worked out with the rest of openwrt to not break existing stuff
<xMff>
the whole desktop and multimedia stack is giving me headaches in particular as it requires a farily complete autotools infrastructure
<xMff>
and we're not quite there yet
<wpwrak>
xMff: is this something where you want to go, or would you be happier if there was no urge to enter this direction ?
<wpwrak>
xMff: (of course, it could just be an interesting intellectual exercise per se :)
<xMff>
even if I do not use the affected packages I'm more than willing to hack on the build infrastructure support
<xMff>
we now have native autoconf, automake, m4, libtool etc.
<wpwrak>
great, thanks !
<xMff>
what is missing is the i18n infrastructure, it is totally underdeveloped on openwrt
<xMff>
we neither have build support for it nor do we package stuff like translation files
<wolfspraul>
wpwrak: the thing I personally care the most about, for my own NanoNote, is boot time. the 12-14 openwrt image I am using, and very happy with, boots in ca. 17 seconds.
<wolfspraul>
that's acceptable, but I hope we can bring it down. 15, 13, 10, etc. it's hard.
<wolfspraul>
the discussion with xMff was for me to understand how we can help openwrt get better in general
<kristianpaul>
My router still taking about ~15 to boot.
<xMff>
another dream of mine is supporting versioned dependencies and conflicts in buildroot
<xMff>
recent opkg supports that
<wolfspraul>
feedback may come in the right or wrong form.
<wpwrak>
xMff: btw, are there other active "desktop" targets for openwrt than the nanonote ?
<xMff>
wpwrak: hmm, afaik not. well you have always generic x86 which can host anything, but the nanonote is the first target really made for desktop
<wolfspraul>
freerunner
<wolfspraul>
that good old thing :-)
<kristianpaul>
What about dingo?
<xMff>
yes, but I always percived that one as proof-of-concept quality
<xMff>
(within openwrt)
<wolfspraul>
wpwrak: I had a great meeting today with Chitlesh Goorah from Fedora/Free Electronic Lab
<wolfspraul>
learnt a lot about ASIC design flow, file formats, etc.
<wolfspraul>
one thing that stood out was he said synopsis/cadence/mentographics all use tcl scripting
<wolfspraul>
tcl seems to be the de-facto standard in that scene, so I will take a quick look whether, if we do more cmdline/scriptable stuff in kicad, we can do something towards tcl as well?
<wolfspraul>
just an idea, I need to check it out...
<xMff>
so if you guys could provide me with a list of packages that are affected by iconv and intl, I'll see how hard it would be to make them switchable
<wpwrak>
wolfspraul: my enquiry is two-pronged: 1) find confirmation/contradiction for my gut feeling that openwrt may not be the best long-term solution for the nanonote, and 2) in case you stick with openwrt as the "standard" distribution, how to make things move smoothly
<wpwrak>
wolfspraul: tcl may just be a historical choice
<wolfspraul>
yes but it seems to be a standard there, well that's what I've learnt at least
<wpwrak>
wolfspraul: probably back when scripting became interesting (and it would become interesting for all of them at about the same side - can't fall back wrt the competition), tcl was the scripting language of choice
<wpwrak>
wolfspraul: nowadays, it would probably be python
<wolfspraul>
yep
<kristianpaul>
Hmm Quartus from Altera uses tcl i think  lekernel can cofirm?
<wolfspraul>
he said that's exactly what they all hate :-)
<wpwrak>
xMff: i think we'll have to wait for kyak & co. to come back for that list
<wolfspraul>
anyway, I'm just listening. I will find out more.
<wpwrak>
wolfspraul: they all hate python ? heh :)
<wolfspraul>
xMff: yeah we have to wait for kyak who for sure will have some apps
<xMff>
wpwrak: I think it will boil down to some $(LIBICONV) / $(LIBINTL) macros which contains appropriate cflags, ldflags and then we need to annotate all packages with that
<wolfspraul>
wpwrak: I don't know what to say about 'long term solution'
<wpwrak>
let's make it lisp then. that's timeless :)
<wolfspraul>
my brain doesn't function like this, I'm so practical
<kristianpaul>
wolfspraul: did Chitlesh Goorah metioned cases of study for FEL in the industry?
