#qi-hardware IRC log for Friday, 2010-11-05

qi-bot[commit] Xiangfu Liu: adjust the rootfs partiton size to 512MB http://qi-hw.com/p/openwrt-xburst/f1c496200:02
mtHi all05:25
xiangfumt: hi06:13
kipjust wanted to know if there is any device that can b placed in a compute that can bypass security and monitor it etc11:40
kipsay if it got fixed and a sort of computer version of a listening but was installed somewhere. is this possible11:41
kipbug sorry11:41
kristianpaulThere is a QT hacker around??12:01
wpwrakhe suspects something, but he hasn't found it yet. heh, heh.12:01
kristianpauli need to find the right /usr/share/qt4/mkspecs/linux-g++/qmake.conf: No such file or directory12:02
kristianpaulthe right qmake.conf used in openwrt build process12:02
wpwrakkristianpaul: that was self-explanatory :)12:02
qi-bot[commit] Mirko Vogt: mark as broken since it does not compile http://qi-hw.com/p/openwrt-packages/f4bd1f212:06
qi-bot[commit] Mirko Vogt: deselect gnuplot since it does not compile (in our tree) http://qi-hw.com/p/openwrt-xburst/589ead112:08
rafakristianpaul: if you need the package which has that it is : qt4-mkspecs12:08
kristianpauli need wmake anyway..12:11
kristianpauli wonder if in can try compile it..12:11
kristianpaulargg dropbeat is easting lot of memory... buw who needs ssh with this emdebbed devices ???12:14
kristianpaulxiangfu: zear is aceptable to you consider that ssh is uncesary for the Ben, as there is no need for security and in the other replace the dropbear will give more memory and high data tranfer12:17
wpwrakkristianpaul: just say no to dropbear. like so many of those "light" programs (busybox, etc.), it has some nasty quirks if you have an even slightly more advanced use.12:17
zearwhat about me now?12:17
kristianpaulwpwrak: hmm12:18
wpwrakkristianpaul: "no need for security". yup, and 640 kB are enough ;-)12:18
zearssh it the way i use to log on the nanonote (typing the commands on nn's keyboard is a suicide) and to transfer the files12:18
kristianpaulsure all we do12:19
zearyou guys want to remove ssh?12:19
kristianpaul*i* just put the idea on the table12:19
zearwhat do you want to replace it with? telnet?12:19
wpwrakzear: only kristianpaul does. and he'll stop doing that once he's taken his medicine :)12:19
wpwrakzear: kermit ! ;-)12:20
kristianpaulokay remove this log and onbody said nothing about ssh12:20
zearwpwrak, ha! One learn something new every day12:20
zeari heard something about kermit, though never really checked it out12:20
kristianpaulbut i just remenber the usb gadget can emulate serial port over ethernet...12:20
Action: kristianpaul needs his medicine 12:21
wpwrakkristianpaul: if you dig deep enough, you'll even find something that multiplexes several sessions over a serial line. i think it had file transfer as well. can't remember the name, though.12:22
zearspeaking about the serial port, i have this little cool thingy rafa gave me during the jlime meeting (i believe it originates from wpwrak?), though i have no skills to solder it in :P12:22
wpwrakkristianpaul: of course, then there's also ka9q. that was very useful on linux, before we had tcp/ip in the kernel :)12:22
zeari don't know, a little thingy with usb connector that acts like serial port for kernel output12:23
wpwrakkristianpaul: (idbg) no no, it's some program. i actualy used it quite a lot but that was a looong time ago. 20+ years.12:23
zearah, you are talking about something else, nvrmnd12:23
kristianpaulidbg <- cool thingy ??12:24
kristianpaulah i guess fdti chips kill it ;-)12:24
wpwrakkristianpaul: (ftdi) dunno. i was planning to replace the mcu with one, but considering the lousy results i got with bit-banging, i'm less confident about this.12:25
kristianpaulyou'll need a really tiny and powerfull CPLD and solve all your problems ..12:26
kristianpaultalking about bit-banging12:26
wpwrakkristianpaul: small, powerful, cheap, with lots of built-in analog components, including an LDO. yes ;-)12:27
kristianpauli'm finding poke very usefull12:34
wpwrakkristianpaul: ah, you've overcome your fear. very good :-)12:35
wpwrakwhat is BLIT ?13:08
kristianpaulan event i guess13:09
kristianpaullekernel: is that pdf made off latex?13:09
lekernela small linux/free software event13:09
lekernelno, it's openoffice13:10
kristianpaulhey that freedom stack look bettyer with comments13:10
lekernelit's not the same...13:11
