#qi-hardware IRC log for Wednesday, 2011-05-11

wpwrakyeah. i added a slightly more positive comment00:09
mthif I compile gmenu2x with GCC 4.3.3, the first libstdc++ function I call causes a segfault, with libstdc++ from GCC 4.5.100:48
mthis that expected or did I do something wrong either compiling gmenu2x or compiling GCC?00:49
mthI thought GCC 4.x was ABI compatible00:49
mthand the version number of libstdc++ is 6.0.10 vs 6.0.14, so that should be binary compatible00:49
wolfspraulhe :-)01:37
wolfspraulthey took it?01:37
kristianpaulyeah, i wonder why actually ;-)01:37
wolfspraulwhy not :-)01:38
wolfspraulI guess the story was good enough, compared to the other things available at the time01:38
wolfspraulcome on this is simple stuff. an editor E constantly has the job to fill up his queue, let's say 10 stories / day.01:38
kristianpaulseems beta-like stories can pass trough01:38
wolfspraulso every day he comes to work, starts... picks the first one, second one, third one, etc.01:38
wolfspraulthat's not how it works01:38
wolfspraulanything can pass01:38
wolfspraulthe editor is under time pressure01:39
wolfspraulif by lunch he is supposed to have finished 5 stories, but he only has three, then he has to speed up in the afternoon :-)01:39
kristianpaulalso under lekernel submissions presure ;-)01:39
wolfspraulwell I'm happy for lekernel, if they take it that's great01:39
wolfspraulwe all should submit more milkymist stories all over the internet :-)01:39
kristianpauloh, sure !01:39
wolfspraulMilkymist has potential for dozens of good stories, they just need to be written.01:40
kristianpaulI'm happy too01:40
wolfsprauland then someone needs to carry them, so we have to shop those stories around, not bury them silently on some unread blog01:40
wolfspraulthat's all. there is no other magic.01:40
wolfspraulwrite it well, good language. authentic, not invented bs. slowly you will build up a reputation with editors and your stories will make it through the competition easier and easier.01:41
wolfspraulkristianpaul: there is a Spanish slashdot variant, forgot the name01:41
wolfspraulyou can try there, in your native language so maybe you can write at a higher level than English, and chances are better that it's taken...01:42
kristianpaulbarra punto, yeah01:42
kristianpaul\. ^01:42
wolfspraulfire away :-)01:42
wolfspraulfollow the footsteps of fantastic Spanish language writers...01:43
kristianpaulhehe i need to take a look to that first :-)01:43
kristianpaulbut yeah, should be easier than english at least01:43
wolfsprauloh you should definitely do that01:43
wolfspraulthe story has a much higher chance to be taken if the language is good01:44
wolfspraulunfortunately most non-English articles will not be picked up elsewhere, but from English to any other language works easily01:44
wolfspraulso you are confined to the Spanish readers. but that's a lot, and you can definitely have an impact.01:44
Action: kristianpaul confined =|01:44
wolfspraulwell it's true. in technology everybody can at least read English.01:45
wolfspraulso the news flow is English ... any other language01:45
wolfspraulbut if some real original news pops up elsewhere, it will not move to any other language01:45
wolfspraulnobody translates Japanese news stories into English01:45
wolfsprauleven though there are a huge number of interesting things going on in Japan01:46
wolfspraulin technology01:46
wolfspraulbut even there it's very rare/never to be translated01:46
qwebirc87931saw the article today on slashdot and looked at the code in milkymist.  There are some possible sim bugs in a few of the files in tmu2.02:54
qwebirc87931it looks like a handful of processes were converted into synchronous processes, but left with blocking assigns to signals used outside of the process, and that are not clocks.02:55
kristianpaulqwebirc87931: hi02:56
wolfspraulyou are very welcome here, but the core milkymist channel on freenode is #milkymist02:57
wolfspraulbut keep going02:57
qwebirc87931ah, i just found this from the wiki/webpage02:58
wolfspraulyou are absolutely in the correct place02:58
