#qi-hardware IRC log for Wednesday, 2012-04-04

pabs3DocScrutinizer: glamo docs are fully public now, Sean got permission to release them01:06
pabs3(well, unless there are some he didn't have or forgot about)01:10
kristianpaulyes it is01:17
kristianpaulin my image..01:17
kristianpaulOkay, confirmed i'll be there http://softwarelibre.info01:22
kristianpauli was told have one hour... oh well, i'll carry my M1, nanonote and see what i can demo or talk, also give away some stickers etc..01:23
kristianpaulyeah.. well, i would like to spare time for more activities but :)01:25
wolfspraulwho is the audience?01:26
kristianpaulstudents, mature people (who also pay tickets? btw..)01:27
kristianpauli'm not very aware of FLOSS audience in the Ecuador so, all is new..01:27
kristianpaulah, i guess people from oter floss based enterprises as well..01:28
Action: pabs3 wonders if kristianpaul will take some opencores stuff too01:43
kristianpauli take milkymist cores too ;-)01:44
kristianpaulbut i dont want get too tech at leas is necesary, remenber is floss event..01:44
pabs3I meant http://opencores.org/01:44
wolfspraulpabs3: have you developed for fpgas? which opencores cores have you found usable?01:45
wolfspraulI would love to understand better what is there and how to (re)use and improve it, but except for "hey there is opencores", typically there is nothing :-)01:45
kristianpaulindeed, i  havent time to look at opencores, perhpas the ethernet and usb ones.. but ... nah to lazy to compare01:46
pabs3given that Linux supports OpenRISC now, I guess something from there is working01:46
kristianpaulit works indeed01:46
kristianpauland openrisc SoC is a fact for that01:46
wolfspraulhave either of you tried it?01:47
rohhey wolfspraul 01:47
Action: kristianpaul waiting stekern release its openrisc port to M1..01:47
rohgot my mail?01:47
wolfspraulkristianpaul: yep :-)01:48
wolfspraulthat would be great01:48
wolfspraulpabs3: do you have a url for "linux supports openrisc"?01:48
wolfspraulwho is using it? have they documented their work?01:48
Action: pabs3 goes looking01:48
wolfspraulof course I know these statements for months, years. but a lot seems to be just hearsay.01:49
pabs3wolfspraul: http://kernelnewbies.org/Linux_3.1#head-37c60fa1253db74ce7d224718a71f5836bd5be0901:49
wolfspraulone-off experiments with lots of unresolved issues, barely enough for some bragging, no code releases, of course nobody else building on top of it because there is nothing to build upon :-)01:49
wolfspraulpabs3: let's find users and documentation01:49
wolfspraulthat would help us and/or stekern01:50
wolfspraulif it's open and it works, anybody should be able to use it01:50
wolfspraulroh: yes, got it - thanks!01:50
rohwolfspraul: hope you can meet. nice guy (mitch)01:50
kristianpaulpabs3: fyi stekern is a guy at #milkymist that hack around openrisc and its SoC01:50
wolfspraulfact is this:01:51
wolfspraulmilkymist soc runs today, and afaik nothing stops us from replacing the lm32 core with an openrisc core01:51
wolfspraulexcept for 'a little' work :-)01:52
kristianpaullittle, ha ;)01:52
wolfspraulbut for some reason Sebastien says the lm32 core is superior, and there must be a reason for that01:52
wolfsprauland we should be careful to dismiss that simply reiterating statements from lots of people where most haven't even tried to use what they claim exists01:53
wolfspraulor: fire up your code editor :-)01:53
wolfspraulI also hear that opencores has a 'great' 'linux supported' usb controller01:53
kristianpaullm32 has its advantages yes, plus still small and not bad documented i think01:53
wolfspraulI try to follow opencores with the best intention for years, read their newsletter, browse projects, etc.01:54
wolfspraulbut somehow it doesn't translate into reality01:55
wolfspraulbut let's see01:55
wolfspraulstekern is our hope01:55
wolfsprauland maybe pabs3 buys a m1 soon and starts opencores hacking01:55
wolfspraulit's one thing that a developer sayss "this works great", or 100 or 1000 users actually agree with him :-)01:56
wolfspraulthat's an entirely different thing in my experience01:56
wolfspraullike night and day - *that* different01:56
wolfspraulpabs3: does this make sense? let's start to *use* opencores stuff!01:57
wolfspraulnot just point to it, that's no news for years01:57
wolfspraulwhen we use it, we find out what really works01:57
pabs3personally I don't have the time nor disposable income to start working on that though01:57
wolfspraulwe find out the nasty little details that make all the difference01:57
wolfspraulsure sure01:58
wolfspraulI did not in any way want to dismiss your input01:58
wolfspraulmaybe one day you find the time and excitement01:58
pabs3understood :)01:58
wolfspraulopencores is great01:58
wolfspraulwe all love it01:58
kristianpaulmilkymist uses conbus core from it !01:58
