**** BEGIN LOGGING AT Wed Oct 23 02:59:58 2013 Oct 23 03:24:53 exactly what all things do the proximity sensor do on n900 other than screen lock/off function while calling? Oct 23 03:36:04 good question Oct 23 03:37:19 but not a precise question, since it's not the sensor doing things, it's things using the sensor. Thus it's not that easy to tell which things are using the sensor since you can't tell from "looking at the sensor" Oct 23 03:39:42 quite a number of apps use the proximity sensor for various purposes, starting at usecase where it's just an additional HID to trigger app-specific actions, by covering it with your finger, to stuff like detecting if device is in a pouch or bag or pocket Oct 23 03:40:58 I dunno if in some apps e.g. speakers are muted when proxy detects possibly your ear next to the speaker Oct 23 03:43:39 not in apps by dedault? Oct 23 03:50:47 by default? what do you mean by default? Oct 23 03:51:03 means without any extra apps Oct 23 03:51:06 the dialer app is known to use the proximity sensor Oct 23 03:51:27 the sensor itself is just that: a sensor. It can't DO anything Oct 23 03:51:59 so only the dialer app uses proximity? Oct 23 03:52:02 see my initial comment above Oct 23 03:52:26 as in witthout installing any extra app.. Oct 23 03:53:41 so the thing is i want to disable it, I dont want the screen to be off while calling. Oct 23 03:53:55 is there any method to accomplish it Oct 23 03:54:13 nobody has done such investigation yet - it's also vastly useless since no device out there has no additional apps installed Oct 23 03:55:11 no, this can't work, since when the screen doesn't get locked, your ear will engage some buttons on the dialer GUI Oct 23 03:55:35 yeah I know, but still.. Oct 23 03:55:55 I also don't see the rationale behind such behavior modification of dialer Oct 23 03:56:17 okay.. my sensor is not sensitive. Oct 23 03:56:36 thus I won't waste processor time of my wetware to come up with a possible solution Oct 23 03:57:27 specify a valid usecase and maybe we can suggest a better solution Oct 23 03:57:37 and it sees as if it is covered by something in low light conditions Oct 23 03:58:39 umm, you say your sensor is defect? Oct 23 03:58:58 yes, its not working as it should Oct 23 04:01:21 check: `` watch --interval=1 cat /sys/devices/platform/gpio-switch/proximity/state ´´ Oct 23 04:02:35 it should say "open" until you cover it with your finger where it immediately and without glitches shall change to "closed" Oct 23 04:04:48 also check your general system status, with htop Oct 23 04:05:08 well, that command Oct 23 04:05:25 proximity sensor (like other events) known to be late or get lost completely when CPU load is way too high Oct 23 04:06:04 its always closed. Oct 23 04:06:10 that's bad Oct 23 04:06:22 do you have a screen protector? Oct 23 04:06:43 does it cover the prox sensor? Oct 23 04:06:44 already removed Oct 23 04:07:27 there might be dust or other debris in the prox sensor aperture Oct 23 04:07:49 or the hardware component (the chip) is defect Oct 23 04:08:22 so your problem is: you never see screen during calls? Oct 23 04:08:50 i have to slide open to see whose calling Oct 23 04:08:56 I see Oct 23 04:09:01 annoying Oct 23 04:09:10 yes Oct 23 04:09:27 obvious solution: fix prox Oct 23 04:09:47 spare flex PCB are available for <10 bucks Oct 23 04:10:24 for a stopgap you could try to deactivate the prox sensor all together Oct 23 04:10:43 how? Oct 23 04:10:45 you need to mess with gpio sysnodes to do that Oct 23 04:11:16 or maybe a bindmount of a static real