wolfspraulqi hardware is a copyleft hardware project, our focus is hardware and manufacturing02:58
wolfspraulwe are manufacturing the Milkymist One video synthesizer02:58
wolfspraulthere is quite some overlap in people between #qi-hardware and #milkymist02:58
wolfspraulmilkymist is the free IC design project (the milkymist SoC), plus the Milkymist One video synthesizer (the first product made around the Milkymist SoC), and the Flickernoise video synthesis GUI02:59
qwebirc87931ah, ok, this really is more on the HDL side, not sure where you draw the hardware boundaries.02:59
kristianpaul(overlap) and lekernel is here to catch you comment as he developed the tmu2 core :-)02:59
wolfspraulSebastien, the founder of Milkymist, hangs out on #qi-hardware as well03:00
wolfspraulI just wanted to clarify the two channels to you first, you an feel at home in both.03:00
wolfspraulcan you give us more information on the tmu bugs you found? can you send a patch to the mailing list? I know very little/nothing about Verilog/HDL...03:01
wolfspraulbut if you found a bug - that's great!03:01
qwebirc87931Ah, it shouldn't be hard to find.  Basically the simulator might not correctly simulate some files.  The synthesizer works differently and will give consistant results03:04
qwebirc87931which, might not match the sims as the sims have the bug.03:04
wolfspraulqwebirc87931: what do you think about the Milkymist project overall? any ideas for the future you may have?03:05
qwebirc87931Well, I'm mainly an FPGA dev, I saw the project on slashdot.03:06
qwebirc87931I'm always a bit skeptical.  There wasn't a lot of details on what specifically makes this better than a CPU, GPU, lowpower CPU+GPU, or lowpower CPU+FPGA.03:07
wolfspraulskeptical is good03:13
wolfspraul'better' in which sense, what are you comparing?03:13
qwebirc87931better in any sense.  an FPGA only is difficult to program and develop for.  a CPU only might not have enough performance.  GPU's can do quite a bit.  in terms of power, they do make low power CPUs and low power GPUs, and obviously low power FPGAs03:15
wolfspraulhey btw, if you have some spare time, there's a number of things that could be added in the fpga03:15
wolfspraulfrom memory - some people started to think about MMU design (it has no mmu right now)03:15
wolfspraulthere is the idea to support s-video and tv-out over the vga (with a converter cable, but needs SoC support first)03:16
wolfspraulreal-time video encoding and streaming over Ethernet would be cool, maybe VP8...03:16
qwebirc87931Well, i really mean in terms of the core processing.  thing like optimized architectures for things like VP8 and such03:17
wolfspraulok but that's just theory03:17
wolfspraulwhat is your point?03:17
wolfspraulsay we make a car, it's a really cool car03:17
wolfsprauland you say "but if the car would be 10 times faster it would be better"03:18
wolfspraulok - yes. you are right :-)03:18
wolfspraulmilkymist one is a product first of all, a video synthesizer03:18
wolfspraulit works, it's very very interesting03:18
wolfspraulnow we invite more people to join, if they want to, if they see something in it that interests them03:18
qwebirc87931I always question why someone would place a CPU on an FPGA.  There certainly would be reasons to do so.03:19
wolfspraulfrom my perspective: saves 1 chip, allows for integration between core instruction processing and peripherals03:20
wolfspraulif you have cpu+fpga, you definitely have a more complicated system in terms of tools than if you just have an fpga03:20
wolfsprauloh btw, since you are an fpga guy, there's also the llhdl/antares project you might find interesting03:20
kristianpaulqwebirc87931: xilinx is already tought on that for spartan703:21
qwebirc87931do you?  In my experiences, having a dedicated CPU and FPGA has always worked better.  mainly because you have good tools for the CPU, and FPGA tools for the FPGA, and can debug each seperatly.03:21
qwebirc87931but having high BW, low latency busses for a CPU does make a lot of sense03:22
