#qi-hardware IRC log for Sunday, 2010-11-21

qi-bot[commit] Werner Almesberger: qpkg: added README and COPYING.GPLv2 http://qi-hw.com/p/wernermisc/78006c502:37
qi-bot[commit] Werner Almesberger: qpkg: Makefile cleanup and added OpenWRT target http://qi-hw.com/p/wernermisc/948193402:37
qi-bot[commit] Werner Almesberger: qpkg/TODO: cleanup and added new task for consideration http://qi-hw.com/p/wernermisc/07a442202:37
qi-bot[commit] Werner Almesberger: qpkg: added copyright headers and changed name to emphasize speediness http://qi-hw.com/p/wernermisc/3d1ef0202:37
qi-bot[commit] Werner Almesberger: qpkg: turn pkg->installed into pkg->flags, allowing for future expansion http://qi-hw.com/p/wernermisc/a63823e02:37
qi-bot[commit] Werner Almesberger: qpkg: towards "struct id" without jrb reference - change iteration and lookup http://qi-hw.com/p/wernermisc/705811502:37
qi-bot[commit] Werner Almesberger: qpkg: towards "struct id" without jrb reference - change ID creation http://qi-hw.com/p/wernermisc/44e443402:37
qi-bot[commit] Werner Almesberger: qpkg: give ID2S macro a more "offical" status http://qi-hw.com/p/wernermisc/dd0ee6602:37
qi-bot[commit] Werner Almesberger: qpkg: record Architecture: tag http://qi-hw.com/p/wernermisc/f14c04102:37
qi-bot[commit] Werner Almesberger: qpkg: perform sanity checks after parsing a package entry http://qi-hw.com/p/wernermisc/83c716402:37
qi-bot[commit] Werner Almesberger: qpkg: added regression test harness and a few tests http://qi-hw.com/p/wernermisc/2c953a502:37
kyakviric: great! could you document it for non-nix users? i.e. kernel options, cmdline, pecularities of ubi creation and then qemu parameters?02:43
wolfspraulkyak: which backend are we using for NetSurf in OpenWrt?02:45
wolfspraulframebuffer? sdl?02:45
kyaksdl backend02:45
wolfspraulwhat are the pros and cons between the two? why did we choose sdl?02:46
wolfspraul(just curious)02:46
kyakits linuxfb is broken02:46
kyaknetsurf's linuxfb02:46
kyakthe guy from #netsurf said02:47
kyak< kyllikki> right now its been reduced to a stub untill i get round to  it02:47
kyakand reason is that02:48
kyak< kyllikki> the API usage changed with the invention of kms and i  never have figured out why netsurfs usage blows up02:48
kyakhowever, netsurf has gtk2 interface02:52
kyakbut i think it will be slower than sdl (need to check of course)02:52
wolfspraulok got it02:53
wolfspraulsdl seems like a reasonable choice then02:53
kyakB_Lizzard tried gtk interface in OE (with X, however). Very slow02:53
wolfspraulsure let's focus on sdl, and maybe linuxfb one day if they fix it02:54
kyakthey fixed already some things "for us"02:55
kyaki don't know if they are willing to go on02:55
kyakcurrently sdl interface has limited support for keyboard02:55
kyakin fact, only arrow keys work :)02:56
kyakbut it becomes more or less usable with keymouse02:56
wolfspraulon the Ben, I think of NetSurf for 2 use cases now: develop 'local' apps in html+css, maybe later combined with some 'server side' (=local) scripting, or JavaScript if NS adds it03:18
wolfsprauland (along the same lines) offline document viewer, for example for offline wiki content03:19
virickyak: I'll try to document that, and send that to qi-hardware05:29
kyakviric: awesome!05:30
qi-bot[commit] Xiangfu Liu: gmenu2x : disable log message http://qi-hw.com/p/openwrt-packages/65934ce09:01
qi-bot[commit] Xiangfu Liu: remove /usr/bin/ben-ash, since we setfont2 in /usr/bin/gmenu2x http://qi-hw.com/p/openwrt-xburst/c8f893a09:58
kristianpauloffline wikipedia in thtml is *huge* i dont see a way there10:42
kristianpaulis the same problem with thhe xml dumps10:42
virickristianpaul: and that compressed file format... I can't remember the name10:42
mstevensoffline wikipedia seems to fit nicely on a wikireader10:44
viricno what?10:44
kristianpaulnoo zim10:44
kristianpaulis like reivent the weel10:44
viricis zim bad?10:44
viricwhat was there before zim? :)10:45
kristianpauli'm not sayin that10:45
viricI only know chm10:45