<wolfspraul>
we have reflash software, we have openwrt images, jlime images, debian images
<wolfspraul>
everything is in flux, hopefully improving
<wolfspraul>
if I try to get something done, and right in front of me, on my ben, it won't work with openwrt, I might just try jlime and see whether that works.
<wolfspraul>
the things I care about are stability / avoiding regressions / test plans / boot time.
<wolfspraul>
that's the 'long term solution'
<wolfspraul>
have a good test plan, avoid regressions, make updating easy, strictly control boot time
<mth>
wolfspraul: we've got Tcl scripting in openMSX (MSX emulator) and we have some mixed feelings about it
<mth>
we're very happy about the scripting itself, but the Tcl language is a pain when you want to do more complex things
<wpwrak>
xMff: sounds reasonable. when generating packages (ipkg), one would then have to choose one or the other at build time, correct ? i.e., one couldn't pick at install time (not sure if there would be a lot of demand)
<mth>
it's fine for simple things though
<wolfspraul>
what other scripting options are there?
<mth>
lua, python, javascript
<wolfspraul>
kristianpaul: not really, only universities.
<xMff>
wpwrak: there will be a menuconfig flag in global build settings which says "I18N implementation: stubbed/full" or similar
<mth>
python might be harder to embed than for example lua
<xMff>
that then is propagated down to all packages
<wolfspraul>
synopsis is most closed in his opinion, cadence doesn't care about the hobbyist stuff, mentorgraphics wants to point people who use pirated mentorgraphics software to free tools
<mth>
blender has python embedded though, so it is possible
<wpwrak>
wolfspraul: (long term) i suffer each time i see someone happily announce the port of something that's almost certainly been in jlime for a decade already
<wolfspraul>
yes and no. it's like comparing gentoo and debian.
<wolfspraul>
why don't we try to reconfigure jlime a little :-)
<wpwrak>
wolfspraul: it's inefficient channeling of resources. imagine what these people might do if they didn't spend all their energy on plumbing.
<wolfspraul>
you are comparing apples and oranges
<wpwrak>
wolfspraul: (reconfigure jlime) what do you have in mind ?
<wpwrak>
wolfspraul: (apples/oranges) maybe. hard to tell.
<wpwrak>
wolfspraul: sometimes you have people who just want to do something useful. if they see a great lack of packages, they tackle that.
<wpwrak>
wolfspraul: if they'd see some other, perhaps more interesting issue, they would go after that
<wolfspraul>
openwrt is a tool to build custom images from scratch
<wpwrak>
wolfspraul: of course, it all depends. harder/different issues may scare people away who would feel comfortable with adding packages to openwrt.
<wolfspraul>
for people who like custom images
<wolfspraul>
openwrt is not a meta distribution, imho
<wpwrak>
wolfspraul: but then, having a richer "standard" distribution may also attract more people and may encourage more contributions
<wolfspraul>
to me (correct me if I'm wrong), it's like embedded gentoo (I never used gentoo so I'm guessing a little)
<wpwrak>
xMff: (metaflag) okay. keeping it free from too much headache :)
<wpwrak>
mth: (lua) haven't played with it yet
<wpwrak>
mth: (python) embedding it is quite popular and python-c bindings are relatively straightforward
<xMff>
I found Lua very easy to work with, especially when interfacing with C
<xMff>
but then I only wrote like ~500 LoC of python in my life
<xMff>
and the lua library ecosystem is.. hm a bit sparse
<wolfspraul>
calling it a day, 'night
<xMff>
definitely no batteries included :)
<wolfspraul>
xMff: thanks for your advice and offers of help!
