**** BEGIN LOGGING AT Fri Oct 19 03:00:01 2012 Oct 19 04:04:44 hi, all, my n900 is partly broken - autolock is not working Oct 19 04:05:30 the keyboard is always open; so I cannot use on-screen virtual keyboard Oct 19 05:54:06 hmmmm, obviously archl been in a hurry Oct 19 08:15:23 think i need to make my own scv layout so i can move almost-useless chars there Oct 19 08:15:48 to make room for a 4-arrow finnish layout with most of the unix-related chars with fn-something or fn-shift-something Oct 19 08:16:40 jaska: ~vi-kbd Oct 19 08:16:52 goddamit Oct 19 08:16:53 jaska: ~vi-kb Oct 19 08:17:03 jaska: ~vi_kbd Oct 19 08:17:18 ~listkeys vi Oct 19 08:17:20 Factoid search of 'vi' by key (16 of 1074): "nvidia's cg compiler ;; airport vid ;; biffview ;; dont be evil ;; favicons ;; ftp logview ;; kvncviewer ;; linux device list, ;; linux virus ;; name service browser ;; nvidia ;; old_viper ;; oldtest hivites ;; openview ;; pogovina ;; problem with the victim doing the decision. Oct 19 08:17:44 ~listkeys vi- Oct 19 08:17:45 Factoid search of 'vi-' by key (3): vi-kbd ;; vi-is-better ;; poor man's vi-mode. Oct 19 08:17:48 http://www.upir.cz/maemo/keyboards/ i think has the compiler.. if wiki is to believe Oct 19 08:18:09 jaska: oh wait, you are talking about virtual keybopards? Oct 19 08:18:22 i mean the special symbol panel you get when you press ctrl-sym Oct 19 08:18:39 so i can move useless stuff like ε, eur sign, gbp sign etc there Oct 19 08:18:44 jaska: That panel sucks. Oct 19 08:18:55 and reuse those for meaningful stuff Oct 19 08:19:00 Does the swedish board use the 4th row? Oct 19 08:19:09 4th level eve, Oct 19 08:19:12 4th level even. Oct 19 08:19:18 nah Oct 19 08:19:29 but some keys on the hardware keyboard dont work with shift-fn Oct 19 08:19:36 like z, x, maybe j and k Oct 19 08:19:43 jaska: I know. Oct 19 08:19:58 otherwise it would have been easy :) Oct 19 08:20:01 But that still leaves like 25 other free keys to use. Oct 19 08:20:51 Hi everybody, hopefully I'm going to have some time to get Maemo hacking again soon, I've been using up my free time trying to get my skiing back up to scratch to pass an instructor test (on Sunday). That along with trying to learn Dutch, and Linux admin work... Oct 19 08:21:07 zeq1: Hey buddy! Oct 19 08:21:20 You earned your R&R, enjoy it! Oct 19 08:21:27 hi vi_ :) Oct 19 08:21:54 jaska: you know about editing the hardware keymap right? Oct 19 08:22:04 you mean xkb? Oct 19 08:22:10 yes. Oct 19 08:22:12 yeah i have a customized one Oct 19 08:22:22 cool. Oct 19 08:22:38 jaska: And STILL not enough space?! Oct 19 08:22:41 yeah :| Oct 19 08:22:50 WOW Oct 19 08:22:51 need more room to map out 4-arrow keypad Oct 19 08:23:17 I have almost every key from a normal english keyboard. Oct 19 08:23:21 which means moving atleast 2 buttons of mappings somewhere else.. and fixing the stuff i mapped behind z and x Oct 19 08:23:30 How many characters is in the swedish aplphabet? Oct 19 08:23:39 28? Oct 19 08:23:50 only 2 more than english! Oct 19 08:23:55 wrong Oct 19 08:23:56 ah, swedish is 29 Oct 19 08:23:57 three Oct 19 08:23:59 Γ₯Àâ Oct 19 08:24:01 finnish is 28 Oct 19 08:24:01 wtf are you using these keys on? Oct 19 08:24:46 but ε is just used in names so im not going to keep it on a hw key Oct 19 08:25:33 You can use the Z+X bug slightly to your advantage. Oct 19 08:25:47 if you like backspacing stuff you typed :) Oct 19 08:25:51 If you map z=[ and x=] Oct 19 08:25:55 heh Oct 19 08:26:06 when you use 4th level z you will get [] Oct 19 08:26:13 when do you ever use a [ Oct 19 08:26:18 and not a ] Oct 19 08:26:19 ? Oct 19 08:26:24 not often Oct 19 08:26:31 so save time and bring both up at once! Oct 19 08:26:40 j and k are another broken pair Oct 19 08:26:47 also works with () and {} and <> Oct 19 08:27:03 jaska: Yup, I have () on them. Oct 19 08:27:09 do we know why they are broken? Oct 19 08:27:13 yes Oct 19 08:27:24 hw issue? Oct 19 08:27:48 it is to do with el-cheapo, poorly thought out hardware design. Oct 19 08:28:08 joerg wrote a good explanation of it on the wiki some where. Oct 19 08:28:09 oh how i wish it had a 4-row pad Oct 19 08:28:30 i dont know how bad the n950 kbd is Oct 19 08:28:30 do you mean 4 arrow keys. Oct 19 08:28:34 do you mean 4 arrow keys? Oct 19 08:28:41 4 rows of keys and 4-arrows :) Oct 19 08:28:54 ah yes. Oct 19 08:29:00 also the n950 kb is shit. Oct 19 08:29:11 yeah ive heard rumors of such Oct 19 08:29:29 not going to pay 1k for a beta phone :) Oct 19 08:30:10 jaska: more like alpha. Oct 19 08:30:19 and I do not mean alpha male. Oct 19 08:30:58 unless doing thibgs the hard way is alpha male, in which case I do. Oct 19 08:31:20 jaska: so swedish KB has the funny 2 arrow design? Oct 19 08:31:25 yeah Oct 19 08:31:31 fi/se layout is the same Oct 19 08:31:43 how did they come up with that? Oct 19 08:31:50 why not just change rx-51 to use the 4 arrow design instead of the 2? Oct 19 08:32:00 that would displace some needed letters Oct 19 08:32:03 zeq1: you have to press blue+left to go up. Oct 19 08:32:11 i am trying to find room to make it 4-arrow :) Oct 19 08:32:13 so intuitive Oct 19 08:32:29 also having to press fn-left to go up makes it hard to map page-up Oct 19 08:32:42 because if you map it to fn-shift it will not show as pgup but shift-pgup Oct 19 08:32:46 jaska: I literally do not understand how you can have run out of letters when your alphabet has only 3 more characters than mine. Oct 19 08:33:01 vi: because i can! Oct 19 08:33:19 also various f-keys behind shift-fn-toprow Oct 19 08:33:26 I have gone as far as mapping stuff like |, ~, %, ^, and ` because there is so much space. Oct 19 08:33:44 escape is fn-backspace Oct 19 08:33:53 toggle fullscreen is shift-fn-backspace Oct 19 08:33:53 jaska: same here. Oct 19 08:34:07 jaska: fn-space. Oct 19 08:34:12 ~vi-kbd Oct 19 08:34:12 rumour has it, vi-kbd is http://wiki.maemo.org/Remapping_keyboard/user_vis_awesome_kbdmapping Oct 19 08:34:15 fn-space is @ Oct 19 08:34:39 jaska: then how do you type in a space when you have fn locked??? Oct 19 08:34:49 i never lock fn? Oct 19 08:35:05 also i use ctrl-i for tab :) Oct 19 08:35:12 You mean you let go of fn to make a space when you are typing in all numbers. Oct 19 08:35:22 yeah Oct 19 08:35:26 or a dot Oct 19 08:35:27 jaska: illogical and filthy. Oct 19 08:35:35 so what Oct 19 08:35:38 ctrl-i is for ctrl-i Oct 19 08:35:55 well, ctrl-i works in a terminal for tab so its ok with me Oct 19 08:35:55 jaska: I am just making jokes. Oct 19 08:36:35 I have tab on fn-enter Oct 19 08:36:58 . on fn-fullstop Oct 19 08:37:19 this lets you type in IP adresses without letting go of fn. Oct 19 08:37:26 yeah :( Oct 19 08:39:36 vi_: hi Oct 19 08:39:49 hi freemangordon Oct 19 08:39:55 LOL Oct 19 08:40:13 zeq1: are you alive? :P Oct 19 08:40:19 just about :) Oct 19 08:40:22 ot it is some bot here Oct 19 08:40:25 *or Oct 19 08:40:31 definitely me... I think Oct 19 08:40:58 where you've been, long time no see Oct 19 08:41:19 see above Oct 19 08:41:31 I've just been really busy Oct 19 08:41:53 not had time to be on IRC at all Oct 19 08:42:47 hmm, that is good news I hope Oct 19 08:43:00 what's been going on here? anything new? Oct 19 08:43:34 well, depends. Oct 19 08:43:51 mysterious Oct 19 08:44:02 lets say nothing new Oct 19 08:44:50 the usual disputes here and there Oct 19 08:44:57 zeq1: so now you have free time? Oct 19 08:45:24 zeq1: BTW i found a bug in gcc4.7.2. Oct 19 08:45:39 okay Oct 19 08:45:41 tried 2012.09, the same Oct 19 08:46:01 is it in the known bug list for linaro? Oct 19 08:46:20 couldn't find it Oct 19 08:46:34 gcc bugzilla? Oct 19 08:47:32 what is the bug? Oct 19 08:47:33 didn;t search. the problem is that gcc issues FPU 64(or 128) bit VFP load instruction on non-aligned data Oct 19 08:47:53 and fails or is slow? Oct 19 08:48:00 kernel traps :D Oct 19 08:48:03 it requires alignment Oct 19 08:48:06 yeah Oct 19 08:48:13 does the kernel handle it? Oct 19 08:48:53 I'll have a look in the gcc bugzilla Oct 19 08:49:08 naah, don;t waste your time for that Oct 19 08:49:17 the trouble is gcc is, umm..., rather complex to debug... Oct 19 08:49:58 zeq1: and it appears only on -O3 Oct 19 08:50:09 with -mfpu=neon Oct 19 08:50:18 i.e. tree-vectorizer Oct 19 08:50:18 that should make it easier to find Oct 19 08:50:41 some constraint not applied I guess Oct 19 08:51:22 well, if you are full of energy wating to be spend on maemo, you'd better make glibc auto-detect kernel pselect() suport :P Oct 19 08:51:44 that's waiting on me still? :O Oct 19 08:51:51 yeah :( Oct 19 08:52:39 seems I am the only one beside you who dare to touch glibc, but I was busy doing other stuff Oct 19 08:53:10 at least noone volunteered :) Oct 19 08:53:14 Okay, I'll put it to the top of my Maemo TODO Oct 19 08:53:21 good Oct 19 08:53:49 zeq1: hmm, something that happened: romaxa joned maemo-ssu :D Oct 19 08:53:55 *joned Oct 19 08:53:57 :) Oct 19 08:53:59 the fuck Oct 19 08:54:04 JOINED Oct 19 08:54:20 vi_: here? Oct 19 08:54:23 that's got to be good! Oct 19 08:54:30 it is Oct 19 08:55:24 unfortunately he no longer works for Nokia Oct 19 08:55:27 romaxa is a top drawer hacker IMO Oct 19 08:55:37 freemangordon: You summoned. Oct 19 08:56:04 zeq1: deffinitely Oct 19 08:56:11 working for Nokia nowadays probably would be too comfortable Oct 19 08:56:13 vi_: any results for mplayer? Oct 19 08:56:19 wouldn't* Oct 19 08:56:29 vi_: I am really curious Oct 19 08:56:50 zeq1: well, depends, if you are good with c#, why not :D Oct 19 08:57:22 zeq1: BTW did you receive my mail? Oct 19 08:57:51 freemangordon: Not yet, although I have found there are some weird iisues wrt to the keyboard. Oct 19 08:57:52 when did you send it? I had a prolonger server outage Oct 19 08:58:19 when playing something with mplayer there are keys for skip forward/back fullscreen etc. Oct 19 08:58:24 3-4 weeks ago. I asked for your ng-crosstool scripts Oct 19 08:58:36 vi_: yes Oct 19 08:58:44 these sometimes work, sometimes not. Oct 19 08:59:14 I had a WD Velociraptor start throwing IO errors in my array Oct 19 08:59:19 sometimes if they are working, then you use volume keys to alter