wolfspraulbut so far I think not 1 line of any opencores core is running on m101:58
wolfspraulthere you go01:58
kristianpaulwich is the thing glue all soc comunication :)01:59
wolfspraulkristianpaul: do you have a url into the opencores site to the original conbus?01:59
wolfspraulI mean the one that m1 reuses/uses01:59
wolfspraulI want to build more of those connections, make people understand where things come from so they can find more...01:59
wolfspraulI repeatedly hear 'openrisc' and also 'usb controller'02:00
kristianpaulwell, the version is not the same, but that one in milkmist re-use most of the code i remenber02:00
wolfspraulkristianpaul: how much was the code changed when being brought over to m1?02:00
kristianpaul20-30% i bet02:00
kristianpaulhavent dont full comparison02:00
wolfspraulthat's the one thing in cores land that still confuses me02:01
wolfspraulhow can people work together, and benefit from each others work02:01
wolfspraulis the m1 'fork' (?) of conbus known to the original conbus devs?02:01
wolfsprauldo they care?02:01
kristianpauli think i wrote him..02:01
kristianpaulbut never got reply..02:01
wolfspraulare cores mostly a one-time heoric effort without much reuse or long-term value?02:02
kristianpaulthen i find answer my self trought milkymist version ;)02:02
wolfspraulnice that you tried to establish contact02:02
wolfspraulno reply, oh well :-)02:02
Action: pabs3 goes to idle on #opencores02:03
kristianpaulpabs3: also sebastien sent back to opencores some cores from milkymist, the memory controller and navre (usb soft core=02:03
kristianpaulpabs3: no no02:03
kristianpaultoo alone :)02:03
kristianpaulreally :)02:03
wolfspraulI think the typical 'open core' (not a pun on opencores) is buggy like hell02:03
wolfspraulthere are too few devs02:03
wolfspraulno quality standards02:03
wolfspraulno standardized test environments02:03
wolfspraulno expectation of reuse even, which makes devs even more reluctant to invest their time in reusability02:04
wolfspraultoo many subtle details in different fpgas, even fpga generations, asic02:05
wolfspraultiming optimizations that eventually lead to a magic binary bitstream that nobody will touch or dare to ask how on earth it was made02:05
wolfspraulbuggy vendor toolchains working against any efforts for better long-term quality and reusability02:06
wolfspraulmore? :-)02:06
wolfspraulkristianpaul: would you agree?02:06
wolfspraulyou are out in the minefield for a while now... you probably know what I mean, or can tell me if I'm wrong02:06
wolfspraultesting in general is under-appreciated in open cores02:07
kristianpaulyeah i know.. :-|02:07
wolfspraulof course because a lot of the value of high-quality and repeatable testing only shows years later after someone reuses the core, which may be never, so nobody invests in that...02:07
wolfspraulactually the problem is not infinite, not at all02:07
kristianpaulalso i followed opencores for while, until found you guys :D02:07
wolfspraulbut there are *a lot* of shaky foundations right now, quicksand02:08
kristianpaulcores with a SoC. well...02:08
wolfspraulfrom a quality perspective, I think opencores is just a random assortment of all sorts of code snippets02:08
wolfspraulI think that's fair to say, althouhg I also don't know how to do better02:08
kristianpaullets wait stekern  news hopefully will give us better info about all this02:08
wolfspraulmilkymist soc doesn't have the problem because there is 1 dev who tightly controls the entire soc tree - of course he can at least from his perspective keep the quality and testing up02:09
wolfspraulkristianpaul: definitely :-)02:09
wolfspraulbut most likely stekern will run into lots of nasty details02:09
wolfspraulon the surface it looks easy, but then...02:09
wolfspraulwill he work through all of them? why?02:09
wolfspraulmaybe just a few quick hacks enough for a demo or to say 'it works', then move to greener lands?02:10
kristianpaulto benchmark, that 'boring' need to be done anyway no?02:10
wolfspraulit's a chicken-egg problem02:10
kristianpaulhehe it works hacks :)02:10
wolfspraulno culture of reuse02:10
wolfspraulno investment in resuability02:10
kristianpaulfully agree, on that one02:11
wolfspraulso let's start to fix the culture first, of course we try to not just talk about 'open cores', but actually use them and fix all the nasty bugs02:11
wolfsprauland invest in high-quality documentation and testing02:11
wolfsprauluntil hopefully one day the real reuse and feed back cycle starts02:11
wolfspraulthat's how I see it02:11
kristianpaulhmm may be we need first list SoC currently using wishbone  and wishbone-like cores, plus compare what from opencores too02:11
wolfspraulwould be good to do that on wikipedia though02:12
wolfspraulI'm editing milkymist-related information in wikipedia sometimes, as I learn. but could do more.02:12
wolfspraulthere was also a port of milkymist soc to altera02:13