file over the /sys/devices/platform/gpio-switch/proximity/state node could work as well Oct 23 04:11:17 oops Oct 23 04:12:20 anything that i can do? Oct 23 04:12:56 sth along the line `` echo open >/prox; mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state ´´ Oct 23 04:14:56 yeah, seems to work, just tested it Oct 23 04:16:31 do this, and then you can control the prox "sensor" state with `` echo open >/prox ´´ and `` echo closed >/prox ´´ Oct 23 04:17:07 do i have to be root? Oct 23 04:17:10 yes Oct 23 04:17:58 prox sensor is owned by root, as is sysfs which you try to mess with (mount) Oct 23 04:18:43 and even / is writable by root only, for the /prox file Oct 23 04:18:58 will you see any result in that command. Oct 23 04:18:59 you can change this to your liking Oct 23 04:19:05 no Oct 23 04:19:11 ok Oct 23 04:19:21 not if it succeeds Oct 23 04:19:56 you however see the difference whe you do `` cat /sys/devices/platform/gpio-switch/proximity/state ´´ Oct 23 04:20:48 well its now open Oct 23 04:20:52 it now shall print exactly what you piped to /prox before, no matter if that's "open" or 2closed" or "foobar" Oct 23 04:22:01 well, screen doenot go black when calling now. Oct 23 04:22:05 since dialer is doing exactly same, you now should see dialer act like when sensor is "open" Oct 23 04:22:50 (maybe you need to killall dialer, to make it drop any open filehandle that still would read from the original /sys/devices/platform/gpio-switch/proximity/state) Oct 23 04:23:29 hmm, nope, prolly not Oct 23 04:23:40 its working just like I wanted. Oct 23 04:24:00 will it revert back if rebooted? Oct 23 04:24:05 yes Oct 23 04:24:21 hmmm.. Oct 23 04:24:23 you need to repeat the mount -o bind during boot Oct 23 04:25:00 GOD I*love* this device!! Oct 23 04:25:17 where from you know all these things. Oct 23 04:25:23 you born like this? Oct 23 04:25:26 decades of linux Oct 23 04:25:34 basically yes ;-) Oct 23 04:26:18 thanks a lot doc, now I have to hard copy that commands,.. :) Oct 23 04:26:26 yw Oct 23 04:26:28 :-) Oct 23 04:26:57 wait a minute, i'll help you some more Oct 23 04:27:19 how? Oct 23 04:29:42 sed -i 's/exit 0/mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state\nexit 0' /etc/init.d/rcS Oct 23 04:29:58 warning, no waranties included, it's untested Oct 23 04:30:44 damn sorry, bad "typo" Oct 23 04:30:59 warning, no waranties included, it's untestederr nope Oct 23 04:31:09 no tiypo Oct 23 04:31:53 what is that? Oct 23 04:31:53 do a `` tail /etc/init.d/rcS ´´ after that, and check if the last two lines look like Oct 23 04:32:04 mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state Oct 23 04:32:11 exit 0 Oct 23 04:32:40 it's a command to edit the rcS sysinit file Oct 23 04:32:53 do automatically do this mount command during boot Oct 23 04:32:58 shoul i start from sed -i? Oct 23 04:33:11 sed -i 's/exit 0/mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state\nexit 0' /etc/init.d/rcS Oct 23 04:33:14 as root Oct 23 04:33:39 then Oct 23 04:33:50 tail /etc/init.d/rcS Oct 23 04:33:55 and check last 2 lines Oct 23 04:34:36 dang!!! Oct 23 04:34:42 sorry huge typo :-D Oct 23 04:34:51 ??? Oct 23 04:35:15 sed -i 's@exit 0@mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state\nexit 0@' /etc/init.d/rcS Oct 23 04:36:53 c&p this line to a root terminal on N900 Oct 23 04:37:21 :-) Oct 23 04:38:17 should do the trick. I hope you won't encounter a bootloop for completely obscure reasons Oct 23 04:38:30 messing with rcS file is a tad "dangerous" Oct 23 04:38:31 bad option i substitution expression Oct 23 