volume they no longer work after that. Oct 19 08:59:28 vi_: how is the performance? Oct 19 08:59:46 (turned out to be a firmware bug!) Oct 19 08:59:49 I suspect it is something to do with window focusing, although I am not sure. Oct 19 08:59:51 I mean does it worth the effort? Oct 19 08:59:58 zeq1: bad Oct 19 09:00:31 vi_: btw there is still configure script --enable-maemo :D Oct 19 09:00:34 freemangordon: I think so, I 'feel' so. But I do not have real numerical proof. Oct 19 09:00:47 'feel' how? Oct 19 09:00:56 naah, forget about numerical Oct 19 09:01:06 we don;t have such for -thumb too Oct 19 09:01:19 freemangordon: I have to confess, I never actually scripted the toolchain build, I have a chroot I use specifically for it, and just rely on bash history(!) Oct 19 09:01:23 I tried to play an mkv file of 'zombie flesh eaters 2'. Oct 19 09:01:40 Mplayer cannot play it, it is too much. Oct 19 09:01:54 so the video progressivly lags the audio. Oct 19 09:02:00 yeah, DSP is still the winner Oct 19 09:02:07 unless you use frameskip. Oct 19 09:02:18 zeq1: well, I finished doing the same Oct 19 09:02:24 I got the impression it was skipping less frames with the new one than with the old one. Oct 19 09:02:38 i.e. less jumpiness. Oct 19 09:02:45 zeq1: though this time I put command lines in a file and saved them :D Oct 19 09:02:51 vi_: yeah, I see Oct 19 09:03:11 In addition to that it is a lot more stable. Oct 19 09:03:19 stable? Oct 19 09:03:24 hmm Oct 19 09:03:32 Old mplayer would just crap out on audio streams. Oct 19 09:03:37 the one in extras is unstable? Oct 19 09:03:40 randomly Oct 19 09:03:44 aah, i see Oct 19 09:03:52 and sit there doing nothing. Oct 19 09:04:02 so you had to ctrl+c to quit and start again. Oct 19 09:04:09 It bugged the piss out of me. Oct 19 09:04:19 this one does not. Oct 19 09:04:21 vi_: any chance to tell kmplayer what to pass to mplayer? Oct 19 09:04:51 I do not know. I have never actually used kmplayer beyond seeing that it worked. Oct 19 09:04:57 :D Oct 19 09:05:00 I do not even know what it is written in. Oct 19 09:05:11 it could be tcl/tk Oct 19 09:05:22 with 'wish' gui extension for what I know. Oct 19 09:05:30 vi_: the more important queston: can you define how backing swap is supposed to behave? Oct 19 09:05:31 (although I do know it is not that!) Oct 19 09:05:40 freemangordon: not slow as shit. Oct 19 09:06:01 naah, try harder Oct 19 09:06:06 correction, frozen shit in an igloo. Oct 19 09:06:16 ramzswap works. Oct 19 09:06:26 if it had backing swap it would be great. Oct 19 09:06:42 But as soon as you enable BS it just stalls. Oct 19 09:06:42 I mean, if I am to be on it, I want to do it in a proper way Oct 19 09:06:55 especially if it correctly handled the flash erase block alignment Oct 19 09:07:20 zeq1: that one is easy Oct 19 09:07:53 i can perform runtime test in module load to find block size and partition misalignement Oct 19 09:08:13 freemangordon: but it isn't done now afaik Oct 19 09:08:20 the problem comes after that, i.e. when to write? Oct 19 09:08:26 zeq1: deffinitely Oct 19 09:08:38 but it seems to me to be really easy Oct 19 09:09:01 I still remember the days of NDD and interleave factors :D:D:D Oct 19 09:09:05 I guess you write when you have a full erase block to dispatch Oct 19 09:09:14 hmm Oct 19 09:09:25 sounds sane Oct 19 09:09:44 vi_: any thoughts on that Oct 19 09:09:51 vi_: if you're interested in kbd matrix flaws: I wrote this lil essay in hw:components iirc Oct 19 09:09:59 I guess ShadowJK can help here too Oct 19 09:10:02 there was a kernel patch (RFC) for it a couple of years ago Oct 19 09:10:09 (if I am not mistaken) Oct 19 09:10:31 zeq1: for mtd or compcache? Oct 19 09:10:43 hmm, not mtd Oct 19 09:10:54 flash is not handles by mtd Oct 19 09:10:57 intelligent handling of swap on flash Oct 19 09:10:59 *hadled Oct 19 09:11:16 zeq1: souns friggin useful Oct 19 09:11:36 but that is what Nokia does Oct 19 09:11:48 I don't think it ever went mainstream because *most* people use SSDs now, and the firmware is pretty good at handling it Oct 19 09:11:53 until swap becomes fragmented Oct 19 09:12:27 freemangordon: it shouldn't become fragmented if the blocks are never misaligned Oct 19 09:12:54 indeed Oct 19 09:13:10 it doesn't matter if they are not sequential Oct 19 09:13:16 DocScrutinizer51: zeq1 was asking, I pointed him to your marvellous article. Oct 19 09:13:17 I know Oct 19 09:14:05 but I still thing some background defragmentation should take place Oct 19 09:14:20 tail-packing? Oct 19 09:14:57 hmm, maybe it will come naturally when I start to play with it Oct 19 09:15:20 usually it happns that way Oct 19 09:15:36 it would be really good to have decent swap performance Oct 19 09:16:13 plus less flash wear Oct 19 09:16:24 the other option would be to use...ummm.. how was that called Oct 19 09:16:35 MM that fennec/firefox uses Oct 19 09:16:44 zeq1: flash wear is a myth. Oct 19 09:17:10 aah, yes, jmalloc Oct 19 09:17:10 on devices that have a controller that takes care of wear levelling. Oct 19 09:18:34 what do you think? about jmalloc type of algorithm Oct 19 09:20:01 BTW I am not sure how exactly backing swap works now Oct 19 09:20:43 does it try to compress and write as soon as there is a page request? Oct 19 09:21:07 vi_: any clue? Oct 19 09:26:03 vi_: (a myth) not anymore nowadays, with multilevel cells Oct 19 09:33:01 hey everyone. i just got myself an N900 and would like to know whats going on in the community today. where is a good place to look for current software/distributions if you want to hack it? i am an experienced linux and bsd user but the maemo platform is new grounds :) Oct 19 09:36:38 raccoon_: you may want to install CSSU on it Oct 19 09:36:49 ~cssu Oct 19 09:36:50 i heard cssu is http://wiki.maemo.org/Community_SSU Oct 19 09:37:12 and if you are bold enough, go for Oct 19 09:37:16 ~cssu-thumb Oct 19 09:37:16 i heard cssu-thumb is [thumb2 microb] indeed seems to render like mad, subjectively, or http://talk.maemo.org/showpost.php?p=1220597 Oct 19 09:38:07 raccoon_: but it really depends what you want to exactly "hack" Oct 19 09:38:26 freemangordon: i'm unsure at the moment, so i'd like to see what the community is up to Oct 19 09:38:45 but i'll take a look at CSSU, thanks Oct 19 09:39:34 raccoon_: CSSU is the community firmware release. Oct 19 09:39:42 vi_: allright Oct 19 09:40:12 raccoon_: Think of it being like cyanogenmod but on hyper-steroids and not wearing a back to the future tee-shirt. Oct 19 09:40:27 hahaha Oct 19 09:40:51 actually, cyanogen came to mind when thinking if there were any similar community projects Oct 19 09:40:58 but thanks for updating me about the no-frills approach Oct 19 09:41:03 i also find cyanogen a little too flashy Oct 19 09:41:18 freemangordon: BS is used when a page cannot be compressed by a certain percentage. Oct 19 09:42:58 This way, uncompressed pages will not eat up your precious ramdisk. Oct 19 09:43:11 aah, yes Oct 19 09:43:15 raccoon_: Then you are in luck! The n900 is anything but flashy! Oct 19 09:43:21 sounds ineffective Oct 19 09:43:49 vi_: sounds like my cup of tea :-D Oct 19 09:44:02 still live in the terminal most of my wake hours... Oct 19 09:44:32 then you will never want another device after this one. Oct 19 09:44:45 vi_: i would rather write everything to flash (compressed) while keeping it (compressed) in ram too, until the point RAM is filled Oct 19 09:45:29 then if i have to store a new page, i'll evict the older in RAM and use it. Oct 19 09:46:29 *oldest Oct 19 09:48:17 hmm, i'll have to recall how was that working Oct 19 09:49:05 raccoon_: The most important thing you want to learn is how to reflash the firmware. Oct 19 09:49:08 raccoon_: better come here asking for first 2 weeks. Maemo has some pretty unintuitive idiosyncrasies, e.g. you should be a bit careful with usual install methods like you know from other linux: apt-get etc Oct 19 09:49:25 ~flashing Oct 19 09:49:26 it has been said that maemo-flashing is http://wiki.maemo.org/Updating_the_tablet_firmware Oct 19 09:49:32 hmm, the wlan is pretty damn laggy with the new access points at work Oct 19 09:49:47 i wonder if its some power saving stuff, sshing into the phone is.. icky Oct 19 09:49:57 jaska: Yes, it is. Oct 19 09:50:02 :nod: Oct 19 09:50:09 the old ap was crap otherwise but it wouldnt do this :D Oct 19 09:50:12 yep Oct 19 09:50:16 jaska: You have 2 choices. Oct 19 09:50:26 vi_: 3 :P Oct 19 09:50:27 1. Turn off wifi powersaving when at work. Oct 19 09:50:29 can i disable powersaving on the n900-end Oct 19 09:50:39 i dont want to change the config of the wlan aps Oct 19 09:50:40 yep Oct 19 09:50:41 2. fiddle with the access points settings. Oct 19 09:50:55 DocScrutinizer51: oh, ok, thanks for the heads up Oct 19 09:50:56 even if i am the person who sets them up :) Oct 19 09:50:57 dtim Oct 19 09:51:07 raccoon_: Oct 19 09:51:07 jaska: w00t! Oct 19 09:51:13 ~jrtools Oct 19 09:51:14 methinks jrtools is http://wiki.maemo.org/User:Joerg_rw/tools Oct 19 09:51:36 DocScrutinizer51: You figured out how to trick the powersaving with beecons or something right? Oct 19 09:51:37 at home everything is fine and it doesnt eat up the bat Oct 19 09:51:40 jaska: you have 3 options on n900 re wlan power savings Oct 19 09:52:01 vi_: dtim=1 on AP Oct 19 09:52:14 and beacons = 10Hz Oct 19 09:52:24 jaska: ^ what he said. Oct 19 09:52:26 all you can do Oct 19 09:52:44 trying to figure out how to do it on the n900-side Oct 19 09:53:25 DocScrutinizer05: I suspect this means beacon interbal of 100 ms? Oct 19 09:53:30 on N900: PSM=none Oct 19 09:53:42 freemangordon: ywep Oct 19 09:53:47 DocScrutinizer51: what? Oct 19 09:53:53 yep Oct 19 09:54:13 PSM=none will burn my battery at home for 6 hours with the above settings Oct 19 09:54:36 my cheapo comes with exactly the same values Oct 19 09:54:44 here is a crazy solution... Oct 19 09:54:47 all you can do on N900 Oct 19 09:54:51 Use a bluetooth access point? Oct 19 09:54:59 not sure if the ap can support multiple power saving schemes on virtual ssids Oct 19 09:55:00 hmmm Oct 19 09:55:16 is there any diff betgween on (max) and on (intermediate)?:) Oct 19 09:55:21 but disabling PSM will burn the battery in no time Oct 19 09:55:32 jaska: 500ms/200ms for something Oct 19 09:55:46 or 500/100 Oct 19 09:56:29 at home the wrt54gl with alternative firmware seems to be fine hmm Oct 19 09:57:30 yep, here too Oct 19 09:57:38 ddWRT Oct 19 09:57:39 derp.. changed powersave to intermediate and it wont connect :) Oct 19 09:58:46 vi_: are you sure BT is less power hungry than wifi? Oct 19 10:00:16 arf, theres no fise.def for scv Oct 19 10:00:38 guess ill have to do from scratch Oct 19 10:05:27 freemangordon: I'm Oct 19 10:06:48 freemangordon: for some reason WLAN RX eats battery like candy, while BT needs some humble 2mA Oct 19 10:07:51 my BT been enabled invisible for last 8 months - yeah I know I'm a moron ;-P Oct 19 10:09:22 yust too convenient to simply power up BT heaset and done. While standby time didn't change any noticeably Oct 19 10:10:05 yeah, this is pretty hungry with psm off Oct 19 10:11:32 are the "100mW" ERP or peak? Oct 19 10:12:15 iirc for GSM it's ERP, averaged over TDMA Oct 19 10:12:29 prolly erp Oct 19 10:12:45 :nod: Oct 19 10:13:02 i could probably live with 10mW, the ap isnt many meters away Oct 19 10:13:20 10mW redcommended Oct 19 10:13:28 -d Oct 19 10:13:52 on anything bigger than a phone 100mW is a non-issue :) Oct 19 10:14:11 the diff is marginal though, consumption wise Oct 19 10:14:12 i want an e7 Oct 19 10:14:25 why are they still so expensive Oct 19 10:14:38 e7 shod cost $99 by now heh Oct 19 10:14:45 should Oct 19 10:15:59 hmm i see. oh well, guess ill have to live with the power nomnom Oct 19 10:16:15 at home i can leave psm on max and itll be fine :) Oct 19 10:19:52 jaska: others did a ping per second, dunno if it's more economic than PSM=0 Oct 19 10:20:09 tried, didnt help Oct 19 10:20:13 the ping would just stall Oct 19 10:20:18 and then continue in bursts Oct 19 10:22:54 dtim is 1, beacon interval is 100ms on the aps hmm Oct 19 10:23:31 i wonder if short guard interval etc do anything Oct 19 10:24:07 a-mpdu aggregatin shouldnt touch 802.11g stations Oct 19 10:25:04 freemangordon: Yeah, I even linked to a paper on the subject some 2 weeks ago. Oct 19 10:25:44 the effects do feel like something a-mpdu would cause tho Oct 19 10:25:49 vi_: ok Oct 19 10:30:34 here is a really good paper on power use in phones: Oct 19 10:30:38 http://static.usenix.org/event/usenix10/tech/full_papers/Carroll.pdf Oct 19 10:30:52 it compares G1, N1 and freerunner Oct 19 10:32:06 Our CPU micro-benchmarks show that dynamic volt-age and frequency scaling (DVFS) can significantly re-duce the power consumption of the CPU. However, Oct 19 10:32:09 this does not imply reduced energy overall, because the Oct 19 10:32:10 energy. Oct 19 10:32:12 run-time of the workload also increases. Our results Oct 19 10:32:15 show (Table 3) that only highly memory-bound work-loads (namely mcf) exhibit a net reduction in CPU/RAM Oct 19 10:32:43 is there a one summary of how they all are related to each other? Oct 19 10:32:49 I wish I could read all of it, but kinda cramped at work :p Oct 19 10:33:37 deepy: save it for later, this is a good paper. Oct 19 10:33:50 I will Oct 19 10:33:56 Good as in a good read, I cannot comment on their scientific method yet! Oct 19 10:42:08 It loooked very interesting from the brief part I read and the description Oct 19 10:52:32 hey, do you think if I painted the camera LEDs with some phosphorus gunk, they would produce white light instead? Oct 19 11:14:28 also, i wonder why its preferring the access point which is the other side of the office instead of the nearer one :) Oct 19 11:14:47 wonder if its that it looks at channel 3 before 9 Oct 19 11:15:16 different channels mean different bandwidth in the spectrum right Oct 19 11:15:25 or how one should put it (not exactly my comfortable terms) Oct 19 11:15:29 different frequency Oct 19 11:15:31 ah Oct 19 11:17:34 vi_: since you've done keyboard remapping, do you have any idea how to use Multi_key as, lets say, Mod4 ? Oct 19 11:18:30 its constantly throwing wlan0: driver reports beacon loss from AP xxxxx - sending probe request inth kernel message buffer Oct 19 11:20:14 Skry: Not exactly, however... Oct 19 11:20:25 brb, dumping rx-51 to psatebin. Oct 19 11:22:22 http://cxg.de/_ece6e8.htm Oct 19 11:23:01 Skry: look at lines 459-476 Oct 19 11:23:18 this is where multikey is defined. Oct 19 11:23:53 I do not really understand xkb Oct 19 11:24:09 it is witchcraft sent back in time from the year 500. Oct 19 11:24:20 i agree Oct 19 11:24:36 kinda Oct 19 11:24:38 However on line 471 we can see the assignments for the 'multi' key. Oct 19 11:24:56 on the first level it is control. Oct 19 11:25:17 on level 2, that is shift+multi it is also control. Oct 19 11:25:38 on levels 3+4 it becomes multi-key. Oct 19 11:26:10 level 3 is fn+multi key Oct 19 11:26:25 sorry, wrong Oct 19 11:26:29 level 4 is fn+multi key Oct 19 11:26:37 level 3 is shift+multi key Oct 19 11:26:48 * jaska hits the wlan chipset with a herring Oct 19 11:26:54 no, that is also wrong. Oct 19 11:27:18 hmm, so i guess i could just use level 2 as super_l or whatever Oct 19 11:27:19 wl2170 chipset only responds to fresh water fish. Oct 19 11:28:01 skry I think level 2 is shift+ctrl Oct 19 11:28:19 I tried to turn it into left alt once. Oct 19 11:28:25 normal, shift, fn, fn+shift Oct 19 11:28:26 I did not get very far. Oct 19 11:28:38 jaska: lost beacon usually means AP is sending them too infrequently, or it is using an insane N mixedmode setup Oct 19 11:28:40 jaska: yeah Oct 19 11:28:57 DocScrutinizer51: Is N even a ratified standard yet? Oct 19 11:29:04 or you have 2 AP on channels close together Oct 19 11:29:08 doc: yeah, its in 802.11n 2 antennas in and out Oct 19 11:29:31 with 40MHz channels etc.. think my phone is the only thing here that doesnt do 802.11n :) Oct 19 11:29:31 vi_: yeah, thats about what I'm also trying, I find it somewhat challenging to use tiling wm without extra modifier Oct 19 11:29:43 11n is no longer draft Oct 19 11:30:13 keep in mind WLAN occupies 2 chan up and down from the center chan Oct 19 11:30:30 You just finished moving to 11n? 11ac is coming very soon - Asus has yet another spec buff router that does ac. Oct 19 11:30:31 in 40MHz mode its a bit more Oct 19 11:32:03 I think I'll be getting a USB network card for my N900 - probably a USB everything else, too. Oct 19 11:32:16 yeah, usb ethernet is starting to sound good :) Oct 19 11:32:39 jaska: so a close AP may spoil your 'home' AP's beacons Oct 19 11:32:47 nothing on the same channel Oct 19 11:32:50 I'm thinking of placing my N900 semi-permanently in a flasher dealie, with USB hooked up to a hub that feeds it with power, ethernet, full size keyboard and mouse. Oct 19 11:32:51 or neighboring channels Oct 19 11:33:06 its not affecting anything else. Oct 19 11:33:08 just the n900 Oct 19 11:33:41 Skry: well, theoretically if you made the first instance of 'multi_key' on line 471 into the modifier you wanted...It SHOULD work. Oct 19 11:33:45 and the old access point worked with it.. those were total garbage that needed daily reboots sometimes Oct 19 11:33:56 However I do not think you could have ctrl+alt combinations. Oct 19 11:34:06 bbl Oct 19 11:34:08 jaska: belkin 54g? Oct 19 11:34:15 vi: hp v-m200 Oct 19 11:34:15 o/ Oct 19 11:34:23 utter shit. Oct 19 11:34:28 jaska, usb ethernet indeed - scp over G-wifi is garbage. Oct 19 11:34:29 * vi_ shudders Oct 19 11:34:42 they pretend to be business aps :) Oct 19 11:34:55 vi_: yeah, well, have to figure something out, thanks Oct 19 11:34:56 they randomly stop accepting associations Oct 19 11:35:42 psm=0 it works just fine tho :) Oct 19 11:40:34 jaska: You could put a script in /etc/network/if-up.d to detect if you are connecting to your work wifi and switch psm on/off. Oct 19 11:40:59 vi: psm is per-network so thats handled by the user interface Oct 19 11:41:12 jaska: Oh yes, so it is. Oct 19 11:41:15 oopz Oct 19 11:41:49 Lets all celebrate, something in ICD that was a good idea. Oct 19 11:41:58 celebrate good times c'mon! Oct 19 11:42:02 ~sing Oct 19 11:42:02 I'm Sexy and I know it! Yea! Oct 19 11:43:40 lol wut. winphone7, you cannot set a static IP. Oct 19 11:43:55 windowsCE7 you mean! :) Oct 19 11:44:05 WinCE Oct 19 11:44:05 i dont think its wince7 yet :) Oct 19 11:44:14 ~define wince Oct 19 11:44:15 vi_: I recall being unable to do that in Android too Oct 19 11:44:15 Use ~dict for definitions. Oct 19 11:44:22 ~dict wince Oct 19 11:44:24 Dictionary 'wince' (1 of 8): \Wince\, n. The act of one who winces. [1913 Webster]. Oct 19 11:44:31 i hope Asus fixes our Nexus7 for free:< Oct 19 11:44:33 ~winces Oct 19 11:44:35 Shoddy quality screen Oct 19 11:44:37 ~dict winces Oct 19 11:44:39 could not find definition for winces Oct 19 11:44:48 And then I noticed that I can't connect my Android to AD-HOC networks Oct 19 11:44:49 worst dictionary ever. Oct 19 11:44:50 WP7 doesn't do static IP? If true, that's a major oversight. Oct 19 11:45:03 Hurrian: wp7 is a pile of major oversights Oct 19 11:45:05 Hurrian: no shit. Oct 19 11:45:08 ooh, wince is at 7.. guess they synced the numbers Oct 19 11:45:09 Everything else has been able to do that since forever. Oct 19 11:45:18 deepy, it's for "security" reasons Oct 19 11:45:20 it is unbelieveable. Oct 19 11:45:35 Relying on DHCP for security reasons? Oct 19 11:45:41 vi_: no, ad-hoc Oct 19 11:45:47 no, I was referring to ad-hoc Oct 19 11:45:50 "security"... emphasis on the "" Oct 19 11:45:54 unfortunately, they think users cannot discern between the icon for ad-hoc and AP-mode wifi Oct 19 11:46:09 I still do not get it. Oct 19 11:46:43 If I was going to steal your passwords and inject goatse images into all your web pages I would set up a fake AP, not use and ad-hoc connection. Oct 19 11:47:58 well, uhh, you see, very few Android devices can do true AP mode, even then they need special firmware provided by the device manufacturer because their WiFi chip is a special star revision Oct 19 11:48:30 ... Oct 19 11:48:51 I would'nt use an android phone to launch SSL MITM attacks. Oct 19 11:48:59 Id use an n900 XD Oct 19 11:49:00 