wolfspraulvery exciting stuff imho02:13
wolfspraulbut william (fpgaminer) also has a limited time budget and it's too hard to unite it back and say "milkymist soc is supported and tested on xilinx s-6 and xilinx s-3 and altera xxx"02:13
wolfspraulone code tree02:13
wolfspraulone set of documentation02:13
kristianpauloh, i need to dollow the editing of that article more often02:13
wolfspraulwould be cool, but lots of work and again: why do it? if nobody reuses anyway :-)02:14
wolfspraullemme see, at least the url02:16
wolfspraultrying to build bridges :-)02:16
wolfspraulpainful to build a bridge when nobody cares to use it anyway02:16
wolfspraulyou rather just use a boat to cross the river once, for yourself02:17
wolfspraulmuch easier, right?02:17
wolfspraulwhich Altera chip was this, one sec?02:17
kristianpaulcyclone III i bet02:18
wolfspraulgreat, can't even find in the README :-)02:18
wolfsprauland (*zero* blame to william here), there is not much effort to merge activities, or at least document/learn from each other02:19
wolfspraulin fact he already did me a great favor by publishing his sources, trying to arrange them a little like the milkymist soc tree, etc.02:19
wolfspraulthat is already pioneering work!02:19
wolfspraulmost people take an open core, tweak, polish, tweak more, hack, tweak, until it works. and then everything they learnt stays unstructured and locally and is not fed back.02:20
wolfspraulI think I must have seen this 10 times now with bits and pieces from milkymist, over the last 2 years.02:20
wolfspraulpainful :-)02:20
wolfspraulso much wasted knowledge02:20
wolfspraulkristianpaul: you think cyclone III?02:20
wolfspraulseems he mentioned DE2_11502:21
kristianpaulah, thats cyclone II i remenber02:22
kristianpaulno no02:22
kristianpaulargh, anywyay02:23
wolfspraulwhat I find looks like Cyclone IV EP4CE11502:24
kristianpauloh much better02:24
kristianpaulhe also i wodering now if he run on to routing problems..02:24
kristianpaulneed to as02:24
wolfspraulwow that chip alone costs about 400 USD02:25
wolfspraulwe have such great value on m1!02:25
wolfspraulsuper high performance at low chip cost, and full focus on the open cores02:25
wolfspraulthen we have noone but ourselves to blame :-)02:26
wolfsprauland the devkit sells for 300 USD :-)02:26
wolfspraulthe de2-115 one02:26
wolfspraulI need to register/login with opencores to even view the sources?02:28
wolfspraulthose kinds of things make me wonder about the platform, better move the sources to github even02:28
qi-bot[commit] kyak: libcss: update to 0.1.2 (master) http://qi-hw.com/p/openwrt-packages/b53ab4205:03
qi-bot[commit] kyak: libhubbub: update to 0.1.2 (master) http://qi-hw.com/p/openwrt-packages/18bd88f05:03
qi-bot[commit] kyak: libnsfb: update to 0.0.2 (master) http://qi-hw.com/p/openwrt-packages/a347d7405:03
qi-bot[commit] kyak: libparserutils: update to 0.1.1 (master) http://qi-hw.com/p/openwrt-packages/e0300a005:03
qi-bot[commit] kyak: libwapcaplet: update to 0.1.1 (master) http://qi-hw.com/p/openwrt-packages/e1e92f205:03
qi-bot[commit] kyak: libnsbmp: initial port (master) http://qi-hw.com/p/openwrt-packages/aa543e605:03
qi-bot[commit] kyak: libnsgif: initial port (master) http://qi-hw.com/p/openwrt-packages/fae184405:03
qi-bot[commit] kyak: netsurf: update to 2.9 (master) http://qi-hw.com/p/openwrt-packages/a32871205:03
qi-bot[commit] kyak: Merge branch 'master' of projects.qi-hardware.com:openwrt-packages (master) http://qi-hw.com/p/openwrt-packages/b5ffbdf05:03
kyakhm.. i forgot how to avoid the "Merge branch 'master'" commits...05:06
kyaki though git pull before git push was enough05:06
pabs3git pull is essentially git fetch + git merge05:11
pabs3I think you want git fetch + git rebase, or git pull --rebase05:11
wolfspraulkyak: do you know the state of vnc on the Ben?05:16
wolfspraulsometimes I keep wondering about it - I vaguely remember once there was a problem because some client required an X backend05:16
wolfspraulbut I've used clients on framebuf before, so not sure. In the packages repo, I find vnc-reflector, vncrepeater and libvncserver05:17
wolfspraulany known vnc clients for Ben?05:17
kyakpabs3: ah, indeed!05:20
kyakwolfspraul: i have no idea.. :)05:21
kyaknever tried that05:21
wolfspraulok sure, I shall investigate05:21
kyaki only see vnc proxy and vnc repeater packages in openwrt05:22
kyakand some vnc server library05:22
kyak"i need new eye"05:23
kyakyou said that :)05:23
kyakif we find sdl vnc client, we might have a chance05:24
kyakbut then, it's just 320x240.. what is the use case you are thinking about?05:25
kyakhttp://www.ferzkopp.net/Software/SDL_vnc/ - this is something really basic05:27
kyakand old, too05:27
wolfsprauluse case anticipates future higher resolutions :-)05:30
wolfspraulI've used a vnc client on framebuf before, but forgot which one it was. there are so many clients...05:30