kristianpaulqwebirc87931: but oddly, or no? altera is ecoraging the use of mips softcores.. so.. there is a way to go o this topic03:22
kristianpaulof course we stick on free/open stuff, but i wanted to make some comparison03:22
wolfspraulgee can't find the llhdl sources right now :-) guess they are being moved03:22
wolfspraulhere's a starting point http://lists.milkymist.org/listinfo.cgi/llhdl-milkymist.org03:22
kristianpaulwolfspraul: yeah moved time ago..03:23
wolfspraulqwebirc87931: again. I understand your theory.03:24
wpwrakwe're probably just at the beginning of seeing serious cpu core development. intuitively, i would also expect a cpu-asic plus an i/o fpga to offer the best performance. but maybe one can shrink the difference to a point where it doesn't matter so much anymore.03:24
qwebirc87931As for the altera/xilinx current FPGAs, one of my other issues is that they don't have the partial reconfiguration support to the point of being useable, and the FPGA has to be programmed even to use the hard-ip CPUs03:24
wolfspraulI think in terms of usability, it depends on each case. I have never seen anyone arguing for adding a non-fpga cpu core to Milkymist One03:25
wpwrakmuch like you're usually faster in assembler than in C, but only few people are desperate/crazy enough to do much programming in assembler.03:25
wolfspraulit would make the design very difficult, and create lots of growth problems in the future, imho (my theory ;-))03:25
kristianpaulwhat aboy  having a 128bit cpu running at 60Mhz? :p03:25
wpwrakqwebirc87931: where's the problem with reconfig ? that the chips can't do it properly or that the tools mess it up ?03:26
wolfsprauland today, the video synthesizer works flawlessly and we have many ideas how to improve it, and none of those ideas include adding another core CPU03:26
kristianpaulor specilized softcores for ie improve lzma descompresion time..03:26
wolfspraulqwebirc87931: yes we love partial reconfiguration, why not. It's not officially supported in the spartan-6 and probably never will be, and the practical usefulness is open for debate too (in our case, video synthesizer).03:27
wpwrakkristianpaul: yes, with an fpga you can "grow" little coprocessors. e.g., how about a cache that can do a bit of math of its own ? people have poked in this direction already, but usually with asics, so this was very slow.03:27
wpwrakkristianpaul: once you take the fpga as a given, you probably find it much easier to move around. also, once the shackles of the proprietary tools are removed ;-)03:28
wolfspraulqwebirc87931: one thing is important I think. we are not an fpga shop. we didn't see fpga and had some theory on how this technology is superior (or not).03:29
wolfspraulthus we are not building an fpga computer03:29
wolfspraulthe thought process of most/all people here goes like this:03:29
wolfspraul1) we love free technology (gpl/bsd), we think we can build really cool products out of free technology03:30
wolfspraul2) we picked a video synthesizer as a starting point because we thought this is a cool first product, and it can later grow into other products03:30
wpwrakkristianpaul: all this reminds me a bit of the situation we had in operating system kernel development before 386bsd and linux. not much happened because it was all centered on proprietary closed code bases. only once that was overcome, all of a sudden lots of people dived into it and started to innovate.03:30
wolfspraul3) we picked a Xilinx fpga as a chip, similar to us picking an Analog Devices video decoder, or Wolfson Micro audio codec03:31
wolfspraulwe are staying away from using proprietary Xilinx fpga features (the design should easily be portable to an Altera, for example)03:31
wpwrakwolfspraul: well, some of us don't care much about 2) but find the thing cool regardless ;-)03:31
kristianpaulyeah :-)03:32
wolfspraulwpwrak: yes I just try to explain the thought process, because some people look at this and say "ah, fpga computer. done that, failed. theory is wrong, bla bla bla"03:32
wolfspraulbut nobody here thinks like that03:32