I would Oct 19 11:49:01 most people are stuck with ad-hoc wifi mode, and thus google's concern that attacks from rooted crap phones would pwn other idiots' phones. Oct 19 11:49:33 vi_, I wouldn't use an android to do anything that requires a lot of CPU power in a glibc userland Oct 19 11:49:50 Hurrian: How well can they chroot? Oct 19 11:49:57 everything is simply ridiculously slow, even in a chroot. Oct 19 11:50:27 a Debian chroot on my SGS2 is /slower/ on XFCE than my N900 is chrooting the same image Oct 19 11:50:35 I want a phone I can hook up to my computer monitor, chroot into debian, hook up my KB+mouse and then do some real shit. Oct 19 11:50:39 like pr0n Oct 19 11:50:49 the N900? Oct 19 11:51:02 Hurrian: How to hook it up to the monitor? Oct 19 11:51:08 it does, uhh, tv-out and bluetooth HID Oct 19 11:51:11 no RCA plugs on your monitor? Oct 19 11:51:34 iirc ubuntu is working on something like that for android Oct 19 11:51:38 but I don't trust it Oct 19 11:51:40 no and I want more than a 640x480 resolution. Oct 19 11:51:48 vi_, shitty RCA to VGA dongles Oct 19 11:51:53 it's disappointing how few phones have any form of TV out nowadays. Oct 19 11:52:07 Hurrian: I dont even have a TV. Oct 19 11:52:08 deepy, it's been vaporware >1 year now Oct 19 11:52:34 [JOKE] How do you know if someone does not have a TV? Oct 19 11:52:46 [PUNCHLINE] They tell you. Oct 19 11:55:11 vi_: [LAUGH] lol Oct 19 11:55:23 I thought miniHDMI or that MHL thing wasn't that rare these days. Oct 19 11:55:25 vi_: also use a usb vga adapter Oct 19 11:55:42 kerio: kernel modules. Oct 19 11:56:02 use a usb vga adapter and compile the kernel module for it Oct 19 11:56:15 kerio: I have no computer right now. Oct 19 11:56:55 vi_: buy a computer Oct 19 11:57:52 NO U Oct 19 11:57:53 vi_, how's that N950 working out for you? Oct 19 11:58:01 Hurrian: It is not. Oct 19 11:58:24 I cannot use it for fear of hamfisting it into a million pieces. Oct 19 11:58:34 freemangordon; re erase blocks etc.. you dont have to wait until you have an entire block to write (well except with kingston, but that's such a pathetically slow device it's pretty hopeless) Oct 19 11:58:36 haha Oct 19 11:58:37 why? Oct 19 11:58:46 fizzie: I have MHL supposedly Oct 19 11:58:51 Also I hate swipe/harmattan. Oct 19 11:58:52 But I don't have a TV to try it with :p Oct 19 11:59:17 worst interface ever. Oct 19 11:59:24 freemangordon; on the good sandisk you can almost write as much as you want, in any random order, inside one specific writeblock Oct 19 11:59:26 an interfeaces if you will. Oct 19 11:59:39 vi_, autorotate a PITA? Oct 19 12:00:02 Hurrian: autorotate? Oct 19 12:00:14 yeah, autorotate on the homescreen. Oct 19 12:00:38 freemangordon; all the issues start when you touch many erase blocks with writes. At that point the card may read/write a 4M block (last sandisk I bought had 4M blocks) for every request Oct 19 12:00:48 I'd imagine it'd be gobs more useful on the N950 than N9 though Oct 19 12:00:54 oh god, I hate autorotate. Oct 19 12:01:12 personally, I find myself using my N9 more as a phone. Oct 19 12:01:13 It always goes sideways when I dont want it to, then I have to wait ~ 1sec for it to come back. Oct 19 12:01:41 vi_: the n900 can do that too! :D Oct 19 12:01:47 holy gawd that, especially when browsing in bed. Oct 19 12:02:56 if you tend to roll around a lot, prepare for lots of orientation twists. Oct 19 12:02:56 freemangordon; some cards can only have one "open" erase block, some cards can even do 4 (and 4 is about the minimum you need before ext* starts running ok-ish with a single active process writing) Oct 19 12:03:15 ShadowJK: sandisk is good, right? Oct 19 12:03:33 i've got a sandisk C4 which is probably an old C2 Oct 19 12:03:39 sadly Oct 19 12:03:49 kerio, 32GB? Oct 19 12:03:51 yep Oct 19 12:04:01 freemangordon; what fremantle's default swap algorithm does is: find largest free block in swap, write all new swapout sequentially into that block, when end of block is reached: repeat Oct 19 12:04:19 they're advertised as C4, and do 4MB/s seq writes, but drop to 2MB/s random Oct 19 12:04:26 i know Oct 19 12:04:33 it's still fairly good, tho Oct 19 12:04:34 ShadowJK: thanks Oct 19 12:04:43 also hahahaha no Oct 19 12:04:46 freemangordon; if there's minimal other activity on the card (like if you use uSD) it then doesn't even matter if it's written 4k at a time Oct 19 12:04:50 2MB/s random is not even close Oct 19 12:05:03 ~600KB-1MB/s, actually. Oct 19 12:05:05 You treat it when writing as if it was a tape drive :) Oct 19 12:05:06 yeah Oct 19 12:05:06 I think. Oct 19 12:05:11 ShadowJK: but I wan't to make it better than Nokia's algo, if possible Oct 19 12:06:00 ShadowJK: are you sure it wouldn't be better is I wait until I have a whole erase block to write? Oct 19 12:06:09 s/is/if/ Oct 19 12:06:09 freemangordon meant: ShadowJK: are you sure it wouldn't be better if I wait until I have a whole erase block to write? Oct 19 12:06:11 freemangordon: wtf is an erase block Oct 19 12:06:25 eMMC and uSD are both perfect block devices, they do their own thing Oct 19 12:06:57 freemangordon; it's better when there's alot of other activity on the card, like filesystems Oct 19 12:08:44 ShadowJK: yeah, my thoughts exactly. If i write data aligned to and having sizeof eraseblock, underlying block driver wuould not interleave my request with fs system requests Oct 19 12:08:56 kerio; yes they do their own thing, and that means worst case behaviour becomes read-modify-write 4M for every write you send. Take a class 4 card, 4M/s write speed, if you trigger a 4 megabyte read-modify-write with a 4kilobyte write repeatedly, the speed has now dropped to 4kilobytes/sec instead of 4M Oct 19 12:09:25 :c Oct 19 12:09:29 that's ungood Oct 19 12:10:06 there's a reason SSDs have half a gig of ram and dualcore arm cpus, they need oomph to work around flash speed issues :) Oct 19 12:10:15 :D Oct 19 12:10:30 ShadowJK: I suspect i can get eraseblock size foe uSD/eMMC somehow programaticaly? or I should benchmark to find it Oct 19 12:10:32 ShadowJK: build me a ram uSD Oct 19 12:10:47 freemangordon: 4MB Oct 19 12:10:52 kerio: no Oct 19 12:11:04 depends on the card/manifacturer,etc Oct 19 12:11:12 yeah but multiples are fine too Oct 19 12:11:26 kerio, if only we had DIMMs addressable as mmc block device, we would've solved the N900's memory problem by now Oct 19 12:11:39 :) Oct 19 12:11:46 or some sort of psram Oct 19 12:11:50 freemangordon; with newer cards and newer kernels, and when you use a non-usb card reader, it appears as preferred_erase_size or something like that in /sys/mmcblk*/, possibly in the queue subdir Oct 19 12:12:03 Guys, when packaging an executable Oct 19 12:12:13 do I need to include some kind of license? Oct 19 12:12:20 wet dreams of 32GB psram usd :D Oct 19 12:12:48 ShadowJK: so I have to benchmark after all. at least to find the begining of the first block Oct 19 12:12:51 jaska: 1GB would be plenty Oct 19 12:13:08 Note: it's not my executable, I only compiled and now I am trying to get it packaged + optified for maemo Oct 19 12:13:10 yeah itd be fine for swap Oct 19 12:13:21 freemangordon; I'd use 4M alignment Oct 19 12:13:22 MrPingu, include it in the debian description Oct 19 12:13:27 hmm Oct 19 12:13:31 and partitions 4M aligned too Oct 19 12:13:39 no need to include the GPL/whatever license, just say it's licensed under x license Oct 19 12:14:24 (and include link back to author/source website) Oct 19 12:15:34 ShadowJK: sy, you propose to get the size of the whole block device, to find the first 4M aligned sector that is part of my swap partition and to mark it as first secto to use? Oct 19 12:15:48 yes Oct 19 12:15:51 ok Oct 19 12:16:01 sounds sane Oct 19 12:16:01 do swap partitions have a header? Oct 19 12:16:03 You don't even need to know the size of the block device Oct 19 12:16:16 ShadowJK: I think I need to know it Oct 19 12:16:30 But if partition is not aligned to 4M, you need to add an offset I think Oct 19 12:16:38 as noone knows what alignment my swap partition has Oct 19 12:17:19 otherwise I should deal with partition table, but that looks more complcated to me Oct 19 12:18:32 In the source package I have a /doc/ folder with some txt files: README.txt and CREDITS.txt Oct 19 12:18:36 though block size and alignment deffinitely should be passed to driver as module parameters Oct 19 12:18:50 For use on "dedicated" uSD, the nokia algorithm could be improved by adding garbage collection to consolidate the small pieces of data that's left all over the device Oct 19 12:19:00 Do I need them to include in opt/usr/share/doc/ in the package or something? Oct 19 12:19:34 I am just trying to do it right and hopefully I will be able to upload it to extras-devel in the end Oct 19 12:20:03 ShadowJK: Nokia algo in swap is like spaghetti, I would prefer to reimplement a better algo in compcache Oct 19 12:20:18 ShadowJK: btw, how do i make sure my partitions are aligned to some sector size? Oct 19 12:20:44 get your starting cluster from partition table Oct 19 12:20:58 kerio, fdisk -u gives you units of 512 bytes instead of whatever the default is.. all fdisk tinkering is destructive, so beware Oct 19 12:21:41 I did an experiment a few years ago, I recorded every write by kswapd and plotted it, the "sector number" on Y: http://enivax.net/jk/kswapd.png Oct 19 12:21:59 Y represents 0 to 768Megabyte or whatever on emmc Oct 19 12:22:20 ShadowJK: ah crap D: Oct 19 12:22:35 ShadowJK: they're not aligned Oct 19 12:22:36 not even close Oct 19 12:23:35 hmm, I still fail to see why is there no enough free space once we hit 768MB Oct 19 12:23:48 aaah, yes Oct 19 12:23:59 There's free space, but in small chunks Oct 19 12:24:00 ShadowJK: is it worth it to make my partitions 4M-aligned on a uSD? Oct 19 12:24:05 kerio, sure Oct 19 12:24:13 even the swap partitions? Oct 19 12:24:16 don't they have some kind of header? Oct 19 12:24:17 especially Oct 19 12:24:20 ShadowJK: naah, the space is free, but swap driver is unaware of it Oct 19 12:24:36 am I getting ignored? This question isn't too hard for you guys ^^ Oct 19 12:24:44 I