kyakhttp://svn.icculus.org/palantir/trunk/ - this is another sdl vnc client05:30
wolfspraulah, directvnc05:31
wolfspraulthat was the one I used05:31
kyakah looking good05:31
wolfspraulyes, sdl might also be an option. I'm looking for the shortest path to something that already exists... but no worries, I'll look around05:31
kyaknot very old also05:31
kyakdirectfb is probably better than sdl if it works05:32
wolfspraulDocScrutinizer: hi good morning :-)06:02
wolfspraulyou mentioned boundary scan tests the other day, in passing with design rule checks06:03
wolfspraulI think for design rules, we are already on a good path, at least knowledge-wise, the rest is a matter of implementing and documenting06:03
wolfspraulbut how about jtag boundary scans? can you describe in a bit more detail what kind of testing you had in mind?06:03
qi-bot[commit] Adam Wang: correcting NC pins to Unspecified electrical type. (master) http://qi-hw.com/p/kicad-libs/b864df807:16
qi-bot[commit] Adam Wang: correct Vcc to Power input electrical type (master) http://qi-hw.com/p/kicad-libs/55ba6c907:16
qi-bot[commit] Adam Wang: added DIN_5_2S with two pin shields (master) http://qi-hw.com/p/kicad-libs/e43dcbe07:16
DocScrutinizerwolfspraul: see http://en.wikipedia.org/wiki/Boundary_scan08:51
DocScrutinizerJTAG originally was meant for boundary_scan, it's just over time that people only know about >>When used during manufacturing, such systems also support non-test but affiliated applications such as in-system programming of various types of flash memory: NOR, NAND, and serial (I2C or SPI).08:53
DocScrutinizerand 08:54
DocScrutinizer>>The boundary scan architecture also provides functionality which helps developers and engineers during development stages of an embedded system. A JTAG Test Access Port (TAP) can be turned into a low-speed logic analyzer.08:54
DocScrutinizer~wiki jtag08:55
infobotAt http://en.wikipedia.org/wiki/Jtag (URL), Wikipedia explains: "{{Refimprove|date=November 2009}} 'Joint Test Action Group' ('JTAG') is the common name for what was later standardized as the IEEE 1149.1 'Standard Test Access Port and Boundary-Scan Architecture'. It was initially devised for testing printed circuit boards using boundary scan and is still widely used for this application. Today JTAG is also widely used for IC debug ports. In the ...08:55
DocScrutinizer"Joint Test" is the keyword08:55
DocScrutinizerwolfspraul: to put it as simple as it basically is: JTAG boundary scan is a huge "cable tester"09:00
DocScrutinizerwith your JTAG-BS aware chip at one end of cable, and either another such chip or some test equipment outside your PCB at the other end of "cable"09:01
DocScrutinizerthen you check each "wire" (aka trace / solder point) for connection and for isolation to neighbours09:02
DocScrutinizerthis is obviously simple to implement for GPIO, without JTAG-BS. Not though for other buses and lines, like addr bus from SoC to RAM or whatever. Also not for lines that have one dedicated other function in normal operation, like on/off-switch or whatever09:04
DocScrutinizerbasically you can "remove" the function core of a bs-aware chip from your circuit, and "replace" it with a logic tester, so you can read in _and_ _set_ logic level of _each_ pin of the chip - except VDD, GND, 4 JTAG pins09:07
DocScrutinizerJTAG quite usually gets daisychained, so all your chips form one long chain of JTAG blocks that's controlled over just one JTAG connector to your PCB09:09
DocScrutinizer>>The ability to perform such testing on finished boards is an essential part of Design For Test in today's products, increasing the number of faults that can be found before products ship to customers.<<09:32
DocScrutinizerand especially for you ;-) >>... JTAG scan chain enables a **low overhead**, embedded solution to testing an IC for certain static faults (shorts, opens, and logic errors).<<09:34
DocScrutinizerwolfspraul: btw DSC like I define it, is an automatic check of properties of one pin against the properties of other pins on same trace. Like "no 2 outputs on same trace", "high level defined voltage (range) same (resp matching) for output and all inputs" etc09:51
wolfspraulunderstood - have to do some background reading but I'll get to that. thanks a lot!09:51
DocScrutinizerof course you regularly need to augment the defualt checks, if you use nifty design which might allow 2 tristate outputs on same trace, when driven correctly09:52
DocScrutinizerbasically for each trace/net you have a DRC that consists of property definitions of all the pins, plus a transformation algo that's empty for plain wire nets but something rather complex for e.g. bus with 50R termination09:54
DocScrutinizerthen some spice-alike equation solver runs each net to check if the equation results in "good" or in "mismatch"09:56
DocScrutinizeropen pins are a very interesting case for that09:57