<wpwrak>
xMff: (sparse) which may be good or bad ;-)
<xMff>
np. night
<xMff>
wpwrak: yeah, at least you don't have to deal with bloat or frameworks of which you don't need 90%
<xMff>
and on x86 there is luajit which offers remarkable performance
<mth>
I think lua fills the same niche as Tcl, but is more modern and better designed
<mth>
python is a full programming language that can be used for scripting, rather than a scripting language
<xMff>
yeah, that about sums it up I think
<wpwrak>
often enough, i just design my own scripting language. it's fun and keeps the blood flowing :-)
<wpwrak>
but lua definitely seems like something worth looking at
<xMff>
reminds me on my failed attempts to implement a javascript interpreter in perl
<wpwrak>
that sounds evil ;-)
<wpwrak>
speaking of which ... i should still have a latex processor in perl somewhere ....
<xMff>
it was fun, learned basics something about parsers and stuff like that. Of course I never finished it
<wpwrak>
perl must be one of the most hostile environments for this kind of stuff ;-)
<wpwrak>
well, /bin/sh would be worse ...
<xMff>
yeah, see libtool :D
<wpwrak>
oh, nice :)
<roh>
i find python a nice language for scripting and see bindings a lot in foss tools
<roh>
its also well suited for integration it seems. (atleast so i am told)
<mth>
I use python a lot for prototypes and for utilities where performance is not critical
<roh>
tcl is ugly and old.
<mth>
I have also written a medium size web app using python and twisted
<roh>
mth: ack. performance stuff should be native.
<mth>
for text conversion, I got better performance from python than from java, by the way
<mth>
I guess the python libs (especially I/O) are better designed or better optimized
<roh>
thats also how all the tools i know do it. provide some bindings into python for native implemented routines
<wpwrak>
you need to skip things it doens't understand (like most of the header) with %beginskip ... %%endskip
<wpwrak>
err, %%beginskip
<roh>
if it needs to be very small and embedded i guess lua is the choice (size in ram and resident memory)
<roh>
ciscos also still use tcl for scripting... resulting in horrible hacks as soon as people want to do advanced stuff
<wpwrak>
isn't it amazing how completely tcl has died out (except for these odd legacy cases)
<roh>
i think openwrt did the right thing by choosing lua as tier one scripting language. for desktops and such python is imho handier, (and the 5-10mbyte interpreter size doesnt hurt there anyhow)
<roh>
wpwrak: hrhr.. yes.. i mostly know it from legacy and 'techie campfire horrorstories'
<wpwrak>
oh, you get used to it after a while. particularly if you know lisp :)
<wpwrak>
what's a bit annoying is that the usual built-ins always differ a bit between perl, tcl, python, etc.
<wpwrak>
so while i'd consider python a nicer language than perl in general, i find it a lot easier to get small to moderate tasks in perl done, because perl is the first scripting language of that class i learned
<roh>
wpwrak: the most stuff from the tcl examples are between 3 and 10 lines in python
<roh>
while having >20 in tcl
<wpwrak>
oh, they all would be just one line in APL ;-)
<roh>
learnt basic.. later awk. and lately (few years ago) python (as much as i need)
<roh>
but i put asm and C inbetween (and all that other s*) ... still dont like perl (i wish there were ellbow-long rubbergloves for touching code)
<wpwrak>
perl is a great write-only language :)
<urandom__>
python was one of my first programming languages but today i always prefer using lua instead
<roh>
wpwrak: ack.
<larsc>
"why is there no obfuscated perl programming contest? because everybody would win"
<roh>
urandom__: lua is definitively on my list (after getting better in python) .. havent't had a reason to dive in hard to be fair
<wpwrak>
i started with basic, then machine code (we didn't have assemblers back then), pascal, assembler, then finally C (but didn't like it at that time), more pascal, then C+Unix, csh, sh, perl, and so on
<wpwrak>
larsc: ;-)))
<roh>
uh. yes.. pascal. that was beween/around x86 asm and c
<urandom__>
lua has tables, they are like arrays but much more powerfull, you can put every type of data in it and mix freely
<xMff>
real fun starts with metatables
<urandom__>
yeah
<urandom__>
pretty sad other languages dont offer anything comparable to it, i really dont know any language that does
<roh>
urandom__: i thougt that python is similar there
<wpwrak>
what do metatables do ?