04:38:48 o.O Oct 23 04:38:51 wtf? Oct 23 04:39:06 prolly busybox again Oct 23 04:40:23 sed: bad option in substition expression Oct 23 04:42:06 you didn't copy the `` sed -i 's@exit 0@mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state\nexit 0@' /etc/init.d/rcS ´´ cmdline Oct 23 04:42:35 the first one I posted been incorrect and spits that error Oct 23 04:42:55 use Oct 23 04:42:56 sed -i 's@exit 0@mount -o bind /prox /sys/devices/platform/gpio-switch/proximity/state\nexit 0@' /etc/init.d/rcS Oct 23 04:43:03 I just tested it, it works Oct 23 04:44:46 don't run the command more than once! it won't print any "ok" when it succeeds Oct 23 04:46:42 http://privatepaste.com/c8c39a90b1 Oct 23 04:47:04 I did a `` cd /etc/init.d ´´ before Oct 23 04:47:49 when using `` tail /etc/init.d/rcS ´´ then you don't need to worry about the cd Oct 23 04:49:18 I nevertheless suggest you fix your hardware Oct 23 04:49:56 probably the proximity sensor chip is dead. Not unheard of Oct 23 04:50:11 it succeeded atlast Oct 23 04:50:59 :-) Oct 23 04:51:02 http://www.ebay.com/bhp/nokia-n900-flex Oct 23 04:51:15 get that ans swap Oct 23 04:51:20 and* Oct 23 04:52:16 you see the right one, you can tell from most offers using same picture Oct 23 04:52:25 cheapest offer <5 bucks Oct 23 04:52:49 I will be using like this for a couple of days. and then I will buy one. Oct 23 04:52:57 it includes camera? Oct 23 04:53:51 yes Oct 23 04:53:56 front cam Oct 23 04:54:35 a real bargain for 4.98USD Oct 23 04:54:55 that one has a fake camera Oct 23 04:55:09 ohh? Oct 23 04:55:16 losing front cam for proximtiy... :) Oct 23 04:55:45 \which one would you choose if you are given such a choice Oct 23 04:55:49 how do you know it's fake? Oct 23 04:56:24 its heading is BRAND NEW FAKE CAMERA FLAT FLEX CABLE RIBBON REPLACEMENT FOR NOKIA N900 c57A... the 4.98$ item Oct 23 04:56:32 * DocScrutinizer05 glares at his 2 flex cable spares he got for 7bucks each Oct 23 04:56:47 LOL Oct 23 04:57:25 I got me 3 of those: http://www.ebay.com/itm/New-Ribbon-Flex-Cable-with-Camera-for-Nokia-N900-/290704781415 Oct 23 04:57:41 not tested yet, but look fine Oct 23 04:58:10 took like 6 weeks for shipping Oct 23 04:58:39 and came with their antistatic bags packed separately ;-P Oct 23 04:59:20 with the flex in antistatic bag they couldn't fold the bags to fit into the package - idiots Oct 23 05:00:29 so they opened the bags, removed the cables, and placed the 3 folded bags into the polystyrene box, then bubble foil, then the 3 cables Oct 23 05:00:47 X-P Oct 23 05:33:22 * DocScrutinizer05 hopes SaiF hasn't completely messed up his rcS with lots of 'mount -o bind foobar' lines that could even throw error and possibly make boot go looping Oct 23 06:10:15 DocScrutinizer05: scraping knock detection, just gonna do the button thingie Oct 23 06:10:27 :-) Oct 23 06:10:28 DocScrutinizer05: I can react to the headset button being pressed with python right? Oct 23 06:10:40 yep, should be possible Oct 23 06:14:56 http://talk.maemo.org/showthread.php?t=21182 Oct 23 06:16:32 cool Oct 23 06:20:29 DocScrutinizer05: 600 mW power er.... handling should be sufficient? Oct 23 06:20:36 for the resistor in parallel with the button Oct 23 06:20:44 sure Oct 23 06:21:06 P=U^2/R Oct 23 06:21:13 U=2V max Oct 23 06:21:28 yeah okay Oct 23 06:25:57 you got the circuit right now? switch in parallel to ~1k resistor from 3rd contact of plug to 4th contact (aka sleeve, ground) Oct 23 06:26:11 nah, gonna do shopping later today Oct 23 06:26:19 ok, same applies Oct 23 06:26:21 you said 2k resistor earlier Oct 23 