DocScrutinizer(nifty design with tristate) usually you get a 3rd class of equation solver results: warnings10:00
DocScrutinizer"WARNING! two outputs on same trace. Mismatch for (p238:out:1 && p317:out:0), (p238:out:0 && p317:out:1)10:03
DocScrutinizerfor GPIO exactly same warning10:05
DocScrutinizerfor a case where one of both not tristate but totempole output, this becomes "ERROR!"10:05
DocScrutinizersame "ERROR!" if for 2 GPIO the power-on defualt of one is incompatible with power-on default of the other10:06
DocScrutinizerall those infos (tristate output, H-voltage range, L-voltage range, power-on default...) need to get stored with the pins of component in CAD10:08
DocScrutinizerthe equation solver engine runs one solution for each of the pin's possible states: H, L, high-Z, input, pullup, pulldown...10:10
DocScrutinizerlike with lint you'll want to "comment out" some warnings, by e.g. defining pull-up and output as illegal for a GPIO10:12
DocScrutinizerall those "lint comments" will get shipped to sw-engineers to let them know what they must avoid to ever do10:13
DocScrutinizera good hw design has an empty such list10:14
DocScrutinizerfor e.g. the simplest case of 2 GPIO connected you do this by connecting them via a 50R, so there's no exceeding load to either of both when one is output:1 and other is output:010:17
DocScrutinizerunless of course your properties of both pins would allow such pathological case without any violation of ABS MAX ratings for fan out10:19
DocScrutinizer(means: one of the GPIO has some "internal 50R" so it's short circuit tolerant and also won't overload the other GPIO)10:20
qi-bot[commit] Xiangfu: nanonote-files: cleanup etc/ files (master) http://qi-hw.com/p/openwrt-packages/78095b110:28
qi-bot[commit] Xiangfu: xburst: nanonote: move Ben special files to it's package (master) http://qi-hw.com/p/openwrt-xburst/c6e40d310:36
DocScrutinizerwolfspraul: probably quite a different approach to DRC than the trace-geometrics check we usually see in layout CAD, hm? :-)10:37
qi-bot[commit] Xiangfu: nanonote-files: config.full_system: remove non-compile packages (master) http://qi-hw.com/p/openwrt-packages/4ff543710:39
qi-bot[commit] Adam Wang: added 6N138, 8-Pin SMD Single-Channel Low Input Current High Gain Split Darlington Output Optocoupler (master) http://qi-hw.com/p/kicad-libs/de234eb10:42
kyakxiangfu: some packages you removed are actually fixed, like centerim, dgclock..10:47
kyaknightsky, netsurf, what else..10:48
kyakit probably makes sense not to remove broken packages, but rather make them =m or mark as @BROKEN in Makefile, so people woudl know10:51
qi-bot[commit] Xiangfu: Revert "nanonote-files: config.full_system: remove non-compile packages" (master) http://qi-hw.com/p/openwrt-packages/e3b513610:55
qi-bot[commit] Xiangfu: nanonote-files: remove nlove, jamvm, pygame. add libnl-tiny (master) http://qi-hw.com/p/openwrt-packages/83f342910:55
xiangfukyak, oh. 10:55
kyaki'd really prefer marking them @BROKEN, so that we know what to fix10:56
xiangfukyak, mark as @BROKEN.  yes agree.10:56
xiangfukyak, but I think just remove from config.full_system and mark as @BROKEN10:58
xiangfu=m is like comment :-)10:58
kyakyeah, if you mark it as @BROKEN, =m doesn't make sense10:59
xiangfukyak, thanks. marking @BROKEN now..11:00
xiangfukyak, I saw you update 'netsurf' a lot, cool11:01
qi-bot[commit] Xiangfu: mark nlove, offrss, pygame as @BROKEN (master) http://qi-hw.com/p/openwrt-packages/5a2adbf11:02
kyakyeah, but if you read the comment to netsurf commit, it's not quite so cool11:02
xiangfukyak, I am try to add new 'gmen2x' icons to nanonote-files. since gmenu2x have broken. :(11:02
DocScrutinizerwolfspraul: with DRC like I define it, odd quirks like the "LED eating massive current" in GTA02A5 never had crept in11:03
xiangfukyak, yes I saw the  comment and 'gmenu2x' revert. :-)11:03
kyakxiangfu: actually, it makes more sense to keep icons in nanonote-files repo rather than gmenu2x repo11:04
DocScrutinizercatching the 1uF (instead 100uF) in hs-audio would have needed a rather complex DRC rules set, but in principle even that would have been detectable11:04
kyakgmenu2x repo is for gmenu2x development, not for new icons commits :)11:04
xiangfukyak, yes. 11:05
xiangfukyak, let's do it. :-) 11:05
DocScrutinizer...depending on powers of equation solver11:05
xiangfuI am moving the icons now..11:05
kyakand if things go well with David's fancy new shell, who knows, maybe gmenu2x may as well have a rest11:06
qi-bot[commit] Xiangfu: move nanonote gmen2x icons from gmen2x.git to here (master) http://qi-hw.com/p/openwrt-packages/cb9045211:08
qi-bot[commit] Xiangfu: move all Ben Nanonote icons to it's nanonote-files packages (master) http://qi-hw.com/p/gmenu2x/a881e7811:09
xiangfukyak, let me stop the current build see if we can get a clean build for next release. (include all recently commits)11:12