had to port/implement swapfree notifications in our kernel Oct 19 12:24:56 freemangordon, consider that it wrote over a gigabyte, but the amount of swap in use at any time never exceeded 140M Oct 19 12:25:06 yes, I know that Oct 19 12:25:33 but a swap sekto is not "freed" until a new request for the same sector comes Oct 19 12:25:41 The normal algorithm, if a swap sector is freed, goes back to that, if it's the "closest" one Oct 19 12:26:52 which makes sense Oct 19 12:27:00 to some extent Oct 19 12:27:46 ShadowJK: do you know what eMMC eraseblock size is? Oct 19 12:27:59 So like, if a gig is written, but, let's say 100M of that is "permanent" in swap, and let's assume that permanent stuff is spread evenly with every write (write 400k, 4k is never modified again), doesn't that result over time in maximum free blocks of 400-4k? Oct 19 12:28:24 mmm, not sure, it's hard to benchmark properly. Oct 19 12:28:25 I see Oct 19 12:28:31 yeah Oct 19 12:29:08 but then reading/writing 4M at a time will still be slower than that Oct 19 12:29:31 hopefully I have big enough cache space Oct 19 12:30:11 ShadowJK: what do you think: Oct 19 12:30:14 So, if over time the average free block size converges towards 400kbyte, it's easy to see why the writes start to jump all over the place with time Oct 19 12:30:21 yeah Oct 19 12:30:23 so: Oct 19 12:31:08 keep compressed pages in RAM(for fast access if needed) and in the same time keep them on disk, uncompressed Oct 19 12:31:38 write back to disk in a work queue, not in real-time Oct 19 12:32:35 when RAM is full, decompress/write the last used pieces to disk, and free the RAM Oct 19 12:32:47 ... Oct 19 12:32:52 or something like that Oct 19 12:32:54 yeah Oct 19 12:33:21 that way fragmrntation would not have any effect, as I'll buffer in compressed RAM Oct 19 12:33:42 Yeah if you write out in 4M chunks fragmentation stops being an issue Oct 19 12:33:56 even if I have to write a page at a time, as it is backed by compressed RAM Oct 19 12:35:22 i.e. all flash access (except for reading a page not in compressed RAM) will be done via a work queue Oct 19 12:35:39 Then you'll have issues when a 4k subsection of something htat's already in swap gets modified by a process :) Oct 19 12:35:40 though 4M seems to much to me Oct 19 12:36:27 should not matter Oct 19 12:36:34 hm, it appears that i *do* have my partitions 4M-aligned Oct 19 12:36:34 aiui Oct 19 12:36:43 16 sectors is 8M, right? Oct 19 12:36:50 no Oct 19 12:36:51 no wait Oct 19 12:36:53 i'm an idiot lol Oct 19 12:37:02 ~16*512 Oct 19 12:37:02 8192 Oct 19 12:38:26 ShadowJK: if something modified in RAM gets modified and needs to be swapped after that, will lead to freeing of the chunk in compressed RAM and allocation of another wone Oct 19 12:38:31 I've benchmarked many cards, but most of my benchmarks don't go far enough to put a number of block size :/ Oct 19 12:38:31 *one Oct 19 12:39:25 I wont put the contents of what needs to be in flash as some buffer assigned to the work queue Oct 19 12:39:50 freemangordon, yeah I just mean that it ends up in the same situation as with the nokia version, except with the difference that you don't take as much penalty from other simultaneous filesystem activity Oct 19 12:40:20 well, stage 2 will be GC Oct 19 12:41:16 but I will have ti find eraseblock size, otherwise the underlying block device driver will mess the things Oct 19 12:41:41 or find a way to tell him to not put my "write" requests in the queue Oct 19 12:42:19 Hm, most the cards I've tested I've only tested with block sizes of up to 512k Oct 19 12:42:29 but recently I've started testing to 4M Oct 19 12:43:15 So if I throw random reads at the device, the throughput should reach a plateau once the correct erase block size is used Oct 19 12:43:52 ouch, we have a penalty for unaligned reads? Oct 19 12:43:59 no Oct 19 12:44:06 Sorry I meant writes. Oct 19 12:44:12 aah, ok :D Oct 19 12:44:23 ShadowJK: do you reckon writing zeros to a partition on a uSD speed up the first writes you do to it afterwards? Oct 19 12:44:54 i mean, are the controllers smart enough to keep the block erased instead of writing zeros to it? Oct 19 12:45:15 Sandisk microsd 32g class 4: 16K: 22.49 kbyte/s, 128k: 127.77 kbyte/s, 1024: 977, 2048: 1854, 4096: 4133 kbyte/s Oct 19 12:45:15 ShadowJK: assuming your partition is eraseblock aligned Oct 19 12:45:35 Yes I test on entire device without partition table Oct 19 12:45:41 aah, ok Oct 19 12:45:46 kerio, I think erased blocks are 0xFF actually, all 1s Oct 19 12:46:36 Now for that sandisk card, each doubling of block size doubled throughput.. I bet, that if I had tested it at 8M, it wouldn't have gone much faster than that approx 4Megabyte/s... Oct 19 12:46:50 ShadowJK: but... there's no /dev/one :( Oct 19 12:46:51 cache? Oct 19 12:47:01 not in use Oct 19 12:48:37 hmm, if I write 4M a time, I would block fs access to the filesystem Oct 19 12:48:43 hm, old sandisk 16 gigabyte USB flash drive, 1M:2M, 2M:4M, 4M:4.5M, my guess is 2M block size for that drive :) Oct 19 12:48:49 s/filesystem/block device/ Oct 19 12:48:50 freemangordon meant: hmm, if I write 4M a time, I would block fs access to the block device Oct 19 12:48:58 ShadowJK: how big is the swap partition header? Oct 19 12:49:14 512? Oct 19 12:49:25 won't that mess up the alignment? :( Oct 19 12:49:26 kerio: good pint Oct 19 12:49:33 point* Oct 19 12:49:40 yes, it will Oct 19 12:49:53 freemangordon, well in theory with 4M erase blocks, in a heavy-write situation, a 4k write is going to be as expensive as a 4M write. The only things that are "free" are reads :) Oct 19 12:50:09 hrmpf, should i move my partitions back one sector? Oct 19 12:50:09 :) Oct 19 12:50:17 reads are not fire Oct 19 12:50:21 free Oct 19 12:50:25 just cheap Oct 19 12:50:28 yes Oct 19 12:50:46 well, that 4k write can be combined with other stuff by block driver Oct 19 12:51:07 while nothing can stop my 4M write request Oct 19 12:51:09 I think one could get away with writing .5M of that 4M block, allow some reads to go through, then write another .5M, let some reads through, etc.. but interleaving other filesystem writes with swap writes is just going to make things slower most of the time Oct 19 12:51:26 Yeah so ideally we'd want a new disk scheduler too.. Oct 19 12:51:40 That better understands the costs Oct 19 12:51:47 ShadowJK: the write itself will be slower, but swap responce not Oct 19 12:51:54 Are you trying to push ramzswap into Maemo5 or what? Oct 19 12:52:01 http://item.mobileweb.ebay.co.uk/viewitem?itemId=271080836262&index=22&nav=WATCHING&nid=38405072568 Oct 19 12:52:06 as it is cached in compressed ram Oct 19 12:52:13 surprisingly expensive Oct 19 12:52:23 RST38h: it is there for ages, but backing swap does not work ok Oct 19 12:52:26 freemangordon, I mean throughput on filesystem will be slower too Oct 19 12:52:36 a linear swapper would be nice too Oct 19 12:52:45 yeah, but we can do nithing in that regard Oct 19 12:52:46 SpeedEvil: it's already linear Oct 19 12:52:58 kerio: for the first pass Oct 19 12:52:59 freemangordon: I have seen people doing it, but it is definitely not there out of the box Oct 19 12:53:13 SpeedEvil, it needs garbage collection to become linear second time round Oct 19 12:53:16 kerio: keeping it non fragmented gets more complex Oct 19 12:53:18 freemangordon: Nokians themselves say N900 has too small RAM for ramzswap Oct 19 12:53:20 indeed Oct 19 12:53:38 nokians said USB host mode was impossible Oct 19 12:53:46 of course it is not in Nokia kernel, but is in KP for more than an year Oct 19 12:53:48 (by implication) Oct 19 12:54:12 still too messy to bepractical =) Oct 19 12:54:19 I'd be interested in using a compressed buffer, to read in the "least occupied 4M block of swap", while device is idle, and idly collect together swap Oct 19 12:54:25 RST38h: and if you keep dedicated RAM to a reasonable value, it helps sometimes Oct 19 12:54:31 sometines not Oct 19 12:54:56 I find it challenging to get "hot" data in ramzswap and "stale" data in emmc swap Oct 19 12:55:12 ShadowJK: basically a circular pointer, with the next written to block being the least occupied block, and you read before write to compress it Oct 19 12:55:23 Obviously the idea of backing swap, but backing swap as it is now is dead slow Oct 19 12:55:23 ShadowJK: or similar Oct 19 12:55:44 ShadowJK: we don;t need compressed data in backing swap Oct 19 12:55:45 where block = 1m or whatever Oct 19 12:56:19 SpeedEvil, it's something like that right now, once it runs off the end, it sets its pointer to start of largest unoccupied area afaik Oct 19 12:56:46 it's just that those areas are ludicrously tiny Oct 19 12:56:57 ShadowJK: yeah' but the lack of integral defragmentation screws it Oct 19 12:57:45 Btw, I've sene situations where something unexpected (like accidentally opening a gigantic image in the browser) has pushed lots of stuff into swap, but after the browser was closed, there was a huge contiguous chunk free in swap, and on that swap graph it has then drawn a straight line through that freed block :) Oct 19 12:58:09 ShadowJK: so... should my swap partitions be aligned to 4M - size of swap header? Oct 19 12:58:17 guys, my idea is to have compressed RAM similar to L1 cache, and backing device similar to main memory Oct 19 12:58:24 kerio, I don't know what it does with the header Oct 19 12:58:47 freemangordon, yes Oct 19 12:59:32 freemangordon: yeah, it's questionable what swap compression gets you. sure t can boost swap bandwidth from 12m to 30m (say) Oct 19 12:59:37 with a L1 written back to the main memory in the background Oct 19 12:59:46 but is that going to help much Oct 19 12:59:51 SpeedEvil: naah, it is much more Oct 19 13:00:18 something like half of the bandwith of the RAM itself Oct 19 13:00:22 or similar Oct 19 13:00:26 You know, I'd want to run an entire mmc behind some translation layer+gc ;) Oct 19 13:00:27 wow Oct 19 13:00:28 that'd solve alot Oct 19 13:00:49 ShadowJK: or a punch through to the raw band Oct 19 13:01:03 NAND Oct 19 13:01:17 Then you'd still need a ftl in software ontop of that :) Oct 19 13:01:36 plus a gig