#qi-hardware IRC log for Wednesday, 2013-01-23

wpwrakheh, fun experiment: build the openwrt toolchain. it'll complain that STAGING_DIR isn't set. since you don't care about this mode of operation, just set it to "1"01:06
wpwrakthen everything will go fine. for a very very long time. until you have a file called "1" in the current directory. then you get a highly mysterious "cc1: error: 1/usr/include: Not a directory"01:07
DocScrutinizer05wolfspraul: actually with antibiotics it's exactly the other way round: if you take too small dosage or too short, the bacteria have a chance to adapt to them and then they're useless01:47
wpwraki think he meant treatments. some people take antibiotics when they have a light cold ...01:50
DocScrutinizer05wpwrak: any clue who's maintaining openmoko.org servers now?01:52
wpwrakdunno. harald ?01:53
wpwrakor maybe they're now fully on auto-pilot :)01:53
rohnobody afaik01:53
wpwrakDocScrutinizer05: perhaps you should maintain them. you seem to be the one who cares most about 'em :)01:54
DocScrutinizer05https://monitor.openmoko.org/munin/bhavani/devi.openmoko.org/cpu.html   https://monitor.openmoko.org/munin/bhavani/bhavani.openmoko.org/cpu.html01:56
wpwraka few  dd if=/dev/zero bs=1 of=/dev/null  ? :)01:58
DocScrutinizer05wpwrak: I actually might ponder having a look wtf is wrong there, but I lost my root access01:58
DocScrutinizer05wiki.maemo.org is *incredibly* slow01:59
wpwrakzombies are often depicted as sluggish :)02:01
wpwrakactually, the correct word would be "portrayed". may may, my language skills are fading ...03:54
C-KeenI have noticed that the .config file on the web is a bit out of sync with the git repo's buildroot11:56
C-Keenplus I get a wget: write error when trying to fetch the m4 sources...11:57
C-Keenah that's due to my wget version for some reason11:58
paul_boddieHello LunaVorax! Anything new?13:13
LunaVoraxStill lost my Ben13:13
paul_boddieSorry to hear that!13:13
LunaVoraxMy girlfriend still use the Ben everyday13:13
LunaVoraxuse her* Ben13:13
LunaVoraxI've ordered a GCW-Zero13:14
paul_boddieWell, I suppose there's some indirect benefit from your loss. :-)13:14
LunaVoraxAnd I'm struggling with C++ classes at school13:14
LunaVorax(not "class" as in "C++" but as in "school")13:14
paul_boddieGCW-Zero looks like a refresh of the Dingoo or something.13:15
paul_boddieWell, I remember learning C++ and finding it nice compared to C and Pascal, but moving on via Java to Python I find the former languages (including Java) awkward and archaic.13:17
LunaVoraxpaul_boddie, it looks like a better Dingoo with all the features the Dingoo/Wiz/Caanoo couldn't get right.13:18
paul_boddieI have to say that wpwrak's frustration with STAGING_DIR is amusing given that I was compiling his code with the OpenWRT toolchain and it started complaining about it. At first I thought I'd actually have to do something about it, but it was just a warning.13:18
wpwrakpaul_boddie: hmm, where did you run into my use of STAGING_DIR ? should be strictly a local phenomenon (unless - by sheer coincidence - you happened to do the same on your side)13:20
paul_boddieLunaVorax: It looks nice, and even has USB Host and will therefore please Ron (especially with Wi-Fi as well) and various other people.13:20
LunaVoraxIndeed paul_boddie 13:20
LunaVoraxWhat have been Qi up too all theses months?13:21
paul_boddiewpwrak: You don't use it, but the mipsel-gnu-linux-gcc (or whatever it's called) seems to like STAGING_DIR for some reason.13:21
LunaVoraxLooks like it's dormant13:21
paul_boddieI'm actually edging very slowly towards getting a program to talk SPI with an Arduino shield, but I want to make sure I know what I'm doing first.13:21
paul_boddieLunaVorax: wpwrak is the man of most action on the Qi front, I think.13:22
wpwrakpaul_boddie: ah yes, for some mysterious reason, openwrt keeps that STAGING_DIR variable around. not sure if there's a "proper" way to transition the toolchain out of the openwrt world bubble into a normal environment. i also don't know what exactly a valid STAGING_DIR ought to point to :)13:24
paul_boddieThe staging directory, maybe. Actually, I think it complained about that, too. :-)13:25
paul_boddieI'm also going to try and get cross-building for Debian to work, but that's not relevant for this project as Debian is running from the microSD card and it obviously won't be plugged in at the same time as the UBB.13:26
wpwrakdon't you wish we had some more uSD slots ? :)13:27
paul_boddieAlso, the Emdebian toolchain may not be completely compatible with my OpenWRT one, and I don't want to introduce more sources of error trying to deploy binaries built for Debian on OpenWRT.13:27
paul_boddieI noticed that SD is supposed to support daisy-chaining.13:27
wpwraktoolchains generally don't mix well. e.g., because you also have different libcs13:28
