#qi-hardware IRC log for Tuesday, 2011-02-15

wpwraktuxbrain: (save to a real editor often) i think we agree on the overall suckiness of web-based interfaces ;-)00:18
kristianpaulwolfspraul: yes did the rework my self, is prety easy i think00:47
kristianpaulyou should try indeed00:47
wolfspraulkristianpaul: I don't have any m1 yet00:52
wolfspraulcannot afford it :-)00:53
wolfspraul(just kidding)00:53
wolfspraulI do need one for a variety of reasons, and I think I will soon get it. need to ask nicely.00:53
wolfspraulI will get one with high-speed fix applied.00:53
kristianpaulwolfspraul: (mm1) i think i read you buyed one at 27C300:54
kristianpaul(need to ask nicel) to whom? ;-)00:56
kristianpaulwell now that wpwrak hang on on #milkymist he should get one as well00:56
kristianpaulor port mm1 to his usrp ;-)00:56
wpwrak(mm1 on usrp) heh ;-)00:59
kristianpaulwolfspraul: noticed peterka thread/talk?01:00
kristianpauli'm confused if that spam, seems not, but...01:00
kristianpaulah nv01:01
kristianpaulhe dint realized it was a mediawiki feature01:01
wpwrakusrp2 has only an "XC3Sxx00". so that's 8k-30k cells vs. 43k in mm1. also, only 294-589 kb memory, instead of 2.1 Mb01:05
wolfspraulkristianpaul: no spam I think.01:06
wolfspraulsorry I am so curious - any gps data news?01:06
wolfspraulI am thinking about a trip to sige, still cannot convince myself it's worth it (maybe a little later...)01:07
kristianpaulno news sorry01:08
kristianpaulwell not about fixes point, thats a NEW01:08
kristianpaullater i think01:09
tuxbrainwpwrak: whats the link of projects for UBB, I'm navigating the wpan tree an I found nothing...01:14
zrafatuxbrain: maybe you need to navigate led blinking01:24
wpwraktuxbrain: http://projects.qi-hardware.com/index.php/p/ben-blinkenlights/source/tree/master/ubb01:26
tuxbrainok :)01:26
zrafathat is I meant :P01:26
wpwraktuxbrain: (i'll just dump all the trivial 8:10 card projects in there)01:26
wpwraktuxbrain: ... and the actual blinkenlights will move to a subdirectory at some point in time01:27
tuxbraincommon , man UBB will be the second product to production made fully with KiCAD :P, some respect man,  at least don't call it "trivial" in public... poor UBB :P01:29
Action: kristianpaul wolfspraul me either01:30
Action: kristianpaul wolfspraul well you can already manufacture that sige-like board, it will work and in parallel way the development can be made01:31
kristianpauli think :-)01:32
tuxbrainpage done, review and if you found something wrong or in accurate, fix it dam it ,is a wiki page01:35
tuxbrainof course geniouses like wolfspraul, wpwrak, kristianpaul .... uff ok any one , report and I will try to fix it ... when ... mmm ... well I will try.:P01:37
steve|mtuxbrain: Gerver files generated from KiCAD? ;)01:40
tuxbrainwpwrak: is not true? there is a proccess in the middle?01:41
tuxbrainany unaccurated info please let me know.01:42
wpwraks/tinny/tiny/;s/it self/itself/;s/striper/stripper/01:42
wpwraktuxbrain: he meant s/Gerver/Gerber/01:42
steve|mtuxbrain: no, it's just a typo01:42
wpwraks/one the/on the/ # werner is dyslexic01:43
tuxbrainok :P fixed01:43
wpwraks/a extension/an extension/ # this guy is really horrible01:43
tuxbrainwell dude time to bed01:44
wpwraki'll just edit the typos directly01:45
kristianpaula google leak ! www.sige.com/files/SE4165T_Preview.pd02:05
kristianpaulphew thats good02:05
kristianpaulwolfspraul: http://gnss.lt/saugykla/Straipsniai/2010%2001%20GPS%20World%20GPS%20RECEIVER%20SURVEY%202010.pdf02:05
wolfspraulkristianpaul: oh wow, that's a lot of data. NIIICE!02:06
kristianpaulyeah i found it time a go, just was too much data for me ;)02:07
wpwrakyears later ... :)02:09
kristianpaulwolfspraul: when you see lots of N/A things get interesting for us i think02:09
kristianpaulmost of then are black boxes IC, not low level black box02:09
kristianpaulzarlink is not there, so wpwrak is right..02:11
steve|mehrm, german google wtf? http://www.google.de/#hl=de&source=hp&q=SE4165T_Datasheet.pdf02:11
wolfspraulkristianpaul: wpwrak is right about what? (wrt zarlink)02:12
wpwraksteve|m: ;-))02:12
wolfspraulthe overview doesn't have much data on ics though, just high-level features. and even those may just be copy/paste from datasheets/tech specs? (in other words sometimes the fine print may take some of it back :-))02:12
kristianpaulwolfspraul: some outdated info..02:14
kristianpaulsteve|m: hehe02:17
steve|m"SE4165T Datasheet Rev1p6.pdf" would be the filename according to their naming scheme.. too bad02:17
kristianpaulhttp://www.sige.com/products/gps/details.html <- 4165 is already there, why sige dont use rss ! :)02:18
qi-bot[commit] Werner Almesberger: atusb.brd: moved some component refernces to improve visual appearance http://qi-hw.com/p/ben-wpan/81818ca02:34
qi-bot[commit] Werner Almesberger: bom: added the ATmega32U2 http://qi-hw.com/p/ben-wpan/68b262702:50