lekernelI merely copied the idea and adapted it13:11
kristianpauli see13:11
mthspeaking of events, anyone coming to t-dose tomorrow?13:11
kristianpaulnetworking is missing13:12
mthhttp://www.t-dose.org/ - open source event in Eindhoven (south-east of the Netherlands)13:12
wpwraklekernel: don't you want to copy the design tools also into the "<< Traditional >> schematics" box ?13:13
lekernelusually when you do a schematics entry, it's for making a PCB..13:14
lekerneland there are already a lot of things to say for a 10-minute talk13:14
kristianpaulahh thats why so short13:14
kristianpauli was about to ask for more sldes :)13:14
wpwraklekernel: (schem entry) yup. you could also just merge the two boxes. it's just a little odd that you duplicate the projects but not the tools.13:16
wpwrakkristianpaul: heh, me too. "where's the cool technology ?" ;-)13:16
lekernelwhat do you want to see, exactly? :)13:19
kristianpaulwell on my side i would like an quick SoC review13:20
kristianpaulnot too much tenical just the concept13:20
wpwraklekernel: nice ! let's hope you can recruit some contributors. we're still quite weak in the mechanical area. the only tangible results so far are the counterweights, a few PCBs, and Jane's bags.13:20
kristianpaulalso for the Milkydrop part13:21
lekernelkristianpaul, i'll demonstrate that anyway (exhibition, and talk if I have spare time)13:21
kristianpaulwell latelly USB development deservers a slide13:21
kristianpaullekernel: sure13:21
lekernelUSB is a detail (and USB sucks)13:21
kristianpaulok lets swich USB by RTEMS as sofware parts is requiren help13:22
wpwraklekernel: some feature list would be nice. it doesn't have to be very understandable, just impress :)13:22
lekernelRTEMS is 80% done by now13:23
lekernelit's not uClinux, you can move fast :)13:23
wpwraklekernel: (feature list) but if there's no time, you can just put such a list in the exhibition area. people will probably stroll by anyway.13:23
kristianpaulwhat people most asked you first about the MM One lekernel ?13:24
wpwrak"does it run linux ?" (-:C13:24
lekernelno, price and availability13:24
lekernelanyway for the "does it run linux" as long as they see busybox and /proc/cpuinfo they're happy13:25
lekerneli even get extra points for running it on the framebuffer console :)13:25
lekernelbut under this appearance, the uClinux port really sucks13:26
kristianpaulahh cool, on flickernoise you did it?13:26
wpwrakin what way does it suck ?13:26
kristianpaulunknow reboots?13:26
lekernelby my standards at last, theobroma systems (who did the original port to the LM32 arch for lattice, in which we fixed many issues already) says it's "production grade"13:27
lekerneldirty code13:27
lekernelmissing drivers13:27
lekernelmissing features13:27
kristianpaulheh better replace it ..13:27
wpwrakis this because of uclinux or just because of the port ?13:27
lekernelbecause of uclinux and gnu13:28
kristianpaulok but RTEMS uses gnu and works better, isnt?13:28
wpwraksee rms, that's what you get for insisting on gnu/linux - now people blame you for the linux bugs as well ;-)13:28
lekernelthere are probably 10 persons on earth who understand the binutils/gcc source...13:28
lekerneland when those pieces of crap break (as they often do when targeting lm32 uclinux) you're left with a huge mess that you need to fix13:29
kristianpaullets invite one of then to join your project !13:29
wpwrakrumor has it that gcc used to be an even bigger mess in the not too distant past ...13:30
kristianpaulsofware is gas as somebody else said..13:31
lekerneloh, and i'll spare you the details of building a software package that uses autoconf and libtool for lm32 uclinux...13:31
wpwraklekernel: you're preaching to the choir ;-)13:32
lekernelsometimes it's faster to throw that junk away and simply run gcc *.c -o executable13:32
lekernelthat's just ridiculous13:32
Action: wpwrak is a card-carrying member of the autotools haters club13:32
Action: lekernel too13:33
wpwraklekernel: only sometimes ? :)13:33
lekernelyes, because sometimes the GNU/Autocrap needs to generate a config.h that is later included by the program13:33
lekernelthose programs are the worst to port13:33
wpwrakseems that you need an autotools-hugger to join and clean up stuff, so that things will be easier in the future13:35
wpwrakso don't talk too badly about it :)13:35
lekernelI made my decision13:35
lekernelI switched over to RTEMS :)13:35
