#qi-hardware IRC log for Wednesday, 2011-08-03

qi-bot[commit] Xiangfu Liu: gmenu2x update (master) http://qi-hw.com/p/openwrt-packages/f8c539101:49
qi-bot[commit] Xiangfu Liu: nanonote-files update build script file (master) http://qi-hw.com/p/openwrt-packages/a6f250f01:55
kristianpaulxiangfu: why you use branches isntead of tags for qi-openwrt releases?01:56
xiangfufor openwrt-package branch is better. since feeds.conf support branch. not tag.01:59
kristianpauloh, good to know01:59
xiangfufor openwrt-xburst.my plan is using tag. 02:00
xiangfuyou can see tag: release_2010-12-14 and release_2011-02-2302:00
xiangfusince we rebase a lot. I think branch is better then tag for us. so for "release_2011-05-28" I using branch again. 02:02
kristianpaul(rebase) ;)02:02
xiangfumaybe I just delete those two tag, using branch as always. 02:02
xiangfuno tags anymore. all branch.02:02
xiangfuthat do you think? tag and branch is mess, only one is better. 02:02
kristianpaulis not better all branch and tags in branch?02:03
kristianpauloh, dint knew it about the mess02:03
kristianpaulwhat linux uses for releases?02:03
kristianpaultarball still? :)02:03
xiangfutag in upstream.02:03
kristianpaulwell.. i really cant advice you, actually i made question in first place as i wanted to survery git users about branch and tags for releases02:05
kristianpaulpersonallu i like tags in the master branchs02:06
kristianpauland let the other branchs for testings/experimenting porpuses02:06
kristianpaulthere is that stash thing but i dont get used to it yet..02:06
xiangfuI am testing. seems the after rebase, tag have some problem. ask in #git for make sure. if tag have problem when rebase, then we have to using 'branch' :)02:15
kristianpaulcan you avoid rebase?02:18
kristianpaulor do it not in master then later merge?02:18
xiangfuI try to keep our commit on top of openwrt's commit. the only way is rebase :(02:31
ignatius_Anyone know how to use the entire NAND's space for userland? Every time I install a .UBI I only get around 500MiB.02:34
kristianpaulignatius_: http://en.qi-hardware.com/wiki/Ubifs_Startup02:36
kristianpaulgood and bad news03:13
kristianpaulfirst the accumlator interrupt signal was OK all this time03:14
kristianpaulbut, and just until i start playing a bit with vertical scale in the scope i discovr this http://downloads.qi-hardware.com/people/kristianpaul/wtf%20accum%20int.png03:16
kristianpaulthat signal should look like this http://downloads.qi-hardware.com/people/kristianpaul/good%20accum%20int.png03:18
kristianpauldifference? well a different fpga pin03:18
kristianpaulthats the bad part... i hope i dint damaged without noticing..03:18
xiangfuI just try "openwrt-xburst.full_system-07272011-1848" images. 03:47
xiangfupython gets better. but still "Segmentation fault", pyclock is working. but 'python' get "segmentation fault"03:47
wolfspraulkristianpaul: you there?03:51
wolfspraulI am wondering whether we should go back to a short usb cable for m1.03:51
kristianpaulhe ;)03:52
wolfspraulI guess going to a long one (I believe that came from you) caused new problems. And I don't want to spend energy on fixing those problems, instead why not just have a short one?03:52
wolfspraulyou complained back then, why?03:52
kristianpaulwell, my current cable is a 80cm long from a canon camera and works ok03:52
wolfspraulif the cable is user in a developer setting (because of the upward pointing USB it's meant to go to jtag), then most likely the m1 will stand right next to a notebook, no?03:52
kristianpaulyes i complain time ago about short cables03:52
wolfspraulok but that one is a dev only situation03:53
kristianpaulbut i never tought this could happen, sorry :(03:53
wolfspraulthe m1 is open03:53
wolfspraulno, all fine03:53
wolfspraulwe must never be afraid to improve things03:53
wolfspraulthere's always a price to pay, but I know this and I'm still passionate about aggressively improving03:53
wolfspraulevery pioneer will at some point realize there are easier ways to live your life :-)03:53
wolfspraulisn't this cable used in a dev only setting?03:54
wolfspraulm1 is open03:54
wolfspraulit must be near03:54
kristianpaulergg, my cable 123cms from canon camera03:54
kristianpaulyes dev03:54
wolfspraulI cannot see why there is any problem if the cable is 80cm, even less03:54
wolfspraulif the cable is used for m1 in an event/performance setting, I can understand03:54
kristianpauland no problem on my side of course, runnon debian stable, what else i can tel..03:54
wolfspraulwe have extra long cables for the power supplies now, for exmaple :-) (2.5m instead of the normal 1.8m)03:54
kristianpauljtag pod with fix03:54