qi-bot[commit] Werner Almesberger: nanonote.patch: cleanup and properly power down on exit (also turns off LED) http://qi-hw.com/p/ben-blinkenlights/0c56f0503:39
qi-bot[commit] Werner Almesberger: cameo: allow comma between parameters of "drill" (makes use in scripts easier) http://qi-hw.com/p/cae-tools/0826d5b05:09
qi-bot[commit] Werner Almesberger: cameo/templates/mkmk-simple: template for milling simple boards http://qi-hw.com/p/cae-tools/b0eb01705:09
qi-bot[commit] Werner Almesberger: mkmk-simple: oops, should be executable http://qi-hw.com/p/cae-tools/4ae218c05:13
kyakxMff: $(TOPDIR)/$(FILES_DIR) is not expanded to what i want05:56
kyaki don't really understand why $(FILES_DIR) expands to ./files.. What's the point?05:57
kyakwhy not expand to the absolute path?05:57
kyakxMff: could be a misunderstnading. I would like to have an expansion to e.g. feeds/packages/utils/vim/files (i.e. the files dir of the package, not the files dir in the top dir of build root)06:08
rjeffries!seem wpwrak06:13
wpwraki only disconnect if there's some network/computer problem :)06:14
kyakyeah.. turning off your computer is so Windows 98 :)_06:14
larscrunning applications on you local machine is so windows xp ;)06:18
rjeffrieshow about letting your computer enter sleep mode?06:20
kyakhaving a PC at all is so... not iPad --)06:20
rjeffrieswpwrak please keep this a TOTAL SECRET yomorrow I drive 80 miles to pick up a Ben Nanonote.06:21
wpwrakwhee !06:21
rjeffrieswhatever you do PLEASE not tell: rejon, tuxbrain, lekernel and the rest of my HUGE fan club.06:22
rjeffriesyou can tell wolfspraul, but he will think it is just a PROOF OF CONCEPT06:23
kyakwhy do you have to drive so far?06:24
rjeffriesI have other business in Santa Barbara where I used to work. I am picking up Nanonote as well as three or so other tasks06:24
rjeffriesgas is expensive these days.06:25
kyakit seems that you have some logistics background :)06:25
rjeffriesnext I will face the nightmare I mean "challemge" of figuring out when it is worth reflashing06:26
rjeffrieskyak not really. I am the kind who knows just enough to screw things up.06:26
kyakand why is that a nightmare? it is as easy as running reflash_ben.sh06:26
rjeffrieskyak I will print that comment out and frame it on my wall.06:27
rjeffriesthe only reason I am borrowing this Ben is to be able to check if Tuxbrain UBB PCB have good mechanical fit. ;))06:28
kyakwell, it had never been harder for me than that06:28
wpwrakkyak: does is also install usbboot on the host ? :)06:28
rjeffriesdo you use Jlime od the qi-haredware OpenWrt or Debian or Vista? just kidding about Vista06:29
kyakwpwrak: it doesn't but it gives a note to install it if it's not found06:29
rjeffriesthis Ben ha sbeen flashed but I do not know when06:29
rjeffriesI figure it does not even have to be turned on to see how well06:30
rjeffriesa UBB fits. ;)06:30
wpwrakkyak: that it does. so it's running the script, plus diligently executing the instructions of your electronic master ;-)06:30
wpwrakrjeffries: you'll almost certainly want to reflash. and i agree that it's not hard. the script is great.06:31
rjeffriesI will be checking the wiki, approaching it as a virgin, seeing if this makes sense to someone with limited intelligence but a great sense if humor06:31
rjeffriesI will not ask any questions on irc, that would be CHEATING06:32
kyakwpwrak: well, if a person runs reflash_ben.sh, sees the pretty explanatory message "can't boot device - xburst-tools setup correctly? device in boot-mode? device connected?" and he can't figure out what to do next, then he just doesn't deserve owning a Ben06:32
rjeffrieskyak NOW you understand why rjeffries hAS NOT PURCHASED A bEN nANONOTE. hE DOESN'T deserve TO WON THIS MILITARY GRADE ELECTRONICS06:33
rjeffriessorry for caps lock06:33
kyakit looks like your keyboard has deserved to be replaced :)06:33
rjeffriesno this idiot lappie has a shift lock that can be hit too easily, and pronlem is worst when a monkey is typing06:34
rjeffrieskeyboard is great I can't type worth beans.06:34
wpwrakxmodmap tends to get rid of undesirable modifier keys quite efficiently :)06:35
rjeffrieswpwrak let me take a wild as guess...06:35
wpwrakas in  remove Lock    = Caps_Lock    :)06:35
rjeffriesthat might be a LINUX program, no?06:35
rjeffriesjust a guess... lol06:36
wpwrakand maybe   keycode 0x42 =  Control_L06:36
rjeffriesbut seriously06:36
wpwrakwell, of course ...06:36
rjeffriesit will be fun, in a perverse way, to see how someone who has not taken the Copyleft Vows manages to get by with a Ben06:37
rjeffriescan Ben read a FAT formatted microSD?06:38
rjeffriesI assume no, or maybe HELL NO06:38
tuxbrain_awayrjeffries: yes it can06:38
rjeffriestuxbrain_away is supposed to be away I am so totally confused. ;)06:39
rjeffriesthat is A Good Thing because this cool Zoom H1 writes .wav files to FAT format ,icroSD06:40