kristianpaulbefore seem in wikipedia we have xml compressed dumps10:45
viriccompressed how?10:45
kristianpaulthe only tink i dont like from zim is that is not easilly maintenable10:45
kristianpaulnot seekables10:46
kristianpaulnot byt it self10:46
viricxml, compressed, not seekable...   it sounds pretty bad10:46
kristianpaulyou can create the seek-index file10:46
viriceven uncompressing a whole bzip2 block is heavy10:46
kristianpaulthis process look bad for me http://www.openzim.org/Build_your_ZIM_file10:46
viricI think the format has to keep in mind the cpu cycles of the 'reading' device10:47
viricand a very specific format for that seems to me like the best approach10:47
viricah, you mean that zim does not take the XML dump?10:48
kristianpauli agree if is easy to maintain, did you see the graphs for builidn zim files?10:48
kristianpaulis crazyness !10:48
viricwell, you dislike the fetching from wikipedia?10:48
viricthe rest looks fine to me10:49
viricsomeone could write another kind of 'fetch data from wikipedia'10:49
kristianpaulall the data is fetched from wikipedia, bt then you need t create other mediawiki serve plus a expotr plugin then compress again..10:50
viricI imagine they did this process in order to get ZIM workign sooner.10:50
kristianpauli dontlilke that cycle10:50
viricBut what is important for the reader device is having software + data that it can use to show things quickly and well, with little filesystem usage10:51
viricThe way to generate the ZIM files is irrelevant for the reader10:51
kristianpauli'm not talkign from the reader side ;-)10:51
viricah :)10:52
viriczim looks hard to beat, on the reader side10:52
kristianpaulviric: hard to beat? dont think so :)10:52
viricI remember this humane pc10:52
kristianpaulrafa: did some work around it10:53
kristianpauland it works !10:53
viricah he has one?10:53
viricone humane pc10:53
kristianpaulno no10:53
kristianpaulthe little computer all we use here :)10:53
kristianpaulis in jliem beta4 i think10:54
viriche used the humane pc software in the nn?10:54
kristianpaulor sorft of10:54
kristianpaulrafa: can exaplain better what he did..10:54
kristianpaulwhen he wakeup or came back from trip10:54
kristianpaulviric: command line vesion http://www.jlime.com/wiki/documentation/user/nanonote/videos check last video10:56
viricI'll check11:08
jekhorgot ncmpc running on NanoNote :)11:51
kyakjekhor: nice! you could post the patch to mailing lists :)12:20
jekhori know.12:20
jekhorkyak, posted.13:01
kyakviric: hey, are you there?15:08
kyakgood :) i'm following your message in mailing lists15:08
viricbrave :)15:08
kyakso far, trying to boot a bare kernel.. however, qemu is just doing nothing -\15:09
virica malta kernel?15:09
kyaki mean, it definitely doing something, but no output15:09
virictry -nographic15:09
kyakno, i tried to build a nanonote_defconfig15:09
kyakyep, tried -nographic, too15:09
viricqemu cannot emulate the nanonote15:09
kyakdoes it care?15:10
viricit's about the *board*15:10
viricat what IO address there is the serial port controller, ...15:10
viricat what other the PCI bus...15:10
viricevery board is different15:10
kyakok then15:10
kyakwill do malta15:10
viricqemu emulates other boards15:10
viricI did not see a big reason not to use the default15:10
viricqemu -M ?15:11
viricqemu-system-mipsel -M ?   I mean15:11
kyaki understand now...15:12
kyakthnaks :)15:12
kyakok, selected block2mtd, ubifs and ubi to be built-in kernel15:14
kyakeverything else is default15:15
kyaklet's see how it works15:15
kyakviric: what is called as "rootfs.img" in your e-mail, is actually the ubi image?15:16
viricresulting from ubinize15:16
kyakso i assume i already got it from openwrt.. only need to pad it with 0xff15:17
kyakviric: /dev/hda,52428815:26
kyakhow you got 524288?15:26
kyakso, despite of the fact that your ubifs is only 50 Mb, block2mtd emulates a 512 Mb drive?15:28
viricthat is the PEB size15:28
virichow much you padded of 0xff?15:28
kyakthe image is ~186 Mb15:29
kyaki padded up to 20015:29