06:26:21 :P Oct 23 06:26:36 I honestly don't know how large the resistor should be Oct 23 06:26:45 since usually that's a mic Oct 23 06:26:57 wait, I'll try to probe it Oct 23 06:32:43 1..2K is ok Oct 23 06:32:49 1.4 actually Oct 23 06:33:01 okay Oct 23 06:33:03 1k should do as well Oct 23 06:35:51 I'll try to calculate the correct value based on circuitry in N900 Oct 23 06:37:22 I can probably take a picture (without engaging the flash) and e-mail it as an attachment too? Oct 23 06:40:26 hmm? Oct 23 06:40:35 picture of whatß Oct 23 06:40:39 ? Oct 23 06:41:53 uh, with the camera Oct 23 06:42:00 the person ringing the doorbell :D Oct 23 06:42:12 btw one of the two schmidt-triggers is built around a 4.7k series resistor to mic, from 2.5V, and the trigger level is Oct 23 06:42:37 ~2.5 / (22 + 6.8) * 22 Oct 23 06:42:38 1.909722222222 Oct 23 06:42:41 volt Oct 23 06:43:17 sure you can take a picture Oct 23 06:44:27 or - rather simple, the resistor value for triggering that threshold is: Oct 23 06:44:46 ~22 / 6.8 * 4.7 Oct 23 06:44:47 15.205882352941 Oct 23 06:44:50 kR Oct 23 06:45:14 that's prolly to detect AV-in connected Oct 23 06:46:29 then there's a second schmidt-trigger that is Oct 23 06:47:26 with a 220R series resistor for mic Oct 23 06:47:49 form an unknown but probably 2V micbias voltage Oct 23 06:48:16 (unknown since I'm too lazy to read the tlv320aic34 datasheet) Oct 23 06:48:40 the voltage level for that trigger is Oct 23 06:48:53 ~2.5 / (22 + 6.8) * 6.8 Oct 23 06:48:54 0.590277777778 Oct 23 06:50:10 given the 220R series this will be ~50R max, and that's probably the max ESR when switch closed Oct 23 06:50:45 ooops, sorry, there's another 2k series in that one Oct 23 06:52:22 so this is basically same as above, just inverted Oct 23 06:52:45 ~2.5 / (22 + 6.8) * 6.8 Oct 23 06:52:45 0.590277777778 Oct 23 06:53:07 err yeah, and Oct 23 06:55:27 ~500R series Oct 23 06:58:21 uh huh Oct 23 06:58:27 * bindi is confused Oct 23 06:58:38 am I still good to go with the 1k ohm resistor or? :D Oct 23 07:00:10 prolly yes Oct 23 07:00:47 http://www.gsma.com/newsroom/omtp-documents-1-0-omtp-local-connectivity-wired-analogue-audio-v1-0 says >LC-0070<: Oct 23 07:01:19 The electrical characteristics of the audio Terminal of the Oct 23 07:01:20 Terminal SHALL operate within the values stated below: Oct 23 07:01:22 Bias Current Oct 23 07:01:23 Min 100 μA Max 500 μA Oct 23 07:01:43 (Voltage of 2.1V generated over microphone component with Oct 23 07:01:45 serial pull up resistor value between 1.90 kΩ and 2.2 kΩ) Oct 23 07:02:32 this is pretty weird spec Oct 23 07:03:23 anyway it gives some hint on ESR of microphone Oct 23 07:03:54 ~2.1/0.0005 Oct 23 07:03:54 4200 Oct 23 07:04:12 minimum R Oct 23 07:04:22 ~2.1/0.0001 Oct 23 07:04:23 21000 Oct 23 07:04:27 max R Oct 23 07:04:59 according to the OMTP specs for headsets Oct 23 07:05:50 I suggest you simply test it Oct 23 07:06:15 what are you guys fiddling with? Oct 23 07:06:19 add 1k, 1.4k, 2k, 4k resistors to my shopping list? Oct 23 07:06:58 sono: custom button triggering an action on the n900 via headset socket Oct 23 07:07:13 ah Oct 23 07:07:33 use CA-75U cable. See what resistor it needs on yellow plug to make the device display headset connected indicator Oct 23 07:08:00 seems you would be fine with a 5k Oct 23 07:08:12 hm, that one comes with the n900? i have one somewhere Oct 23 07:08:14 but I doubt that applies in real life Oct 23 07:08:30 I'd bet a few bucks on 2k just working fine Oct 23 07:12:36 anyway, I just probed a 1400R on a nokia headset, with a good low-voltage