of ram on the SD card Oct 19 13:02:25 well, I think I got enough info for the start, thanks a lot Oct 19 13:02:42 SpeedEvil: you mean a 1GB SD card made of ram Oct 19 13:02:49 no Oct 19 13:02:57 it would be good, tho Oct 19 13:03:14 current microsds are 9chips laminated together Oct 19 13:03:30 if you make 2 of the 8 storage ram. Oct 19 13:03:57 48g ROM/2g ram Oct 19 13:04:13 So I just did my first announce thread: http://talk.maemo.org/showthread.php?p=1282540#post1282540 :D Oct 19 13:04:50 hmm Oct 19 13:05:00 does it understand streams? Oct 19 13:05:14 I mean will it understand Oct 19 13:05:16 pat Oct 19 13:05:21 tern Oct 19 13:07:47 It should Oct 19 13:08:08 tern Oct 19 13:08:09 pat Oct 19 13:08:23 if fragmented and out of order? Oct 19 13:08:59 never used it before, though. I saw a guy asking for it on IRC so I saw it as oppurtunity to learn and serve the community Oct 19 13:09:09 :) Oct 19 13:09:45 SpeedEvil: http://ngrep.sourceforge.net/usage.html Oct 19 13:11:56 I use a TCP logger for that sort of thing. Oct 19 13:12:05 I've fortpgotten the name Oct 19 13:12:14 interesting alternative though Oct 19 13:12:17 SpeedEvil: TCPdump? :P Oct 19 13:12:25 no Oct 19 13:12:53 it reassembles TCP streams matching certain properties and saves them to disk Oct 19 13:13:21 Actually, I also compiled a newer TCPdump, libpcap, hamster and ferret too Oct 19 13:13:25 just needs packaging Oct 19 13:13:37 and testing... Oct 19 13:13:44 :-) Oct 19 13:18:24 ShadowJK: should i also repartition the eMMC to align the data to 4M? Oct 19 13:19:51 i mean, is it worth it? Oct 19 13:20:48 4k align is extremely important Oct 19 13:21:23 also wtf, partition 2 does not end on cylinder boundary? Oct 19 13:21:27 4m is exta bonus Oct 19 13:21:32 the end of the eMMC is not on a boundary? Oct 19 13:21:33 :s Oct 19 13:21:54 oh balls Oct 19 13:22:02 it ends one sector before the end, for some reason Oct 19 13:22:07 cylinders are obsolete Oct 19 13:22:27 yeah but i'm wasting 512 bytes on my optfs Oct 19 13:30:27 dooom Oct 19 13:37:53 jaska: i paid for those 512 bytes! Oct 19 13:47:27 no, those were a free gift on top Oct 19 13:48:05 i'd repartition the eMMC, but... eeeeeh Oct 19 13:49:39 Pali: can i move bootmenu.scr and bootmenu.img.d from MyDocs to the uSD? Oct 19 13:49:51 no Oct 19 13:50:03 :c Oct 19 13:50:10 uboot loading only bootmenu.scr from MyDocs Oct 19 13:50:23 Pali: that's some BS, what if there's no MyDocs? Oct 19 13:50:46 then fallbacking to default menu entry Oct 19 13:50:58 meh, i suppose that's fair Oct 19 13:51:12 but first eMMC parition is needed for maemo (and is FAT32) Oct 19 13:51:14 Pali: is there a way to load a config file from the uboot command line? Oct 19 13:51:17 and uboot use it Oct 19 13:51:29 kerio, it is not config file, but uboot hush script Oct 19 13:51:36 and yes, you can do it Oct 19 13:51:45 ok, so it's not really a big deal Oct 19 13:51:46 cool Oct 19 13:52:01 first need to learn uboot hush scripting Oct 19 13:52:37 boot from sd card also trying to run script 'boot.scr' from any sd card paritition (fat, ext2/3/4) Oct 19 13:54:12 is bootmenu something implemented in the .scr, or is that just a config file for it? Oct 19 13:57:13 it could be neat to load a submenu from the uSD Oct 19 13:57:26 that's only configured there Oct 19 14:19:38 ~seen robbiethe1st Oct 19 14:19:40 robbiethe1st <~robbiethe@50-37-93-93.mscw.id.frontiernet.net> was last seen on IRC in channel #maemo, 12h 24m 14s ago, saying: 'http://maemo.org/packages/package_instance/view/fremantle_extras-testing_free_armel/backupmenu/1.2.0-1/ Give me karma'. Oct 19 14:26:28 kerio, bootmenu is uboot command Oct 19 14:26:59 and bootmenu.scr is only uboot hush script which set some env and run bootmenu Oct 19 14:38:03 weird Oct 19 14:38:34 my second n900 booted up with "dead" display Oct 19 14:38:44 sound, keyboard light etc worked fine Oct 19 14:39:12 shutdown, removal of battery fixed it Oct 19 14:39:56 it has been sitting on my shelf and is hardly used, almost in mint condition. Oct 19 14:51:17 entitled: accidental act-dead? Oct 19 14:51:55 ? Oct 19 14:52:11 now idea what happened Oct 19 14:52:15 no Oct 19 14:52:17 neither do I Oct 19 14:52:47 it's also nearly clean install Oct 19 14:53:50 rootsh, openssh client+server, FAM, that's all Oct 19 14:56:01 just about to reflash it again and install thumb2 Oct 19 15:01:42 yay thumb2 ^_^ Oct 19 15:02:48 * entitled has three n900 with me now Oct 19 15:03:38 third one is friend's, it got wet and won't boot anymore, was erratic for a while (screen flashing etc) Oct 19 15:04:07 i'll try drying it up and if that won't help, then dismantle it and see if there is any rust to be removed Oct 19 15:04:22 entitled: when did you get it? Oct 19 15:07:27 today Oct 19 15:07:37 the damage happened months ago Oct 19 15:07:40 rain water Oct 19 15:07:47 so it might not be that bad Oct 19 15:08:00 with tap water it would be way more corroded Oct 19 15:09:21 fff Oct 19 15:16:22 reflashing is always bit nerving Oct 19 15:16:47 forgot to remove the battery after the EMMC Oct 19 15:17:13 so did it all again just in case Oct 19 15:17:50 not much hope for your rain-dancer Oct 19 15:18:16 :.( Oct 19 15:18:48 electrolytic corrosion Oct 19 15:20:18 we'll see Oct 19 15:20:37 DocScrutinizer51: what's the procedure for water damage again? Oct 19 15:20:43 with a non-empty bupbat Oct 19 15:21:22 rinse with even more water, then with distilled water, then 99%ethanol Oct 19 15:21:47 distilled water and IPA was n my plans Oct 19 15:22:00 no chance for anhydrous alcohol Oct 19 15:22:03 or did you mean the 'remove battery *immediately*' Oct 19 15:22:18 DocScrutinizer51: there are two batteries in the n900 Oct 19 15:22:20 anhydrous acetic acid yeah but no, no alcohol :D Oct 19 15:22:25 and only one is easily removable :) Oct 19 15:22:43 yeah Oct 19 15:22:53 do you have to remove both? Oct 19 15:23:34 btw, is there need to unplug and unplug the usb cable between any rootfs/emmc/emmc flashing? Oct 19 15:23:49 you cant remove bupbat Oct 19 15:23:58 I did the above with an entire laptop (not diaplay) a couple of months at Oct 19 15:24:04 ago Oct 19 15:24:07 DocScrutinizer51: :o Oct 19 15:24:07 why Oct 19 15:24:08 ? Oct 19 15:24:15 pulled ram, HD, washed it all. Oct 19 15:24:21 and damage from bupbat possibly limited Oct 19 15:24:22 it booted up (to some point, can't remember) when I plugged the cable again, after emmc and no battery removing Oct 19 15:24:33 (it had drunk a lot of coffee) Oct 19 15:25:17 entitled: flash combined with -R, and keep holding u, then flash vanilla without -R and then combined with Oct 19 15:25:28 -R is for reboot Oct 19 15:25:38 yeah I know Oct 19 15:26:03 why reboot at the FIRST rootfs flash? Oct 19 15:26:26 because the first flash is to ensure your system isn't too fucked up to flash vanilla Oct 19 15:27:07 you need a working rootfs to flash vanilla, but after flashing vanilla you MUST flash rootfs Oct 19 15:27:11 or at least make sure you never booted from it Oct 19 15:27:33 ah Oct 19 15:28:10 didn't know it's recommended to reboot it once after the first rootfs flash Oct 19 15:30:43 note: if you have a somewhat working fremantle install, you can skip the first or the last combined flash Oct 19 15:31:15 if you don't care about restoring factory conditions for MyDocs, you can skip the flashing of vanilla Oct 19 15:31:31 MyDocs and the rest of eMMC, actually, but still Oct 19 15:31:47 yeah Oct 19 15:32:44 I could have probably skipped the eMMC with this as it was so clean to begin with Oct 19 15:33:24 but I do it just in case if there's any dirt in there Oct 19 15:33:29 indeed Oct 19 15:33:53 i wonder if there's a way to "garbage collect" CAL Oct 19 15:35:03 a proper cleanup would be "dump current CAL data, flash_eraseall /dev/mtd*, write CAL data back, write zeros to eMMC, flash COMBINED/VANILLA/COMBINED" Oct 19 15:49:16 DocScrutinizer05, did you get the scroll wheel to work properly on the Bluetooth MX Revolution? Oct 19 15:50:51 Pali: hm, what's the garbage at the beginning of bootmenu.scr? :o Oct 19 15:51:12 kerio, signature Oct 19 15:51:21 it is also mkimage generated Oct 19 15:51:45 uboot can load only mkimage files (scripts, kernels, initrd, ...) Oct 19 15:51:49 uboot is really demanding! :c Oct 19 15:56:06 Anybody using russian keyboard layout? Oct 19 15:57:38 sure Oct 19 15:57:42 GeneralAntilles: Oct 19 15:58:39 GeneralAntilles: http://maemo.cloud-7.de/mx-revolution/ Oct 19 15:59:13 GeneralAntilles: the nice thing: it's sticky in mouse-mpu Oct 19 16:01:38 The Bluetooth version doesn't work with any of the OS X mouse drivers. Oct 19 16:01:53 So I'm using it on the laptop, but I hadn't been able to get it figured out so far. Oct 19 16:02:14 The RF version works out-of-the-box with Linux Mint, though. Which was cool. Oct 19 16:03:15 I got both here now Oct 19 16:03:20 Thanks, I'll take a look at this tonight. Oct 19 16:03:29 yw Oct 19 16:03:33 I broke down and ordered a used NIB MX Revolution (RF) from Amazon this morning. Oct 19 16:03:36 $135. Oct 19 16:03:46 ouch Oct 19 16:03:51 Wish I had thought to stock up on them when they were on sale being discontinued. Oct 19 16:04:06 Well, it's basically the perfect mouse Oct 19 16:04:13 yep Oct 19 16:04:14 I get about 3-4 years out of them. Oct 19 16:04:15 So Oct 19 16:04:28 though my MMB again fsckd Oct 19 16:05:06 well, not again, since I never changed that critter (SMT) Oct 19 16:07:02 the new one is at office now, and guess what - our sysadm refuses to install the friggin trackpoint.exe or what's the name, for me. So the whole mouse is in default mode, and not even 7-button Oct 19 16:07:25 TWO real and two mousewhell-buttons Oct 19 16:07:44 so even worse than the crappy dell mouse I had to use before Oct 19 16:07:57 modulo the revolution wheel Oct 19 16:08:24 I guess I have to quit this job, they're mobbing me out with such BS Oct 19 16:08:47 (no horizontal scrolling in winXP) Oct 19 16:09:22 :-/ Oct 19 16:10:00 not even outlook (which was evil enough), but lotus domino webmailer OMFG Oct 19 16:11:25 rationale: "we need to keep up productivity - imagine every of our 300 employees comes with *own* mouse" Oct 