lekerneleverything went much smoother since then13:35
wpwrakNIH labs proudly present .... :)13:36
wpwrakNot Invented Here13:36
lekernelhaha, no, there are technical reasons13:38
wpwrakseems that my function generator is accurate to about 1 ppm. i like that :)13:39
lekerneland rtems isn't even that exotic13:39
wpwraklekernel: still .. think of the thousands of packages, say, jlime has. that's still a very long way.13:40
wpwraklekernel: so we should get rafa a MM1 board. then he can solve the obstacles that prevent jlime from building there.13:40
lekernelsure, that's why I leave it to other people (and not myself alone) the care of porting them and beating autocrap/linux/gcc into place13:40
wpwrakrafa, how about learning a bit of gcc, binutils, and autotools internals ? ;-)13:41
wpwrak(well, you probably had your share of fun with autotools already :)13:41
lekernelbesides, linux and x11 are slow13:41
lekernelquite frankly, the few successful Linux ports that i've seen have been done by companies like Codesourcery13:43
lekernelwho probably benefit a lot from the GNU obscurantism :)13:43
wpwrakhmm. if you get linux to compile and run, your toolchain can't be all bad.13:44
lekernelok. then, what would you use as GUI toolkit?13:46
rafawpwrak: sorry, to learn for what? I am not following13:46
wpwraklekernel: whatever works ? :) personally, i like gtk, because it plays nicely with C13:48
lekernelbut requires a bunch of crazy dependencies: glib, cairo, x11, ...13:49
wpwrakrafa: learn about the dark side of gcc and friends to port jlime to the milkymist processor :)13:49
rafawpwrak: ah.. no following.. so there is not a gcc for milymist processorĀ”?13:50
wpwraklekernel: but how many of those things are really a problem one you've properly debugged your toolchain ?13:50
kristianpaulrafa: it is13:50
lekernelwell, first they're slow13:51
kristianpaulrafa: just current uclinux for milkymist seems to lack features plus bugs, and aparently noo usabillity13:51
lekernelthen, how would you integrate support for graphics acceleration like the texturing unit?13:51
lekernelwhen I see the mess the DRI is, I do not want to touch it13:51
wpwraklekernel: that's an optimization problem :) first, i'd be interested in the 99% of the applications where i don't really care about performance13:52
rafalekernel: x11 slow: it depends that you mean with slow.. it looks for me that x11 is a generic word for a huge world of programs doing xwindow.. And I have found xfbdev fast enough many times against porting applications to fb directly..13:52
mthlekernel: I thought Gallium was the future of graphics acceleration13:53
mthor is that layered on top of DRI?13:53
lekernelboth seem equally unappealling to me13:53
lekernelkeep thing simple...13:54
mthbut graphics hardware today is not simple13:54
lekernelmine is :)13:54
mthyou can make a specific API for your hardware of course, but then there won't be much software for it13:55
wpwraklekernel: milkymist can be different things to different people. e.g., i usually don't care much about graphics performance. but the ability to synthesize peripherals that connect efficiently to the rest of the system would be very exciting.13:55
rafawpwrak: if lekernel needs a GUI without X then I would vote for SDL :) .. well, it is not a GUI to do windows/buttons.. but you can easily to use it for that and write a small lib13:55
mthif you want to do windowing, consider DirectFB13:55
mthyou can even run wxWidgets or GTK on top of DirectFB13:55
lekernelI already have a GUI library that does windowing, buttons, widgets, etc.13:56
mthand DirectFB can use hardware specific acceleration if you write some driver code for it13:56
wpwrakrafa: yeah, and people have done amazing things on top of SDL. i always thought it was little more than just a frame buffer, but boy was i wrong.13:56
lekerneland is many times faster than a wxwidgets+gtk+cairo+glib+directfb+sdl+fb bloatstack13:56
wpwraklekernel: ah yes, wxGTK ;-)13:57
lekernelbut well, if someone wants to run Linux, perfect!13:58
lekernelI'm all for it13:58
mthlekernel: wxWidgets + DirectFB is a pretty light stack13:58
lekernelI just do not want to waste my personal time on it13:58
lekernelit's a huge effort, bigger than reinventing some wheels13:59
mthyou can always get something lighter if you write everything yourself, but I'm not sure the difference is worth the effort13:59