viricit does not matter the size of the image15:29
viricmatters the volume size told to ubinize15:30
viric(I think)15:30
viricyou need to pad more than the volume size told to ubinize.15:30
viriclet me look for openwrt...15:30
kyakah ok15:30
viricah it has no volume size15:31
kyakthe volume size ot ubinize is 256 Mb15:31
kyakyea, the size of rootfs15:31
kyakor no?15:31
virictarget/linux/xburst/image/ubinize.cfg does not say so15:31
kyakhm .15:31
kyaki guess it's dynamic then?15:32
viricyes yes15:32
viricit may work for 20015:32
kyaksee, i got thies error15:32
kyakUBIFS error (pid 1): ubifs_get_sb: cannot open "ubi0:rootfs", error -1915:32
kyakVFS: Cannot open root device "ubi0:rootfs" or unknown-block(0,0)15:32
viricthat's ubifs15:32
viricdoes ubi work for you?15:32
kyakwhat do you mean?15:32
viric(look above)15:32
viricabove in the kernel log15:32
kyakUBI error: validate_ec_hdr: bad VID header offset 4096, expected 6415:33
kyaki guess no15:33
viricthe openwrt build with -m 4096 instead of -m 115:33
kyaki remember you got this error :)15:33
kyakor similar15:33
viric(the minimum subpage)15:33
viricblock2mtd has a fixed minimum subpage of 115:33
viricwhile the nanonote has a minimum subpage of 409615:34
viricI don't know how to make the block2mtd have a minimum subpage of 4096, or how to tell ubi to use a minimum subpage of 409615:34
kyakyeah right.. below 4096 is uboot and kernel i guess15:34
viricno no15:34
viricit's about being able of getting blocks of 4096 bytes15:35
viric(I think)15:35
kyaksince i can make openwrt generate rootfs.tar.gz,  i think i will ubinize it by myself15:36
kyakwith appropriate options15:36
kyakhow do you think?15:36
viricyou only need to change -m 4096 to -m 1   in mkfs.ubifs and ubinize15:36
viric(I think)15:36
kyaknah, this will break my regular build :)15:36
kyak(i think)15:36
kyaki suppose also that block2mtd  has some options to be able to support blocks of 4096 bytes?15:37
viricwell, block2mtd supports subpages of 1 byte15:37
viricI don't know how to make it support only 4096byte subpages15:38
kyakcan you share the script or config you used to ubinize?15:39
virickyak, try in the cmdline: ubi.mtd=0,4096  ?15:39
kyakUBI error: validate_ec_hdr: bad data offset 8192, expected 416015:40
viricah. bad.15:40
viricI wish all this were easier :)15:41
virickyak: you need to pass -m 1 to mkfs.ubifs too15:41
viricnot only to ubinie15:41
virickyak: http://nixos.pastebin.ca/199817915:42
kyakwhat's $deviceTable?15:44
viricwell, I don't know if openwrt uses it15:44
kyakand ensureDir?15:44
viricforget those around :)15:44
viricI just copied what I have.15:45
kyakok :) thnaks15:45
viricnix related things15:45
viricThe devicetable is there because I don't want to become root to prepare the image15:45
kyaki don't really understand how the files will get inside the image?15:46
viricwhat files?15:46
viricI have all the files in the subdirectory 'root'15:46
viric-r root points to it15:46
kyakah good15:46
viricHow would I reflash my nanonote with that ubifs?15:47
viricbooting in that special mode, right?15:47
viricM-power ?15:49
kyakand then using usbboot15:49
kyakm-power is to boot from sd card15:49
viricI'll try if my ubifs image works on the nn15:50
viricI think I have nothing of value there15:52
viricwhy usbboot wants to be run by root?15:56
kyakmaybe it needs some low-level access to your usb port, dunno15:57
viricmy user has access to the usb device. I think it simply uses libusb... it should not need root15:58
viricmaybe I have the permissions wrogn15:59
virichm no15:59
virickyak: hummm the '-e' on nanonote and block2mtd also look like being different16:02
kyakviric: i think the best is to generate ubifs manually from staging_dir/target-mipsel_uClibc-
kyakusing your mkfs.ubifs and ubinize settings16:06
kyakwill try do it tommorrow.. time to get some sleep16:06
kyakviric: thanks for your help!16:07
viricwhat is your timezone?16:07
viricah ok.16:07
viricgood night!16:07
viricargh. It looks like 'nerase 16 4080 0 0" erased the uboot too.16:22