19 16:11:47 what the F**K is with MY productivity??? Oct 19 16:11:49 So, what is wrong about that? Oct 19 16:12:55 Doc: Your next email request should contain your manager and that bozo's manager in the CC: field Oct 19 16:13:05 DocScrutinizer05, sure? who? Oct 19 16:13:30 WHo did what? Oct 19 16:13:39 who is using russian keyboard layout? Oct 19 16:13:42 Doc: And contain a clear explanation about whose productivity you would like to keep up, as a comment to the quoted statement from the bozo Oct 19 16:14:03 Doc: And yes, even although he will be in the To: field, he is not the one you are sending it to =) Oct 19 16:15:06 um, I only see 5 items in community ssu (thumb) and (testing-thumb)? Oct 19 16:15:55 (and, if the it bozo were really smart, he would have framed it as a question about who can guarantee that trackpad.exe is not a security risk) Oct 19 16:17:23 my manager is boss of that company, and I'm not sure he really cares Oct 19 16:18:32 akls: ?? Oct 19 16:18:51 Want me to make a phone call? I'll ask for bozo and launch explain.bat Oct 19 16:19:04 DocScrutinizer05, Anybody using russian keyboard layout? Oct 19 16:19:18 I mean.. is there any person here to help me setting it up Oct 19 16:19:50 It's tough using non-optimal inputs. Oct 19 16:20:23 indeed Oct 19 16:21:07 and that lotus domino webmailer is DA FAQ Oct 19 16:21:48 I mean, so fsck it doesn't work at all under FF Oct 19 16:53:00 akls: still fighting with it? Oct 19 16:53:07 freemangordon, yes Oct 19 16:53:08 try ukeyboard Oct 19 16:53:21 akls: ^^^ Oct 19 16:53:55 on screen layouts? Oct 19 16:54:04 I have no problems with virtual keyboard Oct 19 16:54:12 I want a hardware russian keyboard Oct 19 16:54:24 it's there, but if I enable it then it breaks latin layout Oct 19 16:56:42 akls: try it Oct 19 16:57:21 it comes with its own vkb files Oct 19 16:58:00 akls: Oct 19 16:58:09 did you try to change languages order? Oct 19 16:58:13 freemangordon, is it going to help me with my HARDWARE keyboard? Oct 19 16:58:21 i mean the realy physical keyboard Oct 19 16:58:23 real* Oct 19 16:58:35 freemangordon, that's a good idea! Oct 19 16:58:38 let me try it.. Oct 19 16:58:47 akls: I amn not that stupid, thanks. I understand your question pretty well ;) Oct 19 17:00:04 akls: you may also try extkbd if changing order and uleyboard does not help Oct 19 17:00:37 extkbd comes with newer kbd data for RX-51 Oct 19 17:01:11 changing order didn't help Oct 19 17:01:18 freemangordon: wait, what? Oct 19 17:01:23 I'll try ukeyboard now Oct 19 17:01:24 new kbd data for rx-51? :O Oct 19 17:01:37 i thought it was a generic xkb update Oct 19 17:01:55 yes, it is Oct 19 17:02:05 newer version of xkb-data Oct 19 17:02:39 kerio: I have NFC whether someone didn;t messed with akls' xkb-data Oct 19 17:03:03 I've tried reflashing, it didn't help Oct 19 17:03:11 aah, ok Oct 19 17:03:26 extkbd won't help then Oct 19 17:04:11 so, when switching to latin keyboard somehow it doesn't update arrow keys Oct 19 17:04:12 akls: ukeyboard remap not only VKB, but HW kbd too. At least it does that for bulgarian :) Oct 19 17:04:37 is there any file describing latin keyboard layout? Oct 19 17:04:45 akls: I think we were wrong yesterday when we told you to set HW layout to russion Oct 19 17:04:50 *russian Oct 19 17:05:09 you should keep englis(finland or whatever) Oct 19 17:05:21 ok, but how can I use russian then? Oct 19 17:05:39 ctrl-space should work. install ukeyboard and try it Oct 19 17:06:13 installing Oct 19 17:06:18 though I can swear russion should work without ukeyboard Oct 19 17:06:24 damn Oct 19 17:06:27 *russian Oct 19 17:06:46 int_ua: ping Oct 19 17:07:44 freemangordon, maybe I'm doing something completely wrong? Oct 19 17:08:08 freemangordon, current setup: hardware layout: English, Nederlands Oct 19 17:08:15 1st language English (USA) Oct 19 17:08:18 Right Oct 19 17:08:21 2nd language Русский Oct 19 17:08:22 akls: hmm, my second device don't have ukeyboard, lemme see how it behaves Oct 19 17:08:29 Click on the secondlanguage Oct 19 17:08:34 Select Russian Oct 19 17:08:40 RST38h: ^^^ Oct 19 17:08:52 yep, now if I type something I'll get english letters Oct 19 17:08:54 he already did that Oct 19 17:08:57 Be aware that it sucks, misses the soft sign and I think "yu" Oct 19 17:08:57 then I press ctrl+space Oct 19 17:09:08 it says that now it's changed to RUSSIAN Oct 19 17:09:13 good Oct 19 17:09:13 sounds good, but it's still english Oct 19 17:09:29 * RST38h seems to remember this problem Oct 19 17:09:34 Ah, I remember now Oct 19 17:09:40 RST38h: ukeyboard? Oct 19 17:09:47 fuck ukeyboard Oct 19 17:09:50 hehe Oct 19 17:10:06 It is a bug in the Maemo5 firmware that only switches virtual keyboard but not the physical one Oct 19 17:10:18 I have bulgarian phonetic thanks to it, so long live ukeyboard Oct 19 17:10:40 The switch actually DOES work, if you have got a device built for the Russian market Oct 19 17:10:58 This problem is fixable by swapping a few layout files on the device. Oct 19 17:11:10 I have Global thing installed Oct 19 17:11:13 akls: try ukeyboard Oct 19 17:11:20 maybe I should have installed something else Oct 19 17:11:28 freemangordon, already! Nothing has changed Oct 19 17:11:35 hmm, lemme try here Oct 19 17:11:49 I can set hw layout to russian Oct 19 17:11:55 and it's going to work PARTIALLY Oct 19 17:12:05 so when I change back to Latin, arrows keys are still russian Oct 19 17:12:17 maybe there's any way to fix that? Any file describing Latin layout? Oct 19 17:12:30 akls: WFM Oct 19 17:12:42 akls: Read here: there is a layout file for that Oct 19 17:12:59 RST38h, where? Oct 19 17:13:02 RST38h: wait a minute, it works here Oct 19 17:13:12 a moment Oct 19 17:13:13 without messing with .vkb files Oct 19 17:13:19 freemangordon, and what is wfm Oct 19 17:13:24 ~wfm Oct 19 17:13:24 methinks wfm is (Wired For Management Baseline) This is an Intel hardware specification that is designed to allow for compliance with easier management of desktop PCs in a networked environment. The specification calls for computers to be compatible with a pre-boot protocol that can be used to update the system or perform other management options. Also, the computer must be compatible with network/desktop management applications.. Works For Me Oct 19 17:13:27 akls: http://habrahabr.ru/qa/4130/ Oct 19 17:13:30 works for me Oct 19 17:13:51 akls: http://n8xx.com/topic3827-foneticheskie-raskladki-dlya-ne-kirilicheskikh-klaviatur.html Oct 19 17:15:32 ukeyboard has "Русский - Phonetic" layout Oct 19 17:16:06 thou it does not work :D Oct 19 17:16:33 built in kbd firmware works Oct 19 17:16:37 just needs a file Oct 19 17:17:22 other folks (israelis, greeks, bulgarians, georgians) are in a way worse position than russians though, no built in layout for them Oct 19 17:17:57 * RST38h boots up his n900, will have layout file in a moment Oct 19 17:18:01 strangely "Π‘ΡŠΠ»Π³Π°Ρ€ΡΠΊΠΈ - phonetic" with second language "Русский" works ok :D:D:D Oct 19 17:18:16 phonetic is for wussies Oct 19 17:18:41 real men (tm) remember their JCUKENG by touch Oct 19 17:18:53 interesting thing. /usr/bin/osso-product-info says that OSSO_PRODUCT_KEYBOARD='German' Oct 19 17:18:54 interesting Oct 19 17:19:11 akls: got AZERTY? Oct 19 17:19:18 NOPE. Oct 19 17:19:22 it is qwerty Oct 19 17:19:23 weird indeed Oct 19 17:19:27 looks like us keyboard Oct 19 17:19:32 dunno, maybe it was changed Oct 19 17:19:38 that is actually Netherlands :) Oct 19 17:19:49 That is their default eu keyboard Oct 19 17:20:02 There is also crooked russian (no arrows) and normal russian (arrows) Oct 19 17:23:38 so? :\ Oct 19 17:23:56 I'm reading stuff from these links Oct 19 17:24:00 but I can't find a fix Oct 19 17:24:31 oh, by the way Oct 19 17:24:39 this thing WAS WORKING for a little while Oct 19 17:30:01 LOL Oct 19 17:30:31 Date format and decimal separator are defined by device language Oct 19 17:30:37 not by regional settings Oct 19 17:30:38 wtf? Oct 19 17:32:05 akls: the whole ui thing has been dictated by some "user experience specialists" who had absolutely no idea about the real user experience Oct 19 17:32:11 so, do not judge it too harshly Oct 19 17:32:27 motherfuckers Oct 19 17:32:43 well, it is kinda academic now, with Nokia basically dead Oct 19 17:34:19 akls: start with /etc/X11/xkb/base.xml Oct 19 17:36:02 Now, go to /usr/share/X11/xkb/symbols/nokia_vndr Oct 19 17:36:28 You will see file called rx-51 there. You will have to replace it with a file I email you (please give your email address) Oct 19 17:37:53 ok Oct 19 17:38:14 sent Oct 19 17:38:49 anything I should do in /etc/X11/xkb/base.xml? Oct 19 17:39:17 nah Oct 19 17:39:20 not fornow Oct 19 17:39:36 opera-mobile got updated on maemo5 =) Oct 19 17:39:44 those opera folks are so damn nice Oct 19 17:40:01 orhn is king of the bro s Oct 19 17:40:38 He is supporting a dead platform for nothing but the lulz Oct 19 17:41:03 what have I missed? Oct 19 17:41:24 Skry: did you solve the super modifier issue? Oct 19 17:51:11 RST38h, the file is replaced Oct 19 17:51:14 now what? reboot? Oct 19 17:51:20 and which settings should I use now? Oct 19 17:52:39 enable russian as 2nd language and reboot Oct 19 17:53:00 and for hardware layout I should set Enlgish, Nederlands or what? Oct 19 17:53:19 as I have said - set it to English and Russian Oct 19 17:53:31 errr Oct 19 17:53:36 in Text input settings Oct 19 17:53:43 yes Oct 19 17:53:44 there's "Hardware keyboard layout" Oct 19 17:53:48 yes Oct 19 17:53:52 and "1st language" and "2nd language" Oct 19 17:54:02 set 2nd to russian Oct 19 17:54:15 and hardware keyboard layout to english? Oct 19 17:55:27 russian Oct 19 17:55:33 oh Oct 19 17:56:04 holy shit Oct 19 17:56:11 desktop in portrait mode Oct 19 17:56:33 non penis canina, yes Oct 19 17:56:47 It's a first time I saw it working after upgrading :O Oct 19 17:57:29 that is cssu for you Oct 19 17:57:53 YESSSSSSSSSSSSSSSSSSSSSSSSSSSS Oct 19 17:57:57 YESSSSSSSS Oct 19 17:58:10 what? the keyboard finally works? Oct 19 17:58:16 yes! :) Oct 19 17:58:29 RST38h, thank you! Oct 19 17:58:32 just to make it clear Oct 19 17:58:40 ok Oct 19 17:58:43 RST38h, is