kyakyou need to mark gcc-mips as broken11:12
kyaki wasn't able to get to it yet..11:13
kyakthen probably it would all build11:13
qi-bot[commit] Xiangfu: gcc-mips mark as broken (master) http://qi-hw.com/p/openwrt-packages/7eb4f7811:14
xiangfukyak, ok. the current build haved stop. let's see how next build going...11:15
kyakxiangfu: great!11:16
xiangfuthe build will start in next 3 hours. those 3 hours buildhost is busy on milkymist one compile .11:16
kyakwolfspraul: this is directvnc running on Ben trying to connect to vnc server on my laptop running xterm :) http://downloads.qi-hardware.com/people/kyak/tmp/directvnc.png11:16
kyakit looks pretty bad.. has something to do with directfb11:17
kyakbad it somewhat works.. i typed uname -a :)11:17
kyakxiangfu: btw, nmap and tcpdump fail to build for me.. didn't have a closer look yet11:18
kyakwe have tcpdump in full config, might cause problems..11:18
xiangfutcpdump compile fine under build host.11:19
xiangfukyak, http://fidelio.qi-hardware.com/~xiangfu/build-nanonote/openwrt-xburst.minimal-20120402-0952/packages/tcpdump_4.2.1-1_xburst.ipk11:20
kyakoh, ok11:20
xiangfuI use a config.minimal build to check if package compile find.11:20
qi-bot[commit] Xiangfu: nanonote-files: don't use manually edit config file (master) http://qi-hw.com/p/openwrt-packages/2b11c5111:21
xiangfukyak, BTW: the make kernel_menuconfig, in different system give different behavior.11:21
xiangfukyak, I think you commit is works better under buildhost. which is good.11:22
kyakxiangfu: yeah, i remember that kernel_menuconfig was buggy, but i tried this time, and it worked just fine11:22
kyakbtw, what;s the difference between config.autogen and config?11:23
xiangfukyak, I manually edit some config in latest release. since there are too many package not compile in latest release.11:24
xiangfukyak, now back to normal.11:25
kyakah, ok11:25
xiangfukyak, have to go. see you later.11:26
kyaksee you!11:26
zoltanh7211Hi guys11:45
wpwrakDocScrutinizer: (JTAG) interesting ... i always considered "Joint" an attribute of "Group", not of "Test". similar to "junta". in the sense of multi-vendor.11:56
zoltanh7211rejon ping11:56
DocScrutinizerwpwrak: unclear11:57
DocScrutinizerfor me it's "joint test"11:57
DocScrutinizeras that's basically what JTAG-BS is all about11:57
DocScrutinizerboth INT (for bonding) and EXT (for solder joints)11:58
DocScrutinizernah, not exactly< bonding11:58
DocScrutinizerbonding also is EXT11:58
DocScrutinizerobviously :-)11:58
zoltanh7211wolfspraul ping11:58
wpwrakDocScrutinizer: (DRC) we have a primitive test of that kind in eeschema (called ERC, electrical ...). you set pin types and it checks for compatibility. here's a view of schematics symbols with pin types http://people.openmoko.org/werner/gta02-core/gta02-core-expanded-all.pdf11:59
DocScrutinizerwpwrak: nice11:59
wpwrakDocScrutinizer: or the (work in progress) qi-hw variant: http://downloads.qi-hardware.com/people/werner/tmp/out.pdf11:59
wpwrakkyak: (gcc-mips broken) so that's "game over" ? :)12:04
DocScrutinizerwpwrak: if your pin types are sufficiently detailed, in a sense that BiDi(S4C4554) := GPIO_type3(1.8V,pu+pd,fan-out=8)12:05
kyakwpwrak: no, not at all.. just everything has its time.. too many packages broke since last release12:05
wpwrakkyak: (for what it's worth, i ran yesterday into the problem of the owrt build process trying to download unavailable gcc-4.6-linaro)12:05
DocScrutinizerthen this is even better than declaring each and every parameter for each and every pin again12:05
wpwrakkyak: ah, so it builds but some exotic features don't work ?12:06
wpwrakkyak: (gcc download) that was following http://en.qi-hardware.com/wiki/Building_Software_Image12:06
kyakwpwrak: no, it doesn't build :) not with the updated toolchain of openwrt12:06
DocScrutinizera generic BiDi for all levels of VDD_IO of course doesn't help that much12:06
wpwrakDocScrutinizer: yes, i said it's primitive ;-)12:07
kyaknew gcc version, new uClibc version, gcc-mips is broken :)12:07
DocScrutinizerwpwrak: maybe primitive but we're on same page regarding principles12:07
kyakwpwrak: i regularly remove my dl/ dir.. didn't have the problem.. probably the mirror was down.. is it still the case?12:07
kyakwpwrak: are you followign the "Building OpenWrt based on release files" or "Building OpenWrt on last git commmit" steps?12:08
wpwrakkyak: one mirror was down, the others didn't have the directory. that's where i declared defeat :)12:08
wpwrakkyak: based on release files. playing it safe :) and i used the one that explicitly gives a commit (43a86619c3cb9aceaace51097bc35d59b7b8a4fc)12:09
DocScrutinizerroh: ping12:10
wpwrakkyak: imho, the approach of downloading things from all around the world is too fragile. it would be better to have a local (qi-hw) mirror that doesn't automatically propagate deletions12:10