lekerneland a dirty job when it comes to fix the software from the GNUtards13:59
kristianpaulsure not you lekernel send a board to rafa ;-)13:59
lekernelyou don't need a board for the linux port, there's QEMU available for that14:00
wpwraklekernel: there's a big difference between qemu and a real device when it comes to motivation ...14:00
rafalekernel: how do you know that your lib is faster than sdl for example?.. I mean.. maybe I am not following the whole chat here.. but it seems that you just have your GUI and no linux there?.. How is the comparison?14:01
lekernelif there's isn't enough motivation already to overcome using QEMU in a first time, I doubt there will be enough when it'll come to fight with the GNU tools14:01
kristianpaulactually more for the visual tasks i agree with wpwrak14:01
kristianpaulrafa: i think there is no comparison yet14:01
lekernelkristianpaul, oh, I did get SDL apps to run on Linux14:02
lekerneldid you forget those mails already?14:02
lekernelmy initial plan was even to use uClinux and SDL14:02
qi-bot[commit] Werner Almesberger: Print a frequency estimate after each burst. http://qi-hw.com/p/ben-wpan/9361f1414:03
qi-bot[commit] Werner Almesberger: cntr/cntr.c: option -v (report data corruption) was never implemented, oops. http://qi-hw.com/p/ben-wpan/a8d743414:03
qi-bot[commit] Werner Almesberger: More detailed examination of the input circuit problem. http://qi-hw.com/p/ben-wpan/6d4ea6114:03
lekernelthen I gave up after countless problems with the GNU toolchain, porting software and linux kernel driver writing14:03
lekernellinux was slow already, too14:03
lekernelslow to boot, slow to run applications (the crappy executable loader doesn't help, though)14:04
kristianpaulhmm http://lists.milkymist.org/htdig.cgi/devel-milkymist.org/2010-January/000339.html14:04
lekernelah, yeah, got DOOM to work, too14:04
rafakristianpaul: is uclinuc the thing used with openwrt?14:04
rafa(as well)14:04
wpwrakrafa: could you mean uclibc ?14:05
rafawpwrak: probably. .. dont know much about openwrt, so it was ulibc surely :)14:06
wpwrakrafa: uclinux is a variant of the linux kernel that doesn't need an MMU.14:06
wpwraklekernel: perhaps it would help things if you could add some really simple MMU ? could be a throwaway design, just to get things rolling. that way, the system would be "more mainstream", so you'd avoid a number of pitfalls that certainly exist outside the beaten path.14:08
rafawpwrak: ah.. yes, then uclibc is the thing I remember from qi ml :)14:08
wpwraklekernel: (of course, after ccc ... :)14:09
lekernelyeah, I could do that and what not, if I had 72-hour days14:09
lekernelthough I won't reject a patch that implements a MMU14:09
lekernelmy view on the question is that GNU/Linux is a huge and dirty effort that is better shared and put on as many shoulders as possible14:11
wpwraklekernel: i'd love to help with an mmu patch, but i still have that open source synthesis toolchain in my dependency graph ... seems that none of your future selves has invented time travel, or you'd already have the code somewhere :)14:11
lekernelI won't touch it unless I have an army of 100 motivated developers14:12
wpwraklekernel: (many shoulders) it's a huge effort, that's for sure. so anything that makes it easier for people to get going is good.14:12
lekerneland if I were to port an Unix-like OS from scratch myself, I'd rather choose NetBSD14:13
lekernelassuming I have a MMU14:13
kristianpaulrafa: yes libc suck for emdebbed a bit ;-) i was told that14:13
lekernelI'm just too tired of the shitty technical quality of GNU/Linux14:13
lekernelopen source synthesis toolchain: that's something I find way more interesting than fixing GNUtarded software :)14:14
rafakristianpaul: can you tell me who told you that? you often say "i was told that".. and many time I do not agree :D14:14
rafakristianpaul: BTW, you mean uclibc, or libc?14:15
rafakristianpaul: do you remember the wpwrak tests of speed? jlime seemed a lot faster than openwrt.. and it uses libc (jlime). But well, I am not saying here that it is the reason why jlime was faster14:15
kristianpaulrafa: was a guying working on the early port for the avr32 cpu in openwrt14:15
wpwraklekernel: excellent. there are lots of people out there who can make autotools behave, but there are very few who can do the open synthesis toolchain :)14:15
rafakristianpaul: but that is something which both distros could differ14:16