that a file that describes the Latin layout? Oct 19 17:58:57 it is a file that describes all keyboard layouts Oct 19 17:59:06 oh Oct 19 17:59:07 it contains the "correct" definition of the russian keyoard Oct 19 17:59:15 as opposed to the crooked one with no arrows Oct 19 17:59:33 ok Oct 19 17:59:57 can it be added to the next upgrade so I wont need to do that again after reflashing? Oct 19 18:00:58 dunno Oct 19 18:03:08 RST38h, it even has all other characters in right positions! Oct 19 18:03:12 damn it! You're awesome! Oct 19 18:03:17 it still misses YU Oct 19 18:04:16 ... Oct 19 18:04:22 and how am I supposed to type then? Oct 19 18:04:33 I usually open virtual keyboard Oct 19 18:04:39 .......... Oct 19 18:04:40 or put IO :) Oct 19 18:05:01 Well, N97 keyboard missed B, if this makes you feel better Oct 19 18:05:27 is it possible to set down arrow as ю? Oct 19 18:05:46 you have got the file. you have got the text editor. you know what to do. Oct 19 18:06:02 alright Oct 19 18:06:19 but not today Oct 19 18:06:21 it's too late Oct 19 18:06:25 anyway, thank you very much! Oct 19 19:31:02 vi_: not yet at least, was distracted of the matter Oct 19 19:33:08 is it possible to RMA an n900 for broken usb connector even though i don't have any receipt for it? Oct 19 19:33:49 can i just send it to nokia or do i have to go through a retailer Oct 19 19:33:53 when did you buy it? Oct 19 19:34:18 if under 2 years ago, possibly. where are you? Oct 19 19:34:22 sweden Oct 19 19:34:44 and you aren't the person who bought a n900 off eBay a few hours agowi no USB? Oct 19 19:34:46 yeah i bought it second hand this spring, was a christmast present that didnt come to use Oct 19 19:34:47 with Oct 19 19:34:50 ah Oct 19 19:34:56 -t Oct 19 19:35:14 I mean, if it was originally purchased under 2 years ago Oct 19 19:35:23 no i have had since then but the usb connector broke instantly and i got a bit depressed over it Oct 19 19:35:33 yeah it was probably bought in december 2011 Oct 19 19:35:35 or very close to it Oct 19 19:35:36 you will not get a n900 back, even if yo can. Oct 19 19:35:45 oh Oct 19 19:35:48 but i want this one .... Oct 19 19:35:50 :) Oct 19 19:35:54 better get someone else to fix it though Oct 19 19:35:57 *then Oct 19 19:36:14 last year this time, people were getting other phones Oct 19 19:36:16 my nerves are not good enough for the precision soldering of that connector Oct 19 19:36:33 ah, aight... wont send it to nokia then Oct 19 19:37:56 i have an external charger and an extra battery, so it's usable Oct 19 19:38:29 and i really like what i see... but i have to reset the clock each night to charge the battery, and can't use usb for file transfer/modem Oct 19 19:42:19 it is annoying, yes. Oct 19 19:42:28 damn Oct 19 19:42:35 now I'm a bit annoyed Oct 19 19:42:51 I want to change vfat partition to ext3 Oct 19 19:43:13 and there is no good tutorial... Oct 19 19:43:19 http://wiki.maemo.org/Repartitioning_the_flash Oct 19 19:43:32 this one is nice, gives 8 ways, none of them are really usable Oct 19 19:43:35 for my case Oct 19 19:43:38 or I'm dumb Oct 19 19:46:01 from memory, I tarred up everything g from mydocs to home, deleted mydocs, made a new 4g partition with same ids, and fat, then a ext3 partition filling the rest of the space Oct 19 19:46:30 then a hacky script, run through 'queen beecon' to mount it when the ui came up Oct 19 19:59:59 has anyone seen a thread on the maemo boards that mentions any members that might have voulentereed to fix broken usb connectors? for payment of course Oct 19 20:00:19 i see a thread that mentions people that can help with pre-fail fixes but mine is already loose Oct 19 20:01:36 http://talk.maemo.org/showpost.php?p=436216&postcount=2 Oct 19 20:01:38 postfail depends on whether the connector fell off, or if the part of circuitboard it was attached to tore off Oct 19 20:01:43 so I can't change my vfat partition to ext3? Oct 19 20:01:53 or can I? Oct 19 20:02:06 ShadowJK: the connector fell off, the pads are still there Oct 19 20:03:35 i tried soldering it back and it worked for a while but fell of again Oct 19 20:03:48 so i need to get hold of someone with better precision soldering skills Oct 19 20:23:38 raccoon_: bit more solder and enough heat, you can do it :) Oct 19 20:28:51 raccoon_: any good repair shop will fix it Oct 19 20:29:37 probably even those "frenchise" " Nokia (et al)" phoneshops Oct 19 20:30:24 raccoon_: where you're located? Oct 19 20:31:22 i think he said sweden Oct 19 20:31:23 DocScrutinizer05: sweden Oct 19 20:31:26 yup Oct 19 20:31:50 Skry: yeah i think i'll begin with ordering a few of those ports and maybe at least give it one more shot Oct 19 20:32:05 in Sweden there should be similar 'free' phoneshops like here in Germany Oct 19 20:32:11 yeah theres a few Oct 19 20:32:27 i recently fixed my macbook lcd at a place, could go there and ask about the phone Oct 19 20:33:01 I honestly suggest you either let some professional do the job, or you find a few old 5EUR mobile phones to practice on them Oct 19 20:34:15 one advice nevertheless: for the huge solder blobs at left and right of USB you need a rather heavy iron, some 50, maybe 70W Oct 19 20:34:17 I've seen a "Elgigantten" shop in Finland that had oscilloscopes and irons in their repair section, that looked confidence inspiring enough Oct 19 20:34:28 * RST38h suddenly notices that he has never yet sacrificed a racoon to the Tentacled One Oct 19 20:34:31 the ground planes are huge and take a lot of heat away Oct 19 20:34:37 Hm Oct 19 20:35:09 for the contacst you need mor like a 5..7W needle Oct 19 20:35:15 contacs Oct 19 20:35:19 meh Oct 19 20:36:17 ShadowJK: sounds good, yeah Oct 19 20:37:00 yeah, Gigantti, like the last place anyone should ever take anything :) Oct 19 20:40:38 What is it? A Nordic Radioshack? Oct 19 20:42:28 ShadowJK: yeah we have elgiganten here in town too, could ask the mtoo Oct 19 20:42:31 *them too Oct 19 20:42:37 thanks for all the advice people Oct 19 20:44:55 Skry, it all depends on finding the one guy that can operate that repair equipment. I've almost never seen anyone use that equipment though Oct 19 20:47:18 ShadowJK: yeah, at least in this town, those guys over there probably have never even seen an iron before, and generally try to avoid customers if in any way possible. Oct 19 20:47:55 ooh, they found their final professional destiny then Oct 19 20:50:23 is there any skype client or only that integrated thing? Oct 19 20:50:48 integrated thing Oct 19 21:18:50 ~seen marxian Oct 19 21:18:55 i haven't seen 'marxian', vi__ Oct 19 21:19:11 ~seen your peen Oct 19 21:19:12 vi__: i haven't seen 'your peen' Oct 19 21:19:24 ~seen marxian Oct 19 21:19:25 vi__: i haven't seen 'marxian' Oct 19 21:19:32 wtf Oct 19 21:19:41 why the first is different? Oct 19 21:44:53 huh? Oct 19 21:44:58 what's different? Oct 19 21:45:20 aah, the first Oct 19 21:45:31 well, she's eloquent Oct 19 21:45:43 ~xyawn Oct 19 21:45:44 good coffee Oct 19 21:45:48 ~xyawn Oct 19 21:45:50 methinks xyawn is big coffee Oct 19 21:45:55 ~xyawn Oct 19 21:45:55 rumour has it, xyawn is strong coffee Oct 19 21:46:55 ~lart vi__ Oct 19 21:46:56 * infobot takes out a cattle prod and gives vi__ a good jolt Oct 19 21:47:07 ~lart vi__ Oct 19 21:47:08 * infobot offers vi__ some herring Oct 19 22:08:10 Off for the weekend... have fun everyone. Oct 19 22:09:28 congrats to board, and also to woody ;-D Oct 19 22:36:16 useless statistic of the day: full charge equals roughly about 10h as distcc master (through wlan, -j 6) Oct 19 22:36:54 yowie. Oct 19 22:37:22 that's three n900s in a row on eBay with no USB port gone at over 50 quid Oct 19 22:57:50 ouffff Oct 20 00:08:58 hi! i wonder how to style/theme a GtkButton and GtkToggleButton. they just appear flat when all the other apps have a 3d-ish effect to them. i have found set_style() but it does not do anything... Oct 20 00:15:19 I have no clue, really no clue, but my guess would be that there are some special hildon equivalents of GtkButton and GtkToggleButton that gives you the hildon styling Oct 20 00:18:44 and I'm saying that mostly because when you run pure gtk/gnome apps on maemo, they look pretty unstyled Oct 20 00:29:19 ShadowJK, maybe everything else is qt? Oct 20 00:29:31 No. Oct 20 00:29:34 ok Oct 20 00:29:46 Most things in Maemo5 is Hildon Oct 20 00:29:49 i tried HildonButton without any luck :( Oct 20 00:29:55 ah yes of course Oct 20 00:30:02 The time/date setting thing at boot is clutter :) Oct 20 00:31:05 ok Oct 20 01:52:37 My N900 always need to set time and date after restart phone. How to solve this problem ? Oct 20 01:52:58 It's almost impossible. Oct 20 01:53:35 You'd need to replace the backup battery, but that requires a full disassembly and expert soldering skills Oct 20 01:55:34 or software Oct 20 01:55:45 ShadowJK: Is it normal for N900 that the time will be changed after change battery ? Oct 20 01:55:54 netdate first Oct 20 01:56:13 stardiviner: no. however, once the backup battery fails, yes Oct 20 01:56:21 SpeedEvil: I know a program "Sync Clock Now", but it can now set time correctly in my case. Oct 20 01:56:23 and it tends to fail early Oct 20 01:56:47 does it set it from network, or cell operator? Oct 20 01:56:59 Oh .... It's really hateful. I almost change battery once a day. Oct 20 01:57:11 openntpdate I think was the package I used Oct 20 01:57:13 SpeedEvil: from wifi Oct 20 01:57:20 I change battery once a year, and reboot once every 4 months :/ Oct 20 01:57:41 you tell it a NTP server, and leave it Oct 20 01:58:07 My operator does send me time updates, so even if just tap away the time screen, the correct time arrives when I connect to 3g data :/ Oct 20 01:58:14 ShadowJK: no, I mean, I have two batteries, I shift them, after one is used up. and another is charged. Oct 20 01:58:58 SpeedEvil: I have not tried that openntpdate package, I should try later. Oct 20 02:02:29 Yeah there's only been a few occasions over the years when I've ran out of power Oct 20 02:43:35 is there anyway i can downgrade cssu testing camera, its too slow and acts funky **** ENDING LOGGING AT Sat Oct 20 03:00:02 2012