wolfspraulfpga is just a chip, yes it's great because we can run more free technology in it that we could run on an ARM chip03:32
wolfspraulthe ARM chip is missing partial reconfiguration :-)03:32
wolfspraulit's just a difference in what thought/idea is first, and what is a second step/consequence03:33
wolfspraulthe first idea is free tech here, the second idea is fpga. not the other way round.03:33
wolfspraulsorry gotta run, bbl03:33
wpwrakwolfspraul: well, the concerns about fpga being unsuitable matter of course. afaik, little as it is, mm1 is much more efficient than those previous attempts. in terms of gate count (fpga size), processing power, performance of peripherals.03:34
kristianpaulkeep it simple... :-)03:34
kristianpaulwich btw qwebirc87931 mm1 is not that soo complex system some people tend to think it is03:35
wolfspraulwpwrak: yes, but nobody here looked at fpga vs. asic first, and then said "I pick an fpga over an asic because..."03:35
wolfspraulthe thought process of the Milkymist project is different03:35
wpwrakkristianpaul: exactly. someone commented that they had somehow half a system and could have done the rest if they had just had a virtex. well, that big expensive virtex is just the thing we'd rather not need ;-)03:35
kristianpaulyay !03:36
qwebirc87931back.  In anycase I guess part of the project is itself to make an open CPU.03:37
wpwrakwolfspraul: i'd say there's an fpga way and an asic way. sebastien chose the fpga way and this agrees in many ways with what the rest of us like, too.03:38
kristianpaulqwebirc87931: sure thats our basis03:38
wpwrakqwebirc87931: from what i gather, there's a lot of value in having a properly finished soc implementation. finished in the sense that there's a full set of peripherals and they are properly integrated and debugged.03:39
qwebirc87931It's just that, if you are making an open source CPU for audio/video, I would have expected to see more audio-video structures.  though some might be there03:39
kristianpaulIs not just the CPU, as wpwrak pointed, the system (SoC) is the whole thing working torwards a new way of computing,03:40
kristianpaulie for now video manipualtion, this cool milkydrop effects, etc..03:40
wpwrakqwebirc87931: have you seen the thesis from this this originates ? there, sebastien talks a bit about optimizations for video, among other things: http://www.milkymist.org/thesis/thesis.pdf03:41
kristianpauljust matter of find a good fit between kernel, apps, libs and hardware that can do lot of stuff not just been only a CPU :-)03:41
kristianpaulokay time to check some rtem custom commands (WIP)03:41
wpwrakkristianpaul: (kernel) yeah, and get linux properly on the beast ;-)03:42
tkliveanyone i can ask about cpu-z06:13
dvdkcompile error07:36
dvdkcan't open file at /home/pub/spock/src/qi/openwrt-xburst/scripts/kconfig.pl line 32.07:37
wpwrakdvdk: you want to build from plain kernel sources, trust me on this ;-)07:39
kyakdvdk: http://en.qi-hardware.com/wiki/Ben_NanoNote/Kernel - decent instructions how to build kernel from qi-kernel repo. http://en.qi-hardware.com/wiki/Git - some info about your adventures with branches "master" and "trunk" :)07:40
dvdkmorning kyak07:44
dvdkdoes that mean i cannot build from openwrt-xburs.git 'trunk'?07:44
dvdki mean you've been committing there... without testing?07:44
kyakdvdk: hi :)07:45
kyaksure you can build from trunk07:45
kyaki just thought you specifically wanted the bare kernel07:46
dvdktried to rebuild.  saw teh error message i posted?07:46
dvdkno, want to tweak kernel config for next release07:46
kyakthat's a strange message07:46
dvdkbut kernel doesn't compile07:46
kyaki didn't experience such07:46
dvdkmaybe make clean doesn't suffice07:46
dvdkok, maybe my perl version is incompatible?07:47
dvdkrunning ubuntu 11.0407:47
kyaki doubt that.. i prefer to really clean it out before build07:47
dvdkok, it means it can't open the kernel config.  huh?07:48
dvdkit's inside uclibc07:48
kyakno, the kernel config is openwrt-xburst/build_dir/linux-xburst_qi_lb60/linux-
dvdkdo i need to update feeds.conf!?07:49
dvdkno, uclibc uses kconfig for uclibc .conf, too07:49