<roh>
with the tuple stuff
<xMff>
its a bit like operator overloading
<xMff>
you can attach a metatable to an object and have it handle stuff like assign, lower than, equal comparisations etc.
<xMff>
or make stuff callable
<mth>
you can overload most operators in python, you just have to know which method names to use
<urandom__>
roh nah tuple arent comparable i think, they are ummutable
<mth>
it sounds like a lua table is more like a python dictionary
<xMff>
yes
<mth>
a key-value mapping
<xMff>
you cna treat it as array though by using only integer keys
<xMff>
or you can make it having an array part and a dictionary part
<kristianpaul>
larsc: i agree !
<kristianpaul>
I have to deal at work with some sofware wich is reconfiguref by perl scripts, thats a PITA when something fails.. and debug is required :/
<kristianpaul>
Lua is nice, i think i'll focus more on it as i'm doing with C know, lua seems to everywhere from routers (Lucy is written in lua right?) to network tools (nmap), also window managers (awesome)
<kristianpaul>
s/know/now
<kristianpaul>
python is cool but not so portable as i wish, i think lua is in the middle between C and python perhaps..
<urandom__>
python is much slower and bloated
<kristianpaul>
yeap
<kristianpaul>
hey, i found some stuff for SIE in lua, i'll take a look later
<roh>
yes. if you have small bare metal lua is good. if you got a linux kernel and >=32mb ram and similar flash or more, you can also use python. fast stuff will be native in both (if done right)
<kristianpaul>
roh: i tought you where about to said "if you got a linux kernel and >=32mb ram and similar flash or more, you can also still using lua" ;-)
<roh>
kristianpaul: sure. just more pain.
<kristianpaul>
why?
<kristianpaul>
is lua :_)
<roh>
imho. less bindings.
<urandom__>
roh if you are targeting x86 you can use luaJIT which is a lot faster
<kristianpaul>
ah well
<kristianpaul>
thats true
<roh>
in python it feels like everything is there and often already for quite some time. 'well hung' code.. stable apis etc.
<kristianpaul>
I guess python had time with few bindings compared as we know it today
<kristianpaul>
roh: besides that i cant run python on my linksts router unless i add it swap :/
<roh>
in lua i thought about opening a socket and the serial and stopped researching after an hour and hacked up a 20 line python script (also learning the api underways) which is in production use till today in 30 minutes.
<urandom__>
lua has less bindings? sure? i am not
<roh>
kristianpaul: maybe. but you also have to see that lots of linksys routers are 'old crap' now (>5 years old stuff)
<kristianpaul>
pyton for production real stuff? wow, how do you fight against threads?
<kristianpaul>
roh: sure buy i love my crap routers :-)
<roh>
recent plastic boxes have that much ram and rom that i dont care anymore. but sure. you are right. if it needs to fit into somthing tight and you know what to do... / somebody pays for the extra worktime... ;)
<kristianpaul>
Are so trusty and usefull
<kristianpaul>
s/buy/but
<roh>
kristianpaul: i only use the linksys stuff as vlan-switches with ssh managemant now
<roh>
became to slow for recent dsl lines
<roh>
(nat)
<kristianpaul>
what's a recent dsl line in your side of the word btw? (10Mpbs may be?)
<roh>
16/1 adsl2+ is 'normal'
<roh>
one can get vdsl with 25 or 50 mbit down, but that still expensive
<kristianpaul>
is happy with his 1Mb down 500k up
<kristianpaul>
Anyway i jsut replaced my router for a soerkis box, ready to receive adsl2+ someday...
<kristianpaul>
wpwrak: about atusd case, did you find a parametrized way of doing case design?
<roh>
heh.. yes.. got something similar in use here also. a pcengines wrap2c
<kristianpaul>
what about atusb too?