DocScrutinizerroh: what are the plans regarding svn.openmoko.org now?12:10
DocScrutinizerroh: seems quite a number of people still need it12:11
kyakwpwrak: there is such thing.. once downloaded on buildhost, the tarball should be in http://downloads.qi-hardware.com/software/mirror-openwrt-sources/12:11
kyakand the build system will try to pull it from there as a last resort12:12
qi-bot[commit] Adam Wang: added 1). BZX84 Voltage Regulator 2). SN75HVD12D RS-485 TRANSCEIVERS 3). JS28F256J3F105 FLASH (master) http://qi-hw.com/p/kicad-libs/a54fc4a12:12
kyaknow, i wonder why linaro-4.6... if you base on a last release, it should be linaro-4.512:12
kyakwpwrak: how about you do it based on a last commit? :)12:13
wpwrakkyak: (mirror) sounds good. but perhaps it should be used first ? after all, that's what has been tested.12:14
wpwrakkyak: (linaro) ftp://ftp.uu.net/archive/systems/gnu/gcc/gcc-4.6-linaro/gcc-4.6-linaro.tar.bz212:14
wpwrakkyak: (lates commit) i think i'll give it a try again when the dust has settled :) for now, i'm using an old toolchain i found in a backup ... and build static binaries :)12:15
kyakwpwrak: i think the intention is to try to download from the "official" location, then fallback to the cached mirror12:15
kyakbtw, the commit number in wiki is just an example12:16
wpwrakkyak: (official) yes, but does that provide any practical benefit ?12:16
kyakyou should use the real release_ branch or find the correct hash12:16
wpwrakkyak: (example) hmm. why are there so many manual steps anyway ? wouldn't a "git pull ....; make release" be possible as well ?12:18
wpwrakkyak: or maybe "make toolchain". which should be one of the main reasons for wanting to go through all this anyway :)12:18
kyakthe page is about building images, i guess12:19
kyakif you want the toolchain - just download it12:19
kyakwhat do you mean by "manual steps"?12:20
qi-bot[commit] Adam Wang: added 1). FSMRA2JH switch 2). IR 3). Oscillator 4). XLR 3 pole female/male receptacle (master) http://qi-hw.com/p/kicad-libs/e3ebf0a12:20
wpwrak(official source) what i usually see is: 1) upstreams sometimes reorganize, changing paths. 2) upstreams sometimes delete old things (which we may still reference). 3) upstreams sometimes find horrible problems (package corruption or malware). they tend to resolve this in two ways: 3a) upload a fixed version. 3b) more common, delete the bad version and make a new version. 4.13 -> 4.13a or such. 4) upstreams sometimes migrate to a diff12:21
wpwrakerent site. 5) upstreams sometimes simply die.12:21
wpwrakand 6) upstream is temporarily down12:22
kyakfor everything you have mentioned, there is a fallback mirror12:22
kyakand also md5sum12:23
wpwrakin cases 1, 2, 3b, 4, and 5, the local mirror wins. in case 6, it usually wins too, because we need it to fetch other items as well, so it'll have to be up anyway12:23
wpwrakonly in case 3a there would be an advantage of preferring upstream over mirror12:23
kyakwpwrak: could you better tell me, how to make JZ4740 FB work in 16 bpp mode? :)12:24
kyakto avoid this horrible mess: http://downloads.qi-hardware.com/people/kyak/tmp/directvnc.png12:25
wpwrak(fallback) yes, i don't know why it didn't work. the build first failed silently. then i ran it again with V=99 but i stopped it after trying "upstream" mirrors for something like half an hour. so it never got to the local mirror (i didn't expect one at that point anyway)12:25
kyakdirectvnc doesn't support 32 bpp, JZ4740 FB doesn't support 16 bpp12:25
wpwraklooks pretty :)12:25
wpwrakah no, i don't know offhand how to change the color depth. trial and error are your friends ;-)12:26
kyaktrial and error and, probably, larsc :)12:27
wpwrak(mirror) or perhaps there could be some option to change the order ? ideally, it would default to picking the local mirror first :)12:27
rohkyak: there are many colorspaces in 16 and 32bit as well12:27
kyaklarsc: do you have an idea how to make JZ4740 FB work in 16 bit mode?12:28
kyakwpwrak: of course there is an "option" :) in scripts/download.pl12:28
kyakroh: what do you mean?12:28
rohbyte-order, color depth12:29
roh32bit can be 24bit padded, with alpha...   different orders12:29
roh16bit can be 565 or sth. else, different orders...12:29
rohthere are many ways to use 16 or 32bit for encoding colors12:30
wpwrakkyak: nice ! thanks !12:30
rohhttp://projects.qi-hardware.com/index.php/p/openwrt-xburst/source/tree/master/target errors out for me12:32
roherror 50012:32
kyakroh: i'm sorry, i don't quiet get you.. is there any way to make it work in JZ4740 FB driver?12:32
kyakyeah, 500 for me, too.. strange12:33
rohdunno. just noted that there are not only a number of bits to know, but also which order and format the colors are in there12:33
wolfspraulbah the indefero server :-)12:37
wolfspraultoo bad zoltan left already...12:38
larsckyak: on the nanonote?12:48
kyaklarsc: yep..12:53