viricI can't understand why.16:22
wpwrakviric: meanwhile, deep in the mind of your computer: "hmm, he made me erase u-boot in a pretty convoluted way. I can't understand why."16:39
viricconvoluted? :)16:40
viricaren't 16 blocks the offset for the rootfs?16:40
wpwraknaw, i don't know what happened :)16:40
viricI imagine that for ubi, it's always the case that the flash has to be fully erased?16:41
wpwrakdunno. never touched UBI. way too complex for my taste.16:43
viricdon't you use ubi?16:46
viricand, what has that to do with ubi? :)16:46
viricit's uboot that got erased :)16:46
wpwrak(use ubi) well yes, but i don't understand its internals much like i don't really understand the quantum physics going on in transistors, yet i use them ;-)16:49
wpwrak(what to do with ubi) well, you asked about it ;-)16:49
viricah I already had forgot my question about ubi :D16:49
wpwrakyour computer and you *are* similar ! you forget about the question, your computer forgot about u-boot. have you considered the possibility that you might be an AI running in a different virtual machine on the same hardware ? a hardware with memory problems. ;-)16:51
qi-bot[commit] Werner Almesberger: qpkg/qpkg.c: tighten detection of inconsistent use of Filename: http://qi-hw.com/p/wernermisc/892238319:21
qi-bot[commit] Werner Almesberger: qpkg: rearrange things in prereq.c to free the stack paradigm for other uses http://qi-hw.com/p/wernermisc/d0a1a9e19:21
qi-bot[commit] Werner Almesberger: qpkg/prereq.c (prereq): removed confused false error avoidance http://qi-hw.com/p/wernermisc/5a0444f19:21
qi-bot[commit] Werner Almesberger: qpkg/prereq.c (resolve): cleaned up the debugging output mechanism http://qi-hw.com/p/wernermisc/9272f5019:21
qi-bot[commit] Werner Almesberger: qpkg: store dependencies and conflicts in file order http://qi-hw.com/p/wernermisc/6c66fa619:21
qi-bot[commit] Werner Almesberger: qpkg: change prerequisite resolution order to be fully depth-first http://qi-hw.com/p/wernermisc/022905119:21
qi-bot[commit] Werner Almesberger: qpkg: added detection of cyclic dependencies http://qi-hw.com/p/wernermisc/0bc4b6019:21
qi-bot[commit] Werner Almesberger: qpkg/prereq.c (print_debug, resolve): moved debug output to separate function http://qi-hw.com/p/wernermisc/e1404bb19:21
qi-bot[commit] Werner Almesberger: qpkg: option -d to enable debugging at run time, and some cleanup http://qi-hw.com/p/wernermisc/a0c0e8519:21
qi-bot[commit] Werner Almesberger: qpkg: moved comp_versions from prereq.c to id.c http://qi-hw.com/p/wernermisc/dd8bb6e19:21
qi-bot[commit] Yauhen Kharuzhy: Add libmpdclient package http://qi-hw.com/p/openwrt-packages/5e68f8921:01
qi-bot[commit] Yauhen Kharuzhy: Fix ncmpc compilation http://qi-hw.com/p/openwrt-packages/70138e921:01
kristianpauldamn why all people uses the same Cypress MCU :/21:24
kristianpaulsuch honor copying usrp closed hw ideas..21:24
wpwrakhonor or horror ?21:25
kristianpaulgood question :)21:26
kristianpaullast !21:26
wpwrakthought so :)21:31
kristianpaulhmm A-GPS is linked with GSM base stations?.21:46
kristianpaulhow google amps can enable Maps in no gps phones... is it using gsm atena and dsp stuff to process gps data ?..21:47
kristianpaulquestions for future devel.21:47
wpwrakafaik, agps downloads additional data (by whatever means) to improve startup time21:48
kristianpaulahh download21:49
wpwrakyou can use wifi and/or gsm to fake gps, yes21:49
wpwrakgprs, wifi, floppies, ... ;-)21:49
qi-bot[commit] Xiangfu Liu: add msmtp-queue package. when it goto upstream, we will delete it http://qi-hw.com/p/openwrt-packages/f9cc7d521:59
Action: xiangfu set the image default wallpaper to open.png http://en.qi-hardware.com/wiki/File:Wallpaper_Nanonote_open.png22:32
FrankBluesNice wallpaper!22:36
qi-bot[commit] Xiangfu Liu: set the default wallpaper to open.png http://qi-hw.com/p/gmenu2x/6773ad722:43
wolfspraulit should say free, not open :-)22:44
--- Mon Nov 22 201000:00

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