<roh>
basically a pc without fan or keyboard controller on one 16x10cm pcb.
<kristianpaul>
Your molding idea is nice, but on my side i will like to print the case fot the atusd/usb board i hope to buy in some moths :-)
<roh>
wpwrak: how many atusb and atusd boards do you have? i'd like to play around a bit with them trying out some case ideas
<kristianpaul>
roh: to be more specific my router is a soekris and the servr an alix
<roh>
kristianpaul: :) the pcengines stuff is really great hw. rocksolid
<kristianpaul>
roh: parametric is very important, are you good with openscad?
<kristianpaul>
roh: yeap
<roh>
nope. i tried openscad some times, but it was always slow as hell for me and pain to get working (compiletime-brainfuck)
<roh>
had to build all kinds of libs with broken buildsys from hand (not packaged at all) and fix weird issues to get it working
<roh>
why does it need to be parametric?
<roh>
will the boardsize change so often?
<kristianpaul>
heh, nope but just some time ago i tried do an unoffiacial atempt for a atusd case
<kristianpaul>
I jsut was so messy had to modify by hand the design
<kristianpaul>
After i printed it out measure, design, print....
<kristianpaul>
also wpwrak suguestions..
<kristianpaul>
is a chaos !
<roh>
since i know how to use it quite well i really like my qcad. reliable tool
<kristianpaul>
ya, qcad is nice tool
<kristianpaul>
for the cnc owners... ;)
<wpwrak>
kristianpaul: (parameterized) not sure ... openscad goes in that direction but i'm not sure if it's really good enough for this
<kristianpaul>
maybe a script will be enought
<wpwrak>
roh: (#boards) you mean right now ? something like 6 atusd and 4 atusb. all prototypes different from the final ones. some quite drastically.
<wpwrak>
roh: (qcad) ever tried fped for cad work ? :)
<wpwrak>
roh: (parametric) the thing is that *any* change is a pain with non-parametric cad ...
<roh>
wpwrak: parametric only works well for 'simple' stuff in my experience
<roh>
like 'boxes' .. and then you take the output and add the 'special cases'
<wpwrak>
povray ? :)
<kristianpaul>
atusd shape is simple i think
<roh>
scripting these all is hard work.
<kristianpaul>
roh: gears too
<wpwrak>
scripting is hard, i agree
<roh>
wpwrak: maybe you can send me some mechanical prototypes as soon as the mechanical size seems stable enough. i have no issues moving lines in qcad.
<roh>
need to do that anyhow while developing. tolerances and manufacturing is a bitch.
<wpwrak>
i wish there was something like fped for 3d ... i tried once to do a few drawings with qcad, but found it so awkward that i then tried to see if i could abuse fped to do this. worked like a charm.
<wpwrak>
roh: (prototypes) okay, i'll try
<kristianpaul>
fped will be nice indeed
<wpwrak>
roh: (tolerances) that's exactly where parametric shines :)
<roh>
wpwrak: yes.. 3dcad is hard... i still hope heeks will solve that gap for me.
<roh>
else.. there is freecad. and heeks can import the stl and do the cam on it also
<kristianpaul>
heeks is promising, also heekspython
<roh>
designing is still hard or atleast difficult with heeks imho.
<kristianpaul>
the atempt i nade for atusd was in heekscad
<kristianpaul>
roh: lets pay a blender course, after that nothing all will change
<wpwrak>
roh: (prototypes) btw, the layout files (.brd) should give you a pretty good idea already
<kristianpaul>
print it !
<kristianpaul>
or drill/cut, will be nice experience
<wpwrak>
roh, kristianpaulL: can you print/cut from kicad ?
<wpwrak>
s/L//
<roh>
wpwrak: ?
<kristianpaul>
kicad export to dxf?
<kristianpaul>
i think i can
<roh>
kristianpaul: blender sucks for cad. it doesnt even have a ruler.