wpwrakwell, since 1992, the linux kernel can read FAT. before, there were mtools. MMC/SD card support came a little later ...06:40
rjeffriesIf ben can read, that is a start06:40
rjeffriesit just seems that reading a disk format that Bill Gates invented is so, well, morally corrupt. ;)06:41
rjeffriesI am kidding guys06:41
tuxbrain_awayben can read/write/format and surelly recover fat partitions06:41
rjeffriesnever ever take anything I say seriouly. My wife knows that, and so should you06:41
wpwrakyeah, i have to plead guilty there ...06:41
wpwraktuxbrain_away: (recover) there too, i'm afraid06:42
rjeffriesso if I transfer .wav files into Ben, what tool will play them? Is there a soud editor, or must they be converted to .OGG06:42
rjeffriesI am getting worried that I may have found a possible use case for Ben. That can't be good.06:43
wpwrakhmm, i'm getting good at milling ubbs. got one that's 11.02 mm wide. 20 microns tolerance. beat this, industry ! ;-)07:13
wpwraktuxbrain_away: btw, any reaction so far from pulster regarding ubb ?07:19
qi-bot[commit] Werner Almesberger: ubb/cam/Makefile: use mkmk-simple http://qi-hw.com/p/ben-blinkenlights/daa633107:23
rjeffriesis this your final answer?23:11 <wpwrak> getting good at milling ubbs. got one that's 11.02 mm wide. 20 microns tolerance07:32
rjeffriesnot bad for a hand-held Dremel tool. ;)07:33
wpwrakrjeffries: naw, i didn't dremel that ...07:35
rjeffrieswpwrak I know that, but the image was too good to let it pass07:49
rjeffriesoff to bed07:49
qi-bot[commit] Xiangfu Liu: add recently new package icons http://qi-hw.com/p/gmenu2x/bb826e008:16
qi-bot[commit] Xiangfu Liu: update-F4-loading-kernel-from-ubifs_rootfs-bin-uImag http://qi-hw.com/p/openwrt-packages/a0ee30908:17
qi-bot[commit] Xiangfu Liu: gmenu2x, offrss, nethack small fix, update http://qi-hw.com/p/openwrt-packages/0ec093b08:17
dvdkxiangfu: regarding commit http://qi-hw.com/p/gmenu2x/bb826e0 : can use gforth.png icon that i already added to gmenu2x repo10:16
xiangfudvdk: oh. thanks.10:19
dvdkxiangfu: btw can't await the next release with all the recent changes and additions :)10:20
xiangfudvdk: the image I want release was compile complete. : http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-02132011-0750/10:20
xiangfudvdk: :)10:20
xiangfunow I am downloading... and test.10:20
dvdkxiangfu: then i guess you had to remove plplot, since i didn't get around to fixing it?10:20
xiangfudvdk: not include plplot .10:21
dvdkxiangfu: well, next time :)10:21
xiangfualso I this image not include gcc-mips.10:21
xiangfuthe gcc-mips ~70M. so I remove it for saving place.10:21
dvdkxiangfu: btw plplot build for me fine.  looking at your log it looks like your pango stuff in staging_dir is a little broken (pkg-config wise).  will have to try again with all the pango stuff, maybe i will finally manage to break it, just like you :)10:22
dvdk(i.e. plplot only attempts to link pango if it finds it, with my minimal .config, there was no pango anyways)10:23
dvdkxiangfu: moving the image  to 'testing' stage for everyone to try it before release?  or again "direct" release  this time?10:23
kyakxiangfu: can you include the "supertux" in the image by any chance? :)10:25
kyakdvdk: btw, don't bother me seeking you :) i'm just randomly testing the seen feature of the bot. What did he tell you?10:26
dvdktold me 'kyak was looking for you via /msg at 9:00' or something10:26
dvdkwas still sleeping at 910:26
kyakcool :)10:26
kyakxiangfu: i agree to throw away the gcc-mips. Whoever needs it, can install it via opkg.. It's too huge10:27
dvdkkyak: if supertx didn't make it in time, what about hosting the ipk it peple.qi-hardware or something, and dropping a link on irc/ml10:28
dvdkkyak: this way you get some exposition/testing already10:28
dvdke.g. here is where i put my preliminarily ported stuff http://mosquito.dyndns.tv/~spock/openwrt/10:29
kyakdvdk: http://downloads.qi-hardware.com/people/kyak/tmp/supertux_0.1.3-1_xburst.ipk10:29
dvdkkyak: not bad.  btw how did you manage to write to people/kyak?  last time i tried write-acess via webdav, i was blocked10:29
kyakdvdk: i'd prefer not giving the package, which would work most likely.. It is vital that someone else tries to buidl it :)10:30
kyakdvdk: you should ask wolfgang, he would give you an access10:30
dvdkkyak: would be nice if somebody tried. but.. i guess nobody ever will :)10:31
dvdkso far i think only xiangfu ever tries to build the packages i port :)10:32
kyakdvdk: i doubt there are three people on Earth who are playing with openwrt with regards to Ben10:34
dvdkkyak: given how buggy my initial ports were, and how many bug reports i saw, i'd say exacty 3 people tried to build (include me :)10:35
dvdkkyak: or maybe just nobody was interested in the stuff i ported.  i guess supertux will be more famous.10:36