wpwrakrafa: my speed test was flawed. (i think because of gmenu2x messing with the clock, but i'm not entirely sure.) in the end, they're about the same.14:17
kristianpaulrafa: agreed, i think you are more experienced on that field to give a comment, this guy surelly was just complaining14:17
rafawpwrak: somebody did the test without gmenu2x as well IIRC14:17
wpwrakrafa: btw, when you say "glibc", do you really mean "glibc" or "eglibc" ? according to christoph, glibc is a dead end14:18
rafawpwrak: (we did a few tests on #jlime, so I should check my logs)14:18
rafawpwrak: christoph always giving nice news ;-))14:19
wpwrakrafa: (memcpy speed) i don't quite remember the details. just that the difference largely disappeared in the end.14:19
wpwrakrafa: (glibc dead end) the reasons being precisely that it's too bloated and so on. apparently, most major distributions have switched to eglibc.14:21
rafawpwrak: eglibc is the thing14:21
rafawpwrak: the confussion for me is that the package is named libc6, but it is eglibc14:22
wpwrakrafa: yeah, for compatibility :)14:22
kristianpaulHey i like this conding style in SIE sram example :Ā°)15:37
kristianpaulis simple15:40
kristianpauland fun15:42
wpwrakhmm, cas_latency_dmcr[((CFG_SDRAM_CASL == 3) ? 1 : 0)]15:45
wpwraksomeone distrusts his C compiler a lot more than even lekernel distrusts gcc ;-)15:45
kristianpaulwell not so simple15:49
kristianpauli dint noticed that line yet...15:49
kristianpaulwhat ever it does?....15:50
kristianpaulwpwrak: i'm at src folder ATM15:50
kristianpaulhe better i dint looked and build :p15:50
kristianpaulha ! thats ingenic code15:51
wpwrakah, right :) maybe wolfgang can bring them a C book as a present for the next visit :) the code looks nice in general, but there are a few spots where it looks deeply confused15:52
lekernelhahaha, glibc15:56
lekernelanother reason for me to dislike GNU15:56
lekerneljust look at the bug reports and patches and how they're answered... LOL15:56
wpwraklekernel: glibc is dying. long live eglibc ! :)15:57
kristianpaulwill this happen to gcc? ...15:58
kristianpaullekernel: had you ever consider minix3 for the MM?15:59
Action: kristianpaul hides15:59
kristianpaulhandy http://en.qi-hardware.com/wiki/SAKC_GPIOs/es16:20
wpwrakhmm, if this is for gcc, better to use inline instead of #define16:21
kristianpaulah sure16:23
kristianpaulbut gets an idea and could save me read manual pages16:23
wpwrak(inline) because the p and o can cause stray warnings with -Wshadow. otherwise, this looks good. of course,  foo = (1 << o);  ought to be  foo = 1 << o;16:26
wpwrakactually, 1 << (o) :-)16:26
wpwrakat least while it's a macro16:26
lekernelminix3 needs a MMU16:40
kristofferlekernel, how is the feeling of minix3 btw? been awhile since I tested it17:12
lekernelit seems pretty good17:13
kyakNAND DATA partition offset is incorrect now18:37
kristianpaulhow come?18:38
kyakseems to me that the offset should be increased by 25618:40
kristianpaulbecause gcc and other stuff need a 512MB rootfs18:40
kyakyou don't get me18:41
kristianpaulnope :(18:41
kyak.offset = 264 * 0x100000,18:41
kyaki think reflash_ben.sh should be updated, too?18:42
kristianpaulalso wiki18:42
kristianpaulsome nerase command need fix too18:42
kristianpauloh what you mean with updated?18:43
kristianpaulthis changes are not offical yet i thnk so18:43
kyakwhat do you mean "not official"?18:43
kyakit is in git, so as reflash_ben.sh18:43
kyakif one thing changes, other things should change, too18:44
kyakbtw, gcc doesn't seem to compile on x64 and marked as broken now18:45
kyakso maybe 512 MB is not needed after all :)18:45
kristianpaulbut reflash_ben.sh dint pull Ben image from last git commits18:46
kristianpaulindeed, maybe18:46
kyakreflash_ben.sh on the web site is ok18:46
kyakthe one in git should be changed18:46
kristianpaulmake sense18:47
kristianpauli was confused18:47
kristianpaulwow planet.openmoko.org looks like planet.qi.com.. : )18:55
wpwrakkristianpaul: let's fill one planet before we add another one ;-)19:58
qi-bot[commit] Werner Almesberger: Add DFU to BOOKSHELF. Add MMC driver unloading instructions to f32x/README http://qi-hw.com/p/f32xbase/19e87df21:03
qi-bot[commit] Werner Almesberger: Added footprint for U.FL (micro coax) receptacle. http://qi-hw.com/p/ben-wpan/e6c881821:24
--- Sat Nov 6 201000:00

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