digital multimeter Oct 23 07:13:11 thus a voltage of significantly < 2V Oct 23 07:14:04 due to the non-linear characteristics of semiconductors this effective series resistance coukd only get lower on higher voltage, i'd guess Oct 23 07:15:17 in the end it's probably completely irrelevant if your resistor is 500R or 18kR Oct 23 07:16:03 which makes sense since in real life the properties of arbitrary headset mics may vary significantly Oct 23 07:16:53 so as long as 0<< R < a closed switch plus cable and plug parasitic series resistance should still be <<100R, and a CA-75U connected to a CVBS-in is either >10k or ==50R Oct 23 07:19:40 I'll check this in a minute Oct 23 07:21:27 dang! 75R, I could have known Oct 23 07:22:36 this stuff is really puzzling and weird. The general situation as well as the N900 circuitry around it Oct 23 07:23:58 so your resistance between the two contacts when ring button pushed should be <<75R, and the resistance when button not pressen should be some kR Oct 23 07:24:12 uh Oct 23 07:24:28 so that button with resistor in parallel isnt going to work or? Oct 23 07:25:19 it will work, definitely Oct 23 07:25:49 use a normal switch, a 2kR and when this isn't the right components then I owe you a beer Oct 23 07:26:18 ;-) Oct 23 07:26:20 we'll see later today then :P Oct 23 07:31:57 how feasible would it be to record a video with 30s buffer and save 5 last seconds of the clip and e-mail it when the button is pressed? Oct 23 07:32:08 in case person ringing gets to a blind spot or blocks the camera Oct 23 07:32:10 btw on GPIO_182 aka McSPI2_CS1 aka pin OMAP3430:V3 you got the raw voltage from mic_in, labeled "ECI_AD" which suggests this might be either a serial-in for one-wire ECI (multibutton headsets) or even a A/D converter when you mux the V3 pin to another internal function block Oct 23 07:32:38 always record a video with max. 30s of footage that is Oct 23 07:33:07 should be feasible Oct 23 08:41:28 A wonderful fine Sweetmorn (UGT)! Oct 23 09:15:49 morning Oct 23 09:16:50 Hi Oct 23 09:18:25 DocScrutinizer05: do we have complete schematics for the n900? Oct 23 09:18:41 i still have had too little time to really read up, excuse the easily googled questions Oct 23 09:18:45 ~schematics Oct 23 09:18:45 it has been said that schematics is http://wiki.maemo.org/N900_Hardware_Schematic Oct 23 09:18:48 woot Oct 23 09:20:21 that pahe seems kinda garbled and outdated Oct 23 09:20:34 you find them via google easily though Oct 23 09:20:47 nod Oct 23 09:21:02 fascinating.. schematics for my phone o.o Oct 23 09:21:08 i don't think i ever had that before Oct 23 09:21:51 http://www.google.de/search?q=Nokia_N900_RX-51_Schematics.pdf Oct 23 09:22:27 yea. thanks. Oct 23 09:22:30 yw Oct 23 09:22:33 To login in (wiki.)maemo.org i need account in garage.maremo.org, right? Oct 23 09:22:36 * sono stares at mixed tabs and spaces in a .c file Oct 23 09:22:51 err, it's "SSO", yeah Oct 23 09:23:35 using tabs is pretty much deprecated in sourcecode Oct 23 09:24:10 smart editors can convert tab into spaces, and vice versa Oct 23 09:24:37 are you talking to me? Oct 23 09:24:40 and you should set the editor so it inserts 4 spces for each tab you type Oct 23 09:24:43 i was just expressing my hatered Oct 23 09:24:48 no need to lecture me on fucking tabs Oct 23 09:24:51 hatred* Oct 23 09:24:54 yeah, sure Oct 23 09:25:16 in fact, Oct 23 09:25:30 * sono fires up astyle and prepares the mass commit cannon Oct 23 09:25:36 fuck that noise =) Oct 23 09:25:51 hehe Oct 23 09:28:46 tabs are bad.. mixed is .. blyyarghg Oct 23 