kyakwell, you can try and inject a bug into one of your packages. Let's see where the screams will come from :)10:37
Action: dvdk thinks about injecting bugs into *kyak*'s packages10:37
kyakthere are regular reports in mailing lists and here in IRC about something failing to build10:38
dvdkok gotta work10:39
kyakwork bug-safe!10:40
qi-bot[commit] kyak: backgammon won't fail miserably on small terminals http://qi-hw.com/p/openwrt-packages/7495f4510:46
wolfsprauldvdk: no way - many more than 3 are following what you do. as for building, sure, it's just a few people and imo that's fine. I will not overly encourage (or beg) anyone to start building.11:08
wolfspraulwe have a buildhost, if that helps, but I think only 3-4 people have accounts on the buildhost right now.11:09
wolfspraulI think as config.full_system grows, less people will build it from scratch, and I want to work towards upgrades via opkg, instead of full-image reflashing.11:09
dvdkwolfspraul: yeah, was takling about people who attempt to build.  always posting .ipks because i had the feeling the barrier to building was too high.11:11
dvdkwolfspraul: opkg-based updates will be great.  hopefully we have enough place on flash for .ipk download+install11:11
dvdkwolfspraul: the problem with building currently is, i have to remember doing a 'git pull && make world' the evening before i want to try building newly ported packages :)11:13
dvdkso compilation finishes in time :)11:13
wolfspraulxiangfu is also looking into werner's pivot_root suggestion to have multiple rootfs'es in one large (2 gb in our case) partition11:32
wolfspraulI saw your comments about the 512mb one getting full, and I also want to make progress on jlime dual-booting.11:32
dvdkwolfspraul: had a look at the qilb60 schematics+flash datasheet.  maybe one could improve the flash driver to cut down on UBI attach time11:50
dvdkUBI attach is pretty special in the read pattern that is required11:51
wolfsprauldvdk: interesting. you think significant speed improvements are possible?13:56
dvdkwolfspraul: maximum factor of 2, less if the kernel driver is already smartly implemented.  > factor of 2 only, if the kernel driver is much dumber than i expect.  need to take a closer look.14:01
wolfspraulhmm, very interesting - thanks a lot even just for the idea!14:02
wolfspraulbootup time is very important indeed14:02
wolfsprauland the big problem about bootup time is - once it's out of control you can never bring it down again14:02
wolfspraulI'm very happy that the openwrt images we have right now keep around 15 seconds, for 1 year already14:03
wolfspraulit oscillated between 11 and 18 I think :-)14:03
wolfspraulI am watching this number like an eagle...14:03
dvdkwolfspraul: more speed gain would be possible by artifically increasing the erase block size, since UBI attach is proportional to number of erase blocks.14:04
wolfspraulbtw, xiangfu did a quick test with the linux kernel image inside ubifs, and it confirmed larsc suspicion - 4 seconds more!14:04
dvdkyeah, read that yesterday, that made me look into the flash datasheets.14:04
wolfspraulso if we increase the partition size from 512 mb to 2 gb, we may easily bump up even the openwrt boot time to 30 seconds or so - not good14:04
wolfspraulsure, if bigger erase blocks help and the downside is acceptable - why not14:05
dvdkin fact, this is not a problem with ubifs.  the problem is with the UBI layer that reads a short header of every erase block.14:05
dvdkone optimisation would be "dual plane reads" that can read two headers with one mmc command.  the other optimisation is to just pretend erase blocks are bigger by treating subsequent number of erase blocks as one.14:06
dvdkneed to look into the details, so take everything w/ a grain of salt.14:07
xiangfudvdk: not I am include the 'supertux' , generate the image again. include one package. compile rootfs. needs ~1 hours.14:08
dvdkbtw and using ubi layer from boot loader looks like a bad idea.  twice the penalty, once in boot-loader, later in kernel.14:09
wolfsprauldvdk: I was hoping that the u-boot ubifs/ubi implementation would focus on just getting the binary out, without need for full scans14:10
wolfsprauleither that is not the case because nobody optimized for that, or because it is technically impossible (I still cannot believe that but I won't dig into the sources...)14:10
wolfspraulare the blocks totally randomized with no map whatsoever?14:10
dvdkwolfspraul: maybe some optimization is possible, dunno.14:10
mthwhat about a separate small ubifs partition that contains only kernel images?14:10
wolfspraulsure, possible14:11
dvdkmth: that won't help, since the time is involved at the UBI attach, which is required to even know wher that "partition" resides.14:11
wolfspraulthe problem is to keep changes as rare as possible, because it takes so much time to trickle into different distributions, openwrt, jlime, debian, etc.14:11