wolfspraulso you say 80cm for the usb jtag cable is also fine?03:55
kristianpaulwell, i could say that but is not realiable as i never reflashed 90 boards with mine...03:56
kristianpaulyeah. well usually next to netbook.03:57
kristianpaulwell, all people havbe netbooks now, may be i need get one in my home and replace the dekstop computer ;)03:57
kristianpaulbut  YES. cable is only for dev settings i gree03:58
kristianpauland if short works, okay...03:59
kristianpauli think xiangfu have a nice shorter cable, now i realize why :-)03:59
wolfspraulI'm not trying to talk you into liking it, I am trying to understand the specific setting you have in mind.04:00
wolfspraulin your case, the m1 is on your desk, but the cable runs down to a desktop computer sitting on the floor?04:00
wolfsprauland what is the cable length you need to reach down there?04:01
wolfspraul80cm will not be enough, right?04:01
kristianpaulwell. in my particular case a little tight but you work04:02
kristianpaulhe, i hope my feedback helped... well... at least i'll keep mind changing cable lengts is not  good always..04:06
wolfspraulwell, anything can make a difference, in hindsight you know.04:06
wolfspraulbut that cannot stop us from making changes, otherwise we may as well close shop.04:06
wolfspraulso all fine04:06
wolfspraulwe learn04:06
kristianpaulafortunty usb cables are cheap in you side i guess :)04:08
kristianpaulgn8 :)04:42
qi-botThe build has FAILED, see log here: http://fidelio.qi-hardware.com/~xiangfu/compile-log/openwrt-xburst.full_system-08022011-0429/07:49
kodeinoh dear08:07
xiangfuthe recently openwrt updata break the 'at-3.1.12' package.08:42
LunaVoraxHello everyone !11:16
wolfspraulLunaVorax: hi11:23
Aylawpwrak: hello?11:40
LunaVoraxwolfspraul, how have you been ?14:04
wolfspraulLunaVorax: busy!14:09
wolfspraulworking on upcoming Milkymist One14:09
wolfspraulour video synthesizer14:09
wolfspraulLunaVorax: wanna buy one? :-) 499 USD...14:09
wolfsprauland how have you been? what's brewing?14:10
rix1234has anyone done with usb-redirection for citrix receiver14:23
LunaVoraxwolfspraul, passing semester 2 at college, still learning C and SDL, working on some project, building a NAS RAID-5, lot of stuff :)15:50
LunaVoraxwolfspraul, also tried to learn asm x86 but lost motivation15:50
Aylalosinggeneration, what are you doing here? :)15:54
losinggenerationidling :P15:55
Aylayou bought a nanonote?15:56
Ayladone with your didj?15:57
losinggenerationseems like a logical place to idle since I have an openmoko neo freerunner and like open hardware stuff15:57
losinggenerationoh, not done with it, just put on the back burner for a bit, life is a bit hecktic right now, and I've been trying to work on some sms code lately15:57
losinggeneration*hectic* even15:57
Aylasms as in sega master system?15:58
losinggenerationit is a bit ambiguous when I'm in non-sega irc channels :P15:58
Aylakeep me informed, we need a SMS emu :p15:59
losinggenerationwell, I ported CrabEmu to the Didj which meant writing an SDL backend15:59
Aylathe Didj is 320x240?16:00
losinggenerationoff the top of my head, I think so16:00
Aylathere's no ASM in it?16:01
losinggenerationpure C16:01
Aylaheh, interesting16:02
losinggenerationit needs a few more minor optimization, but it runs at pretty much full speed16:03
Aylatell me when the code is mature enough, so that I can port it to the dingoo A320...16:03
losinggenerationit's actually not in terrible shape IMO, but another set of eyes on it may be useful :)16:04
Aylawe'll release a new kernel for it in a few, a new emulator would make some advertisement16:04
losinggenerationoh, it does depend on Lua (for configuration) also, but that shouldn't be too hard to make for pretty much any system16:06
Aylait's one case to check on buildroot, and we have Lua on the system16:07
Aylayou could put KOS lib on the dcload disc16:16
Aylathe elf to send on the cable would be MUCH smaller16:16
losinggenerationYes, it would indeed be smaller. I'd also like to get KOS to a point where it's modular, so only, for instance, the base OS, and filesystem stuff loads up instead absolutely of everything16:24
Aylathat's for KOS 3.016:25
losinggenerationyeah :)16:26
losinggenerationmaybe for ten years from now :P16:26
losinggenerationI've also been thinking it may be useful for it to support more devices/consoles... but the DC still has a pretty loyal (albeit small) following 16:28
Aylait's too close to the hardware to be portable16:29
losinggenerationwell, it used to have initial support for ps2 and gba, but those were never finished/matured16:29
AylaI wouldn't mind seeing it on x36016:30