<kristianpaul>
ah yes
<roh>
same as inkscape. designer tools. not technical cad
<kristianpaul>
i used the herber from kicad in heeks
<kristianpaul>
wpwrak: but my printer is to 3d too limited for those small things
<wpwrak>
roh: i mean, given a kicad design, can you cut a pcb ? (the outline, not the traces)
<kristianpaul>
2d printer is enought/was in that time
<wpwrak>
kristianpaul: hmm yes, you need about +/-0.5 mm accuracy, better if possible
<roh>
ah. and caduntu got renamed librecad
<roh>
it is in essence, a ported to qt4 version of qcad
<kristianpaul>
wpwrak: i have that
<roh>
wpwrak: i have no kicad installed right now in a qi-compat. patched version
<kristianpaul>
remenber the pics from the case atemtps?
<roh>
wpwrak: need to work myself into it first. i know eagle... but also havent done pcbs with it yet.
<wpwrak>
roh: the normal version should be fine as along as it's recent
<wpwrak>
>= ~bzr 2685
<roh>
wpwrak: nah.. my build is from 2008
<wpwrak>
urgh :)
<kristianpaul>
roh: who needs rules, you cane measure the vector it self :-)
<wpwrak>
kristianpaul: yeah. okay, so you could print the pcb as well :)
<wpwrak>
off to the chemical lab for a bit ...
<roh>
hates it when he needs to use custom toolchains all the time. that damn unpractical when one uses N instead of 1 computer.
<kristianpaul>
actually i tought do it but never..
<roh>
kicad is on my list of todos...  ;) i just hoped there would be fresher packaged builds
<kristianpaul>
in fedora may be
<kristianpaul>
Build: (2010-05-09 BZR 2366)
<roh>
hm,. there is a ppa with bzr2720
<roh>
is still on karmic
<roh>
add-apt-repository ppa:paxer/ppa
<wpwrak>
ti-tick ti-tick ti-tick ... toner transferring to the 2nd side ...
<wpwrak>
oh, and the witchbrew that's supposed to wake me up in the morning has been in the freezer for quite a while. better not to take that "ice" in "ice coffee" too literal ...
<wpwrak>
still mostly liquid. good.
<kristianpaul>
wpwrak: how is the process for you pcb in order to join both sides of it?
<kristianpaul>
I really like 2 layer stuff, i think this year i'll try ro get my 0.8 pcb
<kristianpaul>
s/ro/to
<wpwrak>
kristianpaul: toner transfer to the first, cover the second with adhesive tape, etch, remove the tape, clean, toner transfer to the 2nd side, tape the 1st side, etch again. done :)
<kristianpaul>
bo you need make holes i guess?
<wpwrak>
i now do al the cutting and drilling with the mill
<kristianpaul>
drill with hand tools or your cnc thing?
<wpwrak>
manual drilling results in too many broken drills as the number of vias goes up
<wpwrak>
all cnc
<kristianpaul>
oh nice
<kristianpaul>
I need a cnc
<wpwrak>
i'm actually writing a description of the process ... maybe tomorrow i'll finish ...
<kristianpaul>
or a better way of hold th drill :)
<kristianpaul>
nice !
<wpwrak>
a cnc mill is something very nice to have :)
<kristianpaul>
later
<kristianpaul>
well maybe a tiny cnc , my space is some reduced here
<kristianpaul>
wpwrak: why you dint etch your pcb?
<wpwrak>
i do etch my pcbs ...
<wpwrak>
toner transfer is for the etch resist
<kristianpaul>
i mean the vias
<kristianpaul>
s/vias/path
<wpwrak>
well yes, i do etch.
<kristianpaul>
hmm, i'll way for you doc better :-)
<wpwrak>
1) mount a bare pcb in the mill. 2) drill all the small holes. 3) mill all the large holes. 4) mill the outline. 5) toner transfer and etching, see above
<wpwrak>
next lab round
<wpwrak>
mission accomplished
<wpwrak>
now ... a little protective tin coating, a bit of scrubbing, and i'm done for tonight ...