mthdvdk: I mean a NAND partition, not a partition on a block device14:11
wolfsprauldvdk: then 2 separate partitions, not 2 ubifs 'volumes' inside the same partition14:11
dvdkmth: you mean a mtd partition?  this is how it is done currently, isn't it (albeit without a file-system on top).14:12
dvdkwolfspraul (keep changes rare): i think any tuning could be (should be) done as part of the qilb60-specific kernel drivers.14:13
dvdkmth: wolfspraul: note that much of what i wrot is still wild guessing.  i'll know more by next week.14:13
kyakxiangfu: thanks!14:14
mthI'm surprised that building the map takes so long though14:15
wolfspraulmth: if the blocks are really anywhere (totally randomized), with zero maps or hints/guides/whatever, then what can you do14:16
wolfspraulI cannot believe that really, but who knows. ubifs is supposedly a fairly sophisticated nand file system.14:16
mthin OpenDingux we scan the file system on the NAND (non-ubi) and while I haven't benchmarked it I doubt it takes more than 1 second for 4GB14:16
dvdkwolfspraul: ubifs is sophisticated and mounts quickly.  the problemn is the ubi layer below that.14:16
dvdkmth: maybe you're lucky and your erase-blocks are just larger?14:16
mtherase size is 512K14:17
wolfspraulI can see how central maps/tables cause big problems with wear-leveling. so maybe they have no such thing at all and blocks are 100% random? :-)14:17
wolfspraulI don't know, just wildly guessing.14:17
mthso 8192 blocks14:18
wolfspraulif there is not a lot of write activity, larger erase sizes shouldn't matter much, right?14:18
mthwe only read the first OOB area of each block though, I don't know if that's the same for UBI14:18
wolfspraulit only means that a write will cause a bigger amount of reshuffling to preserve the rest of the (larger) eraseblock14:18
dvdkmth: 8192*64 bytes to be transfaired, afair.  not even half a megabyte, but what about the latencies of the flash accesses?14:19
mthoobsize is 128 bytes for this flash14:19
mthI very naievely implemented an FTL driver for the native (eCOS/II based) OS of the Dingoo14:20
dvdkmth: do you use any kind of command queueing optimization for the 8192 transfers?14:20
mthsince it was fast enough, I never benchmarked or optimized it14:20
mthno, just mtd->read_oob() inside a loop14:21
dvdkmaybe the linux driver transfers full pages instead just the header?  i mean the full page is loaded to flash-internal buffers anyway, so we're suffering from a MMC bus bottleneck?14:21
dvdkbtw for the NN erase blocks are also 512K iirc14:23
dvdkjust looking at the timing, full page transfers would approximately explain the long ubi attach time (4096 blocks a 4096 bytes in 30 seconds)14:25
mththe UBI scan routine mentions it reads EC and VID headers, but I have no idea what those are14:27
wolfsprauldvdk: I think it takes about 8 seconds to attach the entire 2 gb14:27
wolfspraulnot 3014:27
mtherase counter and volume ID?14:27
dvdkwolfspraul: didn't you write 30s yourself?14:28
dvdk"increase the partition size from 512 mb to 214:28
dvdk                   gb, we may easily bump up even the openwrt boot time to14:28
dvdk                   30 seconds or so"14:28
mthnote: the FTL driver was for Linux, but designed to be compatible with the native OS14:28
dvdkthat's not purely the attach time?14:28
wolfspraulI meant if we move the Linux kernel into the ubifs, and increase the ubifs from 512mb to 2gb, I have a feeling without further improvements the total openwrt boottime will go up from 16 sec or so where it is now to > 30 sec.14:29
wolfspraulbefore your ideas, I didn't have genuine ubi/ubifs performance improvements on the radar14:30
wolfspraulI was just trying to measure how long it takes to mount a 2gb partition, and to move the Linux kernel inside a partition (if that is even a benefit, we can think about it...)14:30
dvdkI'll look into the datasheets and sources and do some math to see how much can be gained, then put the results into a wiki-page.  but not before saturday.14:32
wolfspraulno rush, this would be very valuable in the long run, so I'm very much looking forward to it14:34
kyakdvdk: i see you are pushing towards the 2Gb rootfs. But have you thought about having a safe place across between reflashes?14:34
dvdkkyak: mmc card :)14:35
dvdkkyak: but i only need 128MB of safe place :)14:35
kyakmy mmc card is 256Mb and it has Jlime on it. moreover, the mmc card drops off easily and i don't like it14:36
kyaki need safe place for video/music/pictures14:36
xiangfudvdk: quick question. when I run 'gforth' in new release. it give me "can't load library 'libffi.so.4'"14:36
Action: dvdk scratches his had14:36
xiangfudvdk: I found the 'package' libffi have update to '5'14:36
dvdkxiangfu: but it updated *after* you compiled gforth?  how's that possible?14:37