dvdk"  /home/pub/spock/src/qi/openwrt-xburst/scripts/kconfig.pl -n   ./config- > /home/pub/spock/src/qi/openwrt-xburst/build_dir/toolchain-mipsel_gcc-4.3.3+cs_uClibc-
kyakconfig- wtf?07:50
dvdkit's uclibc07:50
kyakyou are sure you are on trunk?07:50
Action: wpwrak giggles07:50
kyakit's uClibc-0.9.3207:50
dvdkgit checkout origin/trunk07:50
dvdkyou're right it shows 0.9.3207:51
dvdks toolchain/uClibc/07:51
dvdkconfig-0.9.32  Config.in  Config.version  Makefile  patches-0.9.3207:51
dvdkmaybe 'make distclean'?07:51
kyakdvdk: maybe you could start by really cleaning up the mess left from backfire :)07:51
kyakyes, distclean07:51
dvdkrunning, will take a few minutes07:52
kyakcheck that you don't have env variables left like ARCH or CROSS_COMPILE (they could be left if you tried compiling the kernel manually)07:53
xiangfudvdk: (make distclean) better backup the dl/ folder :)07:57
dvdkxiangfu: too late :(07:57
dvdkbut running a squid proxy anyways07:57
wpwrakit's really not a good idea to lead people in the direction of using a distro build process as sort of a super-make for development on individual applications08:00
wpwrakwe had that mess at openmoko for some application development and it was a total nightmare. people must have spent half of their waking time just chasing build system excursions or just waiting for vast quantities of completely irrelevant stuff to be built over and over again08:01
wpwrakbetter to have a proper cross-development toolchain and environment, with the ability to install -dev packages directly into the cross-development environment, and build exactly what you need08:02
kyakwpwrak: there is an SDK08:06
kyakxiangfu: btw, how was your buildhost progress for master/trunk?08:06
xiangfukyak: trunk: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.trunk-full_system-05112011-0430/08:07
xiangfukyak: master is :  0 19 */2 * * /home/xiangfu/bin/compile-openwrt-xburst.sh full_system08:07
xiangfucrontab ^08:07
kyakxiangfu:  error: conflicting types for 'versionsort'08:07
kyakseems you haven't pulled the openwrt-packages/trunk08:08
wpwrakkyak: so ... why isn't dvdk using it ? :)08:08
kyakthe versionsort patch was relevant for backfire, but must be removed in trunk08:08
xiangfukyak: yes. add switch to openwrt-package/trunk to build script file now.08:09
kyakwpwrak: i don't know :) i don't use it because i like to watch it all build :)08:09
wpwrakkyak: tv so bad in your area ? ;-)08:10
kyakwpwrak: you are close, i rarely watch television :)08:12
kyakxiangfu: i see there are still those efl/xfce/phone feeds in feeds.conf. Maybe they should be disabled? (i only have qipackages/packages/desktop)08:17
kyakthese efl/xfce/phone contain outdated/unusable packages anyway08:17
xiangfukyak: yes. then let's just remove them. 08:18
kyakxiangfu: in nanonote-files/data/qi_lb60/conf/feeds.conf they are pinned to a specific svn revision.. Should we leave it?08:20
kyaki wonder how you do it in your build script..08:21
xiangfukyak: those 'svn revision' is for 'release', it's always keep the latest release package svn revision.08:21
kyakahh, oh, i remember now08:21
xiangfukyak: I am not using the feeds.conf, only when release. we modify that revision. 08:21
kyakok, i see08:23
xiangfuthe 'make distclean' will not remove the 'feeds.conf' :)08:24
qi-bot[commit] Werner Almesberger: tools/lib/usbopen.h: moved to tools/include/, to allow for sharing http://qi-hw.com/p/ben-wpan/6fc212808:31
qi-bot[commit] Werner Almesberger: libatrf: new function usb_rescan to force next open_usb to scan tree again http://qi-hw.com/p/ben-wpan/3deac4108:31
qi-bot[commit] Werner Almesberger: tools/lib/usbopen.c: make vendor and/or product optional http://qi-hw.com/p/ben-wpan/55d220b08:31
qi-bot[commit] Werner Almesberger: tools/usbwait/: new tool to wait for a USB device to appear http://qi-hw.com/p/ben-wpan/703785808:31
qi-bot[commit] Werner Almesberger: tools/usbwait/usbwait.c: new option -r to require removal before appearance http://qi-hw.com/p/ben-wpan/62a03bb08:31
qi-bot[commit] Werner Almesberger: atusb/fw/Makefile: new target "update" to update the application via DFU http://qi-hw.com/p/ben-wpan/b82db5508:31