kyaklarsc: i'm trying to launch directvnc, which doesn't support 32 bpp (so i laucnh it in 16bpp). So i see something like this http://downloads.qi-hardware.com/people/kyak/tmp/directvnc-xeyes.png (the image is compressed by the half of screen)12:55
kyakor probably it is directvnc that needs to be fixed.. since i see exactly the same picture on my laptop - just a half of the screen13:12
larsckyak: yes that sounds like a better option.13:25
larscthe framebuffer needs it in 32bit13:26
wolfspraulkyak: I have no idea how this patch ends up in my people folder, but if you want to see a really crude version for 32bpp supports in directvnc, have a look at http://downloads.qi-hardware.com/people/wolfgang/tmp/directvnc-0_7_5.patch13:30
kyaklarsc: i see13:36
kyakwolfspraul: heh, nice, giving it a try13:37
wolfspraulit's just a few lines, and did the trick on a notebook I setup a little while back13:37
kyaksomething has definitely changed13:44
kyakit is full screen now, but still mangled13:44
zoltanh7211wolfspraul ping15:08
zoltanh7211rejon ping15:22
kyakmth: there is another regression in gmenu2x (also somewhere around these commits).. the "workdir" parameter in icon file is ignored16:26
kyakah, i read "Removed ability to configure custom working directory."16:27
kyakwell yeah.. it turns out to be useful16:27
kyaklike when we launch ash from gmenu2x, we want it to change to home dir16:28
kyakit's ugly when it starts in /usr/bin16:28
wpwrakkyak: use a wrapper script ?16:40
kyakwpwrak: frankly speaking, i don't want to.. the icon file itself is already a "wrapper"16:42
wpwraki woud seem to be16:45
wpwraklet's try this again16:45
wpwrakit would seem to be the unix way - don't create omnipotent applications but rather solve the problem with modular pieces. a little #!/bin/ash  cd $HOME  exec /bin/ash "$@"  doesn't seem overly troublesome :)16:47
wpwrakbut of course, i can't make you like it :)16:48
wpwrakoh, and why i tried the image build to get a toolchain: that's where google leads you when looking for "nanonote" and "toolchain". not sure if there's a more streamlined process anywhere.16:49
mthkyak: is it just ash that has this problem or more programs?16:54
mthcurrently gmenu2x is rather difficult to maintain and use, mostly because it has dozens of features16:55
mthwell, the way the code is structured is the actual cause for the maintenance problems, but the more features there are, the harder it is to restructure16:56
kyakmth, wpwrak: i understood and agree with your points.. indeed, the cleaner gmenu2x's code, the better for us.17:47
kyakwpwrak: if you just need the toolchain, i can get it here: http://downloads.qi-hardware.com/software/images/NanoNote/Ben/2011-11-13/17:49
kyakbtw, you are also checking out a pretty old release (2011-05-28).. maybe that's why you have troubles downloading the sources17:50
whitequarkDocScrutinizer: what do you think about the article?18:15
DocScrutinizerindeed lol, regarding it does full pearl regex but implicitly assumes /g18:50
DocScrutinizerumm, this article... coreboot, ChromeOS... errr :-S18:55
whitequarkwhat's with it?19:19
whitequarkI thought coreboot was a good project19:19
DocScrutinizerwell, probably it is19:45
DocScrutinizerthough I'm not sure if it's just another BIOS, or that nonsensical bios based mp3 player19:45
whitequarkerr, a FOSS BIOS19:45
whitequarkno idea what mp3 player you're talking about19:46
DocScrutinizerthere's a BIOS that does media and stuff without any OS booting19:56
DocScrutinizersure a FOSS BIOS is a nice thing19:57
GNUtooit's a very nice thing, I did the port on my desktop mainboard21:24
wpwrakkyak: (toolchain) thanks ! for now i'm good. i'll probably give it another try in a week or so.21:33
dvdkhmm, mplayer segfaults in uclibc23:18
dvdkhmm, have to disable freetype support in mplayer, to make it run with 'fbdev' output.  accelerated output still crashing, though23:45
dvdkcorrection, mplayer crashes somewhere in /lib/ld-uClibc-0.9.33.so, wtf?23:48
wolfspraulhow good is our debugging on the Ben setup actually?23:56
dvdkwell, not at all? :)23:57
wolfspraulshould we auto-build images or packages with debug info turned on?23:57
dvdkwolfspraul: debug info is not kept on nanonote, but on build system.23:57
wolfsprauland how would one be able to capture that then in case of a crash?23:57
wolfspraulyeah, I know it's tedious, so wondering23:58
wolfspraulevery time there is a crash it's like the world stops :-)23:58
wolfspraulnormally you would just want to know the source code line and maybe a few local variables and that would allow you to go one step further in 95% of cases23:58
dvdkwell, currently i'm debugging with gdb but without debug info.  assembler dumps plus checking memory mapping 23:58
dvdkcurrently the easiest way is to recompile a buggy package with debug info, then test that.23:59
--- Thu Apr 5 201200:00

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