xiangfudvdk: by make symlink: libffi.so.4 -> libffi.so.5.0.10 . gforth works fine.14:37
dvdkxiangfu: can we somehow push that change into the release image?14:37
kyakseems that libffi was recompiled after gforth has been built14:38
dvdkbut libffi is referenced by gforth's dependency line, is it not (or not correctly?)14:38
xiangfucompile by crontab jobs. not manually re-compile .14:39
kyakin theory, make should've rebuilt gforth after its dependency (i.e. libffi had been rebuilt)14:40
kyakbut who knows &14:40
dvdkhopefully the libffi problem doesn't break other script languages lua/python/octave etc.14:41
kyakdvdk: btw, i'm reflashing pretty often. Sometimes several times a day.. and i'm really happy with these 1730Mb of safe space14:41
dvdkkyak: 1500Mb nowadays, isn't it?14:41
kyakstill 1730 for me.. i like 256 mb rootfs :)14:41
Action: dvdk hasn't even formatted the unused part14:41
xiangfulua, python, octave works fine. as least 'prompt is come up'14:42
dvdkxiangfu: ok, so what are you going to do about the problem?14:44
Action: dvdk is going to check libffi dependency in gforth Makefile14:44
xiangfuthere are two folder under build_dir/.../  libffi-3.0.9/      libffi-sable-3325/14:45
xiangfufor me. temporary work around. create a symlink when first boot. :)  really don't want re-compile the whole openwrt.14:46
xiangfuit's needs ~30 hours.14:46
kyakyou could just rebuild gforth, won't take that long..14:47
kyakafter dvdk adds this dependency :)14:48
dvdkcan we patch in the symlink without a rebuild, then just re-create the ubifs?14:48
dvdkkyak: afair, i added the dependency14:48
kyak  TITLE:=GForth14:48
kyak  DEPENDS:= +libltdl14:48
Action: dvdk shakes his had in disbelief14:48
xiangfudvdk: manually we can. by using openwrt command "make target/install" it needs 1 hours in buildhost.14:49
dvdkat least now i now how important this dependency stuff actually is:)14:49
xiangfuopenwrt try to rebuild  a lot of package even there is no change. :(14:49
kyakhm, i never noticed such behaviour14:50
dvdkxiangfu: patching the symlink would be sufficient for now, since libffi isn't actively used in gforth currently.14:50
kyakdvdk: did you actually took it seriou swhen we discussed about injecting bugs in the morning? ;)14:50
dvdkxiangfu: but i'd like that patch to be part of the ubifs image, not everybody is going to understand how to patch it himself14:51
xiangfudvdk: sure. easy add then to openwrt-package.git/nanonote-files/scripts-files/etc/ubi-default/14:51
dvdkxiangfu: thanks a lot.  now i can play "brainless" chess out-of-the-box.14:52
xiangfudvdk: this folder "etc/uci-defaults" will only run once at first boot.14:52
Action: kyak is ready to switch to 512 Mb rootfs just to test the image when it's released14:53
dvdkxiangfu: didn't know about that, adding it there is ok, sure.14:53
dvdkkyak: but where to but your safe data before reflashing? :)14:53
kyakdvdk: this time, i'm ready to spend the time and back it up14:54
kyakcopy to host, then copy it back14:54
kyakxiangfu: http://en.qi-hardware.com/wiki/Format_Data_Partition, ubimkvol /dev/ubi1 -s 1400MiB -N data - is it correct?14:55
kyaki mean, 140014:55
kyakand i'm going home, too.. see you later14:57
xiangfunot exactly , 1400~1500.14:57
xiangfu1480 is MAX I think. just manually test.14:59
xiangfuI have test 1500, 1490, 1480. 1480 is success create a volume on data partition.14:59
xiangfukyak: see you.15:04
xiangfujust test 1486 is max in my nanonote, maybe there are different bad block in nand.15:04
xiangfuwiki page update.15:07
qi-bot[commit] Xiangfu Liu: work around for gforth http://qi-hw.com/p/openwrt-packages/504dfdc15:10
qi-bot[commit] Xiangfu Liu: increase the data partition size to 1480MB http://qi-hw.com/p/openwrt-packages/ebafec915:10
Action: xiangfu can't wait for compile goto sleep. check that tomorrow: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-02132011-0750/15:28
rjeffriesin smuxi irc timestamps are my local time. jsut sayin'15:28
zrafa(Xiangfu Liu: increase the data partition size to 1480MB) new layout? :)17:03
zrafakyak: or was it already that size before and he forgot to commit?17:04
kyakzrafa: he's just adjusting the leftover size17:24
kyaki.e. boot and rootfs haven't changed17:25
zrafakyak: ah.. okey, so only new size for data partition then17:42
wpwrakargh ! i just realized that i can probably make atben a lot smaller18:55
kristianpaulwpwrak: how come?19:55
Jay7hm.. seems there are some people anknowledged with UBIFS internals :)20:06
Jay7kexecboot need ubifs support20:07
Jay7at least for detection that there is ubi/ubifs on nand20:07
kristianpaulJay7: detection, no more?20:14
Jay7kristianpaul: well.. this is start move :)20:15
Jay7first thing we need20:15
Jay7then it is possible to implement attaching and mounting20:15
Jay7seems I should do this for LVM as well20:16