dvdkxiangfu: i'm currently try to get CONFIG_PROC_PAGE_MONITOR=y into kernel config08:47
dvdkis the 2.6.37 or .38 going into next release?08:47
dvdkneed the PAGE_MONITOR for cleaner DMA implementation in mplayer video08:47
xiangfudvdk: you can add it to 'openwrt-package.git' trunk branch, which is for the 2.6.37 or .38 kernel.08:48
dvdkxiangfu: so add it to both kernels?08:48
xiangfusorry. it's openwrt-xburst.git08:49
dvdkxiangfu: already working on a trunk checkout of openwrt-xburst.git08:49
dvdkit's using gcc 4.5 now, is it?08:49
xiangfuby default it's using 2.6.37, but you can push both I think :)08:50
xiangfuyes. 'trunk' using gcc4.508:50
dvdkso next release will be gcc 4.5, too?08:50
dvdkmaybe gives us some speedup for mplayer/theora.  that can never hurt.08:51
xiangfudvdk: I am not sure. I guess 80% still backfire. let's see how 'trunk' build on build-host. 08:51
dvdkxiangfu: ah, ok so you have to merge between branches for new built08:52
xiangfudvdk: here: http://en.qi-hardware.com/wiki/Git#openwrt-xburst.git_rebase_on_upstream_trunk08:54
wpwraktoday, also slashdot.jp picked up ubb-vga. nice to see the wave go around the globe ;-)08:54
xiangfumost of the patches is useless in 'trunk', also kyak have finished a lot of work, :)08:54
dvdkwpwrak: (slashdot.jp): "hack that uses micro sd card slot as vga port" no word aof abuse08:55
wpwrakthey're too polite ;-)08:56
dvdktrying to decipher the story.  pretty much rewritten.  08:57
wpwrakthe comments are quite a bit more disciplined than over at /.09:01
dvdkwpwrak: that's not unexpected, is it :)09:02
dvdkcool, i now have a 5 digit slashdot ID09:06
dvdk... slashdot.jp ID :)09:06
dvdkwpwrak: now i'm curious how much visitors we get through slashdot.jp.  enough for another slashdot effect?09:10
wpwrak95 until ~8:30 am UTC09:11
wpwrakabout 159 from /. for MM1. but that's qi-hw. milkymist.org as the primary landing site probably has a lot more09:12
wpwrakalso 86 visitors from downloads.qi-hardware.com ;-)09:12
dvdkwpwrak: can't seem to find anybody from /.jp on http://en.qi-hardware.com/piwik09:13
dvdkah, it says date=yesterday here09:14
dvdkwpwrak: it now even shows japan on the map09:17
wpwrakyeah. the world at a glance. just what we need for our domination plans :)09:18
qi-bot[commit] Werner Almesberger: tools/dirtpan/: quick and dirty IPv4 over 802.15.4 tunnel (in progress) http://qi-hw.com/p/ben-wpan/dbad7ae10:40
kyakdvdk: i was running your mplayer-jz in trunk, haven't noticed any difference to backfire :)11:03
kyakerror: pathspec '97dc86b793efb9c6ac604cdfff4027fe27efa12c6' did not match any file(s) known to git.11:06
kyakdamn.. not this again11:06
kyaki removed the dl/ dir, it became > 2 Gb :)11:06
kyakhm, ffmpeg devs somehow managed to screw up this revision, too11:09
kyakdespite of the fact that i got their whole git tree11:09
kyakthe only explanation now is that specific commit had been reverted11:11
kyakdvdk: the latest mplayer from svn won't compile -\ sticking to r33341 and ffmpeg is updated to 3b6bbfa0631d237f2bbc85a7b43907733bea1e8211:28
dvdkkyak: don't update mplayer just for fun.  it takes quite some time to test through all kinds of sample files to make sure still everything works13:29
dvdkhmm, cannot push13:33
dvdkerror: failed to push some refs to 'git@projects.qi-hardware.com:openwrt-xburst.git'13:33
dvdkgit pull:13:33
dvdkYou are not currently on a branch, so I cannot use any13:33
dvdk'branch.<branchname>.merge' in your configuration file.13:33
qi-bot[commit] David Kühling: linux kernel: add CONFIG_PROC_PAGE_MONITOR=y to allow for clean user-space DMA http://qi-hw.com/p/openwrt-xburst/e9141de13:50
dvdkok needed --track -b when doing git checkout (?)13:50
kristianpauldvdk: just git checkout branch_name i thikn14:02
dvdkkristianpaul: did that (i think), but then i couldn't push.14:03
dvdkwell --track -b did the job this time.  going to try a pure 'checkout' next time14:03
kristianpaulgit push origin branch_name ?14:04
kristianpaulwell, origin or whatever name for remote repository14:05