losinggenerationinitial apps wouldn't be portable from one device to another (which would be annoying) unless we put a generic hardware interface in place16:30
losinggenerationoh, all the big picture projects I'd love to have enough time to work on :/16:31
Aylaheh, I know that16:34
AylaI have one big picture project called AICA/OS& ;)16:34
Aylawhich has been in pause for a long time16:36
losinggenerationdid you end up ditching newlib, I forget16:36
AylaI have some code on github for it, but I fear I'll have to start it again, and implement the RPC on the syscalls of newlib16:37
Aylathe goal is to be able to open say "/cd/file.txt" directly from the ARM, and read/write it, with nothing more than fopen, fread and fwrite16:40
losinggenerationthat would be pretty cool, also, fclose is overrated (or underrated for not being mentioned :P)16:44
Aylayou'll have fclose() too, indeed16:45
Jay7LunaVorax: raid5 may be one-way-ticket for your data :)17:36
LunaVoraxJay7, you sure ? I though it was safe17:37
Jay7LunaVorax: it have one problem - write hole17:37
Jay7check wikipedia about RAID517:38
LunaVoraxI did17:38
Jay7failed driver increasing load of next drive in array17:38
Jay7at least install spare drive and run checkum re-checkin regularly17:39
LunaVoraxJay7, which RAID would you use ?17:40
Jay7mostly RAID1, but have 2 hardware RAID6 too17:41
Jay7now thinking about ZFS with RAID-Z..17:41
Jay7it's RAID5 but w/o write hole because of ZFS's CoW algo17:41
Jay7I dislike HW raid's.. it's hard to recover data from it always17:42
Jay7if card is dead you should find card by same vendor (and may be even from same series)..17:42
Jay7after ~5-10 years it's very hard to do :)17:43
Jay7LunaVorax: other option is RAID10 (or 1+0)17:44
Jay7but you will have less space when compared to RAID5..17:45
Jay7but disks are cheap now :)17:45
LunaVoraxJay7, we don't have the same conception of cheap then17:47
LunaVoraxI paid 333¬ for 4*2Tb17:48
Jay7it's better than 3333 euro :)17:48
LunaVoraxAnd, yes, I need the 8tb :/17:48
LunaVoraxI'm not that ready to pay 333 more euros only to make a RAID 1017:49
Jay7well, use RAID5 but buy one spare disk 17:49
Jay7just to lower possibility of double fault17:49
Jay7or change drives every 2 years :)17:49
LunaVoraxJay7, what software are you using for your raid ?17:50
Jay7LunaVorax: mdadm on Linux, gmirror or zfs on FreeBSD17:50
LunaVoraxI was thiniking about using FreeNAS since I'm a noob17:51
Jay7FreeNAS is FreeBSD-based17:51
LunaVoraxI know17:51
Jay7there you may use RAID5 because it is RAID-Z17:51
LunaVoraxYou mean RAID5 using ZFS formatted disks ?17:51
Jay7but you should buy at least 4Gb of RAM for zfs to run smoothly :)17:52
LunaVoraxOh damn17:52
Jay7better to have 8Gb (if you will enable deduplication with next FreeNAS release)17:52
LunaVoraxNo ZFS then17:52
LunaVorax4gb for a filesystem ? that's insane17:52
Jay7not for FS, but for NAS17:53
LunaVoraxyou lost me a bit17:53
Jay7zfs may work with 1Gb 17:53
Jay7but you will get more speed with 4Gb 17:53
Jay7because you may enable prefetching17:53
Jay7or use OpenFiler e.g.17:54
Jay7it's linux-based (lvm + ext3 iirc)17:54
LunaVoraxThat's getting even more complicated than I though17:54
Jay7or use freenas but with UFS17:55
Jay7but I'm unsure it will create raid5 then.. you should check this before17:55
Jay7LunaVorax: nothing really complicated... it depends on your NAS load17:56
Jay7for low load you may use almost everything17:56
Jay7for high load I'll prefer to have ZFS with 8Gb of RAM :)17:56
LunaVoraxJay7, It's just a little NAS for me on centralise all of my data mand make it safe17:59
LunaVoraxAnd I'm the only one using it18:00
LunaVorax-on + to18:00
Jay7LunaVorax: try OpenFiler then18:05
Jay7it should provide all you want :)18:05
Jay7and iirc, it does not require separate storage for itself (like FreeNAS does)18:06
LunaVoraxhum key18:06
Jay7but I'm not sure18:06
kristianpaultrue, raid is a mess when you want ot recover lost data18:41
kristianpaulwher i work, at month or so, there are such us cases18:42
kristianpauland hardware lockerd to vendor most of the time..18:42
kristianpaulNAS is good way, at least if you can sync tow of this boxes (hopefullt remotelly separated)18:43
Jay7btw, nice place for open HW - open hw raid cards :)18:44
Jay7with open metadata format and cheap batteries :)18:45
Jay7and sure with open drivers and tools :)18:46
kristianpaulyeah... when time allow :)18:48
kristianpaulstorage indeed is full of lock, just see those cheap usb memory sticks now days..18:49
kristianpaulor even worst external hard drives, with usb support, that uses AES, but no body knows the key aparently..18:50
kristianpauloh well :)18:51
--- Thu Aug 4 201100:00

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