wpwrakkristianpaul: i realized that, if i allow a minor length difference in a pair of RF traces, i can rotate the chip such that all the digital signals are directly on the ben side. then i don't need to run all the traces around that 90 degree corner. also the crystal ends up at better location.20:18
wpwrakkristianpaul: i'll get a phase difference of about 1 degree, which according to simulation, shouldn't have any noticeable effect20:25
kristianpaulsignals directly is almost perfect !20:26
wpwrakkristianpaul: yeah. i expect that the much cleaner layout will make up for anything i get in terms of imperfections due to the mismatched RF traces20:28
tuxbrainwpwrak: (atben size) well better realize now than already in production :)20:30
wpwraktuxbrain: yeah :) that's pretty much the last moment :)20:32
wpwraktuxbrain: btw, regarding UBB, any reaction from pulster ? i would assume he'd be your largest buyer, no ?20:34
tuxbrainfor the moment just a 10 units bag to analize the product (as I also do) but surelly he will become my best costumer...as allways :) well maybe ron will compete with him this time :P20:36
tuxbrainI think I will do preorders but I will not wait any amount to place the order I will place it anyway. UBB must be done20:37
wpwrakyes ! that's the spirit ! :)20:41
kristianpaul+1 tuxbrain ;-)20:54
rjeffriesthis rocks, totally. http://nokiaplanb.com/21:42
qi-bot[commit] David Kühling: gforth: add missing libffi dependency http://qi-hw.com/p/openwrt-packages/cee4a8d21:44
qi-bot[commit] David Kühling: plplot: disable detection of pango for now, since that can break compilation http://qi-hw.com/p/openwrt-packages/9ab42d321:44
qi-bot[commit] David Kühling: gforth: bump version number (this version already distributed via release?) http://qi-hw.com/p/openwrt-packages/5f81a8d21:44
dvdkxiangfu: btw looks like the image you're building will also miss the Brainless chess game?21:45
dvdkxiangfu: mind to still add it?  does not require any compilation, as it's just some bunch of forth source code.  gmenu2x icon&starter for Brainless is already in the image.21:46
kristianpaul"Increase the lifespan of Symbian to a minimum of 5 years." no pleasee noo !!22:01
wpwrakare there no pictures of elop with horns yet ? :)22:06
qi-bot[commit] David Kühling: use gforth.png icon for gforth http://qi-hw.com/p/gmenu2x/bd8ee7222:06
qi-bot[commit] David Kühling: add launcher for Octave http://qi-hw.com/p/gmenu2x/f1c4bd822:06
rjeffries@segphault: How many Elops does it take to screw in a lightbulb? None, he has Steve Ballmer do it because he's too busy screwing Nokia.22:09
mthsounds like they are not just shareholders, but developers at Nokia22:10
mthsince one of the problems they name is lots of outsourcing and layers of management trying to make the outsourcing work22:11
wpwrakhmm, i'm getting good at this. redoing the atben layout only takes me 2-3 hours now. i can almost do it with eyes closed :)22:11
Jay7rjeffries: look at nokiaplans.com22:14
wpwrakmth: they also seem to quite strongly dislike R&D outside of helsinki. i mean, helsinki is a nice place in summer, but ...22:14
mthI was surprised that Nokia turned to Microsoft for help: while Nokia has been lagging behind their competitors, but Microsoft failed in a much bigger way in the mobile world22:15
qi-bot[commit] David Kühling: add octave icon http://qi-hw.com/p/gmenu2x/58eae7422:15
mthwpwrak: I don't think it's just their love of Helsinki, but the problems they see with outsourcing22:15
mthdeveloping in multiple locations is already difficult within a company, but if the other location is another company it get really difficult22:16
mthand if it's not just one company but several, it's even worse22:16
rjeffriestuxbrain do you have firm quoutes and schedulke with your pcb fab house yet?22:16
wpwrakmth: (outsourcing) yeah. don't know that part of nokia.22:17
mthI don't know the insides of Nokia either, but I've experienced problems with multiple locations in another company22:18
mthcommunication issues, cultural differences, different priorities of different managers etc22:18
wpwrakof course, he _we_ are, scatted all over the planet ... :)22:18
tuxbrainnop I have to pay for so,   take it easy dude22:19
wpwraks/scatted/scattered/  # erhm22:19
dvdkshould write an open letter to our shareholders22:19
mthyes, it can work, but it does require good communication and having common goals22:19
mthit's not just a matter of signing a contract and being done22:19
wpwrakmth: particularly among management and developers ;-)22:19
qi-bot[commit] Werner Almesberger: atusb-pgm/cam/Makefile: use mkmk-simple http://qi-hw.com/p/ben-wpan/110619622:20
qi-bot[commit] Werner Almesberger: atben: rotated the CPU for much tighter layout; removed C9 http://qi-hw.com/p/ben-wpan/c75274f22:20
mthwpwrak: maybe it's actually easier to have every single person at a different location than having teams at a few locations22:26