wpwrakdaisy-chaining ? that's new to me. i've seen multiplexing, but not chaining.13:28
paul_boddieMaybe not daisy-chaining. I'll have to look at Wikipedia again.13:30
roh'supposed to support'... well.. thats... a interresting definition *g*13:37
rohso it means 'doesnt do it in reality'?13:37
roh*scnr* .. i havent seen daisy-chains there too... just hacky '2 sd on one controller' crap with extra 'select' lines13:38
paul_boddie"The SD protocol envisioned the ability to gang 30 cards together without separate chip select lines. The host device would broadcast commands to all cards and identify the card to respond to the command using its unique serial number. In practice, cards are rarely ganged together because open-collector operation has problems at high speeds and increases power consumption. Newer versions of the SD specification recommend separate lines to each c13:42
paul_boddieGiven the relation to SPI, this doesn't surprise me that much.13:42
larscthere are even SD card multiplexer chips13:44
wpwraklarsc: yes, that's the sort of thing i've seen13:45
wpwrakpaul_boddie: interesting ... an ID space wired-or approach. nice to see that used in real life. well, almost :)13:45
paul_boddieJust trying to keep you at the cutting edge. ;-)13:46
wpwrakbtw, speaking of fun things, UBB can also sample external signals at a high constant speed. alas, not continuously. now trying to bring the trigger delay down ...14:03
C-Keenin the wiki http://en.qi-hardware.com/wiki/Building_Software_Image there is a section talking about creating a symbolic link for buildroot but somehow the ln command is garbled, what's the correct step I need to do here?14:14
wpwrakhmm, odd. can't find my openwrt build notes :(14:32
C-Keenadding ones own stuff would be a lot easier if the stuff this wiki page talks about would be integrated in git already. The patches to upstream feeds for example or the feeds config14:36
wpwrakyeah. i don't know why there's such a long manual process14:38
wpwrakand to make things worse, it always gets out of sync, so you need to poke around to find the correct names and paths14:38
C-Keendo you know a better description on how to reproduce the last release images?14:42
wpwrakif i could find my darn build notes, then i would have one that's at least more up to date ...14:44
wpwrakbut according to grep, they have vanished :-(14:44
paul_boddieWhich ln command?14:48
C-Keenpaul_boddie:  $ ln -s feeds/qipackages/nanonote-files/data/qi_lb60/files14:48
paul_boddiewpwrak: Would Ben/UBB be usable as a logic analyser?14:48
wpwrakpaul_boddie: with certain limitations, yes14:49
paul_boddieC-Keen: Doesn't that just make a link called "files" in the current directory? I'm sure I actually did this.14:49
wpwrakthe main limitations: 1) you can't record continuously, so you'd have a break every ~8 ksamples14:49
C-Keenpaul_boddie: but what's the current directory in this context?14:50
paul_boddieThe top of the buildroot.14:50
paul_boddieI always found it a bit confusing, really.14:50
wpwrak2) you also have a trigger delay in the order of 1 us if you start capture after seeing the trigger14:50
C-Keenok and I guess I need to adjust it too, as with the release_2012-10-11 branch this results in a dangling symlink14:50
wpwrak3) you need to come to terms with the pull-up resistors built into the ben14:51
paul_boddieI'll admit to not really building very much since last year, and only then with 2011 snapshots, I think.14:51
C-Keenpaul_boddie: as in the 'data/qi_lb60' part of the path seems to be gone now14:51
paul_boddieMaybe the git repository history can help here?14:52
wpwrakpaul_boddie: incidently, ...14:55
qi-bot[commit] Werner Almesberger: ubb-la/hw/: proof-of-concept passive LA adapter board (master) http://qi-hw.com/p/ben-blinkenlights/d4c1f5614:55
qi-bot[commit] Werner Almesberger: ubb-la/: data capture via UBB and MSC+DMA (WIP) (master) http://qi-hw.com/p/ben-blinkenlights/2c5b36214:55
C-Keenno the feeds/qipackages consists of two commits14:56
paul_boddieC-Keen: I will admit to not really knowing how the buildroot is maintained.14:56
C-Keenpaul_boddie: I think there is a space missing in the wiki docs14:57
C-Keenit should read ln -s feeds/qipackages/nanonote-files/data/qi_lb60/ files14:57
C-KeenI think14:57
paul_boddiehttp://projects.qi-hardware.com/index.php/p/openwrt-packages/source/tree/master/nanonote-files/data/qi_lb60 is probably what you want.15:04
C-Keenpaul_boddie: yep15:05
Action: wpwrak feels stupid15:05
paul_boddieSure enough, that directory used to have a files subdirectory: http://projects.qi-hardware.com/index.php/p/openwrt-packages/source/tree/release_2011-11-13/nanonote-files/data/qi_lb6015:05
wpwraki kept the build instructions right in the top-level directory. the only places where i didn't search ...15:05
wpwrakthe "git branch -a" is for locating a suitable branch15:06