dvdkkristianpaul: ah, that might have done the job.  14:12
plaeshave there been plans for nanonote v2?15:30
wpwrakplaes: ideas, many :)15:31
wpwrakplaes: but there's no concrete work on making one. one little issue is that this would cost a bit more money than anyone around here has at hand.15:34
Action: plaes has been pondering about data acquisition platform supporting various sensors15:36
plaesI know about Vernier LabQuest, but these are all stupidly expensive :S15:37
wpwrakyou could probably make sensors that connect via UBB. for that money, you can make quite a few ;-)15:40
wpwrakof course, the ben isn't quite so rugged ...15:41
plaesUBB is basically a SPI bus + plain GPIO?15:44
wpwrakit's mainly plain GPIO. you have three typical configurations: 1) 6 GPIO + 3.3. V power on/off; 2) 5 GPIO + 3.3 V power on/off + clock output (frequency is configurable, up to at least 56 MHz); 3) MMC/SD/SDIO15:47
wpwrakin the case of ubb-vga, i switch between 1) and 3), but that's an unusual mode of operation.15:48
wpwrakin 1) and 2), you can bit-bang SPI but you don't hardware-acceleration for SPI, so it's not so efficient if you have to move lots of data15:50
wpwrakperhaps you can actually drive some SPI devices with hardware support, if using the tricks i used with ubb-vga. i realize with a bit of sadness that the signal arrangement of atben doesn't allow for any optimization along these lines15:58
wpwraknot sure if it would actually be faster, though. such a thing would be more useful for SPI devices that transfer continuously but at a low bit rate. that way, you could just fill the fifo or set up dma, then go away. atben works in relatively fast bursts.16:02
dvdkon the master branch I get now after clean checkout and recompile:16:06
dvdkconfigure: error: in `/home/pub/spock/src/qi/openwrt-xburst/build_dir/target-mipsel_uClibc-':16:06
dvdkconfigure: error: C compiler cannot create executables16:06
dvdkok, my fault: renaming the openwrt-xburst directory is not possible (without distclean), it seems16:20
kyakdvdk: i wouldn't touch mplayer if i didn't have to update ffmpeg git revision.. But i got your point :)16:27
qi-bot[commit] David Kühling: mplayer_jz47xx: upgrade to 0.1.5 version that supports /proc/<pid>/pagemap http://qi-hw.com/p/openwrt-packages/54ae31816:37
qi-bot[commit] David Kühling: mplayer_jz47xx: fix md5-sum for updated version http://qi-hw.com/p/openwrt-packages/46bda1e16:38
dvdkkyak: did you break mplayer?16:59
dvdkerror: pathspec '97dc86b793efb9c6ac604cdfff4027fe27efa12c6' did not match any file(s) known to git.16:59
kyakdvdk: me?? i was the one who reported this problem17:03
dvdkkyak: didn't you say you just upgraded ffmpeg?  i had to downgrade it again17:03
dvdkset to FFMPEG_REV:=95f163b33b4f13cfe650f53ee2c3746788122def17:04
kyaki did upgrade it, but i didn't commit17:04
kyakand before, i said:17:04
kyak15:06:16 < kyak> error: pathspec '97dc86b793efb9c6ac604cdfff4027fe27efa12c6'  did not match any file(s) known to git.17:04
dvdkkyak: sorry, wasn't here when you wrote that.  scrolled off.17:05
dvdkkyak: maybe you better commit?17:05
kyakdvdk: i can do this. but actually i'm now wondering how should we do17:07
kyakcause we have master and trunk in openwrt-packages17:07
kyakand they are starting to diverge17:07
dvdki never commit in packages' trunk17:07
kyaki never did, too, until yesterday17:08
Action: dvdk hates this branches. too many headaches17:08
qi-bot[commit] kyak: update ffmpeg to a working git revision http://qi-hw.com/p/openwrt-packages/73ca35717:10
kyaki commited it in master.. i guess we will tag this branch when we migrate to next openwrt release.. and then commits from trunk branch will go on top17:11
kyaksomething like this...17:11
kyakin other works, the trunk branch is to temporary keep patches required for openwrt trunk, but not working/tested in backfire17:12
kyaki think xiangfu said it yesterday, but i wasn't very attentive..17:13
dvdkkyak: now i get on 'git stash pop'17:16
dvdkmplayer/Makefile: needs merge17:16
dvdkwhat do i enter now?17:16
dvdkto take your version of the file?17:16