mthbecause if everyone is somewhere else, you are forced to communicate electronically and then you might as well broadcast it to everyone22:27
wpwrakmth: indeed. but it depends a lot on the people. some adapt naturally. others don't.22:27
wpwrakrjeffries: i already answered your mail yesterday ... you need to work on that read/write ratio ;-)22:29
rjeffrieswpwrak I have a flood of email. I assume your reply went to Doug Bull then?22:30
rjeffriesI guess I need to ask him for the next rev of fab document22:30
mthin open source, someone who doesn't communicate problably never gets to the point of being offered commit rights22:30
wpwrakrjeffries: to you and to him. "Ron K. Jeffries" <rjeffries@gmail.com>22:35
wpwrakrjeffries: subject  Re: WERNER: questions re UBB fab specs22:35
wpwrakmth: so their only choice is to start their own projects, then enjoy each day the complains from people who have sent patches and got ignored :)22:36
wpwrakrjeffries: did you find it ? Date: Mon, 14 Feb 2011 21:15:25 -030022:38
rjeffrieswpwrak you've got mail.22:39
rjeffriesI want him to polish/finalize the spec, which I think is getting close22:40
wpwrak... waiting for mail to arrive ...22:41
rjeffriesyesterday was Feb 14 in USA, a holiday we call Valentines Day. spent it (mainly) doing fun stuff away from computer with my wife22:41
wpwrakah, valentine's day is an official holiday ? didn't know that22:42
rjeffriessent 14:36 my tme (PDT) to you and Doug22:42
rjeffrieswell I do not think it is OFFICIAL but someone married for many year like me has learned it is IMPORTANT holiday22:42
wpwrakmail arrived. took a while to make the first hop.22:43
rjeffriesdoes Tux ok22:43
rjeffriesdid tuxbrain get firm UBB quote and schedule? just curious is all. I assume 800 Ben owners are screaming for your UBB22:45
wpwrakright now, they're quite silent. i think we'll need a few "ice breakers" who will do something cool. then the word will spread.22:50
wpwrakcool but simpel :)22:50
wpwrakrjeffries: regarding batch sizes, this depends on the pcb fab. what i've seen so far is you generally have a certain setup cost and a more-or-less constant per board cost. this then gets somehow translated into a price for you.22:52
wpwrakrjeffries: i.e., you may find that 100 and 500 units have almost the same total cost. the pricings i've seen in this context all make 500-1000 a reasonable order size, while 100 or lower makes no sense.22:55
rjeffriesI think I'll ask for quotes for 500 and 100022:55
rjeffrieswe will get a fisrt article to review I think.22:56
rjeffriess/fisrt/first  that is pretty stanadsrd22:56
rjeffrieswhat is the ticjness tolerance you want to spcfy?22:57
rjeffriess/ticjness/thickness   sigh22:58
rjeffriesI'll be back later have to go pick up a Ben Nanonote22:59
wpwrak(sample) hmm, maybe. the per board production cost (for the fab) is low. probably well below 50 cents. and just setting up the production run may cost them USD 50-100. they may prefer to just do the run, and split the risk - if they did something wrong, you get another one, if the bug is in your data, your bad luck22:59
bartbeskyak: hey, I just added 2 games to the repo23:00
rjeffriesI think you are correct besides ther material cost approaches zero on this tiny job23:00
bartbesEarthDefenders and CartesianCardinalZap23:00
wpwrak(thickness) something like 0.7-0.95 mm.23:00
rjeffriesis the rumor that Ben makes a really functional paperweight correct? or is that jsut talk? ;)23:01
wpwrak1.0 mm may be pushing it.23:01
wpwrakrjeffries: (paperwight) you need to install the counterweight for doing this properly23:01
rjeffrieswpwrak we work in inches as you see from his spec. ;)23:01
rjeffrieswpwrak maybe the counterwight can be my next MONSTER project23:02
wpwrakrjeffries: (in) "units" is your friend :)23:03
rjeffriesis this total thickness or tolerence ? 14:58 <wpwrak> (thickness) something like 0.7-0.95 mm.23:03
wpwraktotal thickness !!23:04
rjeffrieswpwrak I only do this for you my friend LOL23:04
wpwrakin any case, there's little you can do about the thickness. it's a standard PCB thickness and then a few microns are added for copper, plating, etc.23:06
rjeffriesso you would say 0.85 mm plus 0.10 minus 0.15 ????23:06
rjeffrieswell the big boys spec it so I will as well. the fab must work to a final dimension for our buy off23:07
wpwraklet me put it this way: if the fab output isn't what fits, then the board cannot be produced efficiently ;-)23:08
wpwrakso the main parameter is the thickness of the base board23:08
rjeffriesok back later23:08
wpwrakrjeffries: i don't know the right keywords for the other parameters, so i don't try to specify them. chances are that they'd only cause confusion or (if i don't guess the "normal range" right) may accidentally put the board into an unusual category with high production cost23:12
--- Wed Feb 16 201100:00

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