paul_boddieIt seems to be a collection of files that get copied to the filesystem on the device, all under /etc.15:06
wpwrakthe "do not comp" is for eliminating the restriction prohibiting building as root15:06
C-Keenwpwrak: something like this is more detailed: http://en.qi-hardware.com/wiki/Building_OpenWRT_on_Debian_615:07
C-Keenwpwrak: thanks15:08
wpwraknp :)15:10
C-Keenwpwrak: do you remember why you download the latest .config and then overwrite it with the minimal one from feeds?15:10
wpwrakheh, good point :) i was just following the instructions. didn't notice that dubious step15:11
paul_boddieAs I recall, the build process does some fancy stuff with the .config anyway.15:26
Action: C-Keen throws 16 cores at the build15:36
wpwrakwasn't that roughly what they did at chernobyl ? :)15:37
C-Keenthe threw that against the building :)15:37
C-Keenhm building glib-dbus is broken: /home/ckellerm/proj/nanonote/openwrt-xburst/staging_dir/target-mipsel_uClibc- error: #error "Only <glib.h> can be included directly."16:48
C-Keenand feeds/qipackages/gs needs a +libpthread as dependency16:49
wpwrakthe fragility of that process is scary16:55
wpwrakit seems hard to get things right, though. i remember pretty much the same sort of issues from OE. there, it was worse, though, but that was probably because of the large number of packages in the system16:56
C-KeenI think the system in general grew too large for buildroot16:57
paul_boddieAnything related to Gtk+/Glib is probably going to be horrible.16:58
C-Keenthere is no proper support for resuming builds, cleaning only custom targets and rebuilding them etc16:58
C-Keenpaul_boddie: #t16:58
wpwrakare you building owrt just for the toolchain ? or do you need something else from that build ?16:58
C-KeenI want to use it as a demo for using buildroot but I don't think it will be a good example. And yes originally I just wanted the toolchain and I may go back to that17:00
wpwraki wonder if we could trick xiangfu into making some pre-built cross-toolchain packages17:04
roheh.. openwrt has a target to only build a toolchain afaik17:05
wpwrakgetting the toolchain seems to be by far the most common reason for getting into that process. and basically all the weird things that happen then tend to quite overwhelm the respective victim17:05
rohi think they call it 'SDK'17:06
rohintended to be useable to build fitting binaries to installed images from the same release/snap17:06
wpwrakthat's pretty much what we should have (packaged)17:07
rohactually i think the documentation is much better than what we had at openmoko for oe... also things are much easier ;) http://wiki.openwrt.org/doc/start is a good starting point17:08
wpwrakOE at OM was the mother of all clusterfucks :) but the basic flaw was the same: drag people who shouldn't have to worry about the distribution into having to build it17:09
wpwrakthings got pretty smooth for me once i had figured out how to create a barrier between me and OE :) (with zecke providing the packages and my myroot combining them into a rootfs)17:11
rohyes and no. we had no seperation of distro maintainers and coders17:11
rohso yes that fucks stuff up17:11
C-Keenwpwrak: the toolchain isn't the problem he already provides those in the download folder together with the images17:12
rohand i dont think that was the coders fault or the distro guys.. that was a faulty management decision from what i see now.17:12
wpwrakC-Keen: ah, i see. looks good. let's see if it works :)17:13
rohi think management didnt understand that building a distro and developing sw is 2 different pair of shoes with different goals and or methods17:14
C-KeenI do wonder though why the releases aren't tagged. Together with all changes that were required to make the build17:14
wpwrakroh: yes, wolfgang never cracked that one. that mode of operation had existed from the very beginning, though.17:15
wpwrakone problem was also that developers developed inside the OE build system, basically running bitbake as if it was "make"17:15
rohwpwrak: ack. but i dont think it was his fault either... there was just no understanding from higher up that such a thing would have been neccessary.17:15
rohwpwrak: they should be free to do so if they want. but i think nobody should be forced to17:16
wpwrakroh: he simply inherited the mess. ordered it a little, but didn't kill it :)17:16
rohheh.. like we did with the admin stuff? *cough* ;)17:17
wpwrakyeah ;-)17:18
wpwrakhmm, nice. i found a way to squeeze improvements in the order of 10-20 ns out of the trigger delay, by increasing the MSC clock speed. that's in the range of 1/4 to 1/3 of an MSC clock cycle. somehow, this doesn't seem to be worth it ... :)17:20
wpwrakhmm. average trigger delay = 1 sample + 440 ns. kinda sluggish.17:38

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