kyakdepending on what changes you have inside your version of Makefile17:17
dvdkonly a different ffmpeg rev from yours17:17
kyakso you can just take my version17:17
kyakgit co mplayer/Makefile17:17
dvdkgit: 'co' is not a git command. See 'git --help'.17:18
kyakthen you can git pull17:18
kyakoh, sorry.. i have some shortcuts :)17:18
kyakgit checkout mplayer/Makefile17:18
dvdkmaysbe just git stash drop17:18
kyakheh, maybe, i didn't know abotu that :)17:18
dvdkseems to have worked17:18
dvdkkyak: but lost some other change.  not important one.17:19
kyakah yes.. checkouting this individual file would've been better then..17:19
dvdkgit checkout mplayer/Makefile17:19
dvdkerror: path 'mplayer/Makefile' is unmerged17:19
dvdkgit rm first?17:20
kyakhm ok, now you have to manually resolve the conflict i think...17:20
dvdkwhat's the command. svn would be 'svn resolved'17:20
kyakthere should be lines like >>>> inside mplayer/Makefile17:20
kyakmarking the conflict17:20
dvdkgit reset --hard HEAD17:21
dvdkkyak: you mean it parses inside these files?  that would be cool.17:21
kyakoh yeah, that one would work.. also smashing your changes :)17:21
dvdkbut too late already did 'rm' which didn't help, then reset17:21
qi-bot[commit] David Kühling: mplayer_jz47xx: upgrade to 0.1.5 version that supports /proc/<pid>/pagemap http://qi-hw.com/p/openwrt-packages/36bcb9e17:39
qi-bot[commit] David Kühling: mplayer_jz47xx: fix md5-sum for updated version http://qi-hw.com/p/openwrt-packages/f342c3b17:39
qi-bot[commit] kyak: update ffmpeg to a working git revision http://qi-hw.com/p/openwrt-packages/71d83e917:39
kyakok.. learned how to pick commits from one branch to another :)17:40
wpwrakdvdk: i'm not sure if it parses the files (didn't dare to commit a file containing unresolved conflicts yet :), but it marks them inside the files. kinda Augmented Reality :)17:48
kyakgit is kinda smart on resolving conflicts btw17:51
kyaki'd say, in most cases it would resolve automatically17:51
wpwrakhmm, i'd say it's more or less at the smartness level of patch(1)17:55
wpwrak(if it doesn't actually use patch(1) :)17:55
dvdkbtw the mplayer_jz47xx now uses /proc/self/pagemap if present.  should make sure that the next version really has the kernel support CONFIG_PROC_PAGE_MONITOR=y18:29
kyakdvdk: you should include this change in openwrt-xburst/master, too. xiangfu told he was going to release another backfire-based image18:31
kyaki think it's only in trunk so far18:32
dvdkkyak: you mean change the 2.6.32 config?18:32
dvdkoh the headaches come back18:32
kyakgit is a beast, right? :)18:33
dvdkmy brain is too small for it18:33
kyakwhen i think that i "mastered" git enough for my daily needs, something always shows up.. And i have to google.. every time :)18:35
dvdkhmm, does it automatically rebuild the kernel, if i just touch the config-2.6.32?18:39
dvdknot sure whether it did anything.  18:39
dvdkstill compiling, though.  does it recompile the full userspace?18:41
dvdklet's just see whether *uImage.bin is new in the end18:41
qi-bot[commit] Werner Almesberger: tools/dirtpan/: -d now generates terse output; -d -d dumps full content http://qi-hw.com/p/ben-wpan/29e56c719:51
qi-bot[commit] Werner Almesberger: tools/dirtpan/: introduced terse debug messages also for timeouts http://qi-hw.com/p/ben-wpan/a1a508e19:51
qi-bot[commit] Werner Almesberger: tools/dirtpan/dirtpan.c (open_net): turn off WPAN_WANTACK http://qi-hw.com/p/ben-wpan/171500e19:51
qi-bot[commit] Werner Almesberger: tools/dirtpan/dirtpan.c (timeout): fencepost error when normalizing tv_usec http://qi-hw.com/p/ben-wpan/354789d19:51
qi-bot[commit] Werner Almesberger: tools/dirtpan/dirtpan.c (rx_pck): immediately ack all data packets http://qi-hw.com/p/ben-wpan/2c7d06f19:51
qi-bot[commit] Werner Almesberger: tools/dirtpan/dirtpan.c (rx_pck): break tie if both sides are in s_tx http://qi-hw.com/p/ben-wpan/e17881c19:51
qi-bot[commit] Werner Almesberger: tools/dirtpan/: if given an interface conf command, run it with tunX in $ITF http://qi-hw.com/p/ben-wpan/de72af319:51
--- Thu May 12 201100:00

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