09:29:29 tab space tab space space tab ;-P Oct 23 09:29:45 yeah, i cry blood and run an indenter when i see that Oct 23 09:29:56 Unchanged /home/benjamin/src/XLU-Freescale/src/task.c Oct 23 09:30:01 "hmm..?" Oct 23 09:30:06 benjamin@ubuntu-dev:~/src/XLU-Freescale/src$ file task.c Oct 23 09:30:06 task.c: empty Oct 23 09:30:16 * sono opens the window Oct 23 09:30:18 * sono jumps Oct 23 09:30:25 DAT CODE Oct 23 09:30:51 great task ! XP Oct 23 09:31:28 >>file task<< >task is empty<< Oct 23 09:31:31 that's epic Oct 23 09:31:42 there is some epic shit in our repos Oct 23 09:32:09 void dosendparameter( std::ostream & ostr, int Parameter, char laenge, char schalter ) Oct 23 09:32:12 { Oct 23 09:32:14 static char text[20]; Oct 23 09:32:17 if ( schalter == true ) Oct 23 09:32:19 { Oct 23 09:32:22 ostr.put( 0x1B ); //ESC = 0x1B Oct 23 09:32:24 ostr.put( 'Z' ); Oct 23 09:32:27 ostr.put( 'O' ); Oct 23 09:32:29 ostr.put( 'K' ); Oct 23 09:32:32 } Oct 23 09:32:34 itoa( Parameter, text, 10 ); //Wandeln in ASCII Oct 23 09:32:37 strrev( text ); //Wenden Oct 23 09:32:39 if ( Parameter < 0 ) Oct 23 09:32:42 { Oct 23 09:32:44 for ( int x = strlen( text ) - 1;x < laenge - 1;x++ )text[x] = '0'; //Führende Nullen setzen Oct 23 09:32:47 if ( laenge > 1 )text[laenge-1] = '-'; Oct 23 09:32:50 } Oct 23 09:37:57 sorry, no idea what i was thinking Oct 23 09:38:06 i suppose i just had to share the pain Oct 23 09:38:11 np, just don't do it again please Oct 23 09:38:19 not like i don't know better Oct 23 09:38:54 (you have to admit that is probably among the most terrible code you ever saw, tho) Oct 23 09:39:02 yes, indeed Oct 23 09:39:06 (that code *is* alive) Oct 23 09:39:19 just not my project (phew) Oct 23 09:39:21 makes me want to have breakfast, so I can throw up a bit Oct 23 09:39:23 ... function names, variables in non-english? Oct 23 09:39:25 yea Oct 23 09:39:41 jaska: not just that. they're also completely nondescriptive even if you do speak the language Oct 23 09:39:49 "schalter" is just "toggle" Oct 23 09:40:06 or switch, if you wish. Oct 23 09:40:24 that whole code is a single abomination Oct 23 09:40:45 there's hardly a line in that function that is not. Oct 23 09:40:46 strrev( text ); LOLWUT?!! Oct 23 09:40:53 ahha yes. Oct 23 09:43:11 and the missing spaces after ) of a condition Oct 23 09:43:16 great style Oct 23 09:45:00 but spaces everywhere else almost Oct 23 09:45:10 albeit inconsistently Oct 23 09:45:30 the algorithm is the best part Oct 23 09:45:40 if you want to call it that Oct 23 09:46:11 i'm just glad i don't have to deal with that project =) the guy who does, tho... well he kinda deserves it i guess Oct 23 09:46:17 he must Oct 23 09:46:20 devel must've been on crack Oct 23 09:46:25 otherwise it would be cruel, and unusual Oct 23 09:46:44 he's long left the company. i am told there were two guys Oct 23 09:47:14 both fresh out of a special vocational school (not uni!) where they learned to code from 50 year old teachers that learned to code from a textbook Oct 23 09:47:22 they had no mentoring, nothing Oct 23 09:47:37 in a way, it is surprising that the project does anything at all Oct 23 09:47:48 that doesnt mean anything, nobody ever mentored me, i never even finished highschool. Oct 23 09:47:51 <.< Oct 23 09:47:53 well yes Oct 23 09:47:56 same here Oct 23 09:47:57 still, ... Oct 23 09:48:17 you did not learn how to program in a vocational school Oct 23 09:48:20 i did write some ugly code early on. Oct 23 09:48:20 they did. Oct 23 09:48:25 certainly Oct 23 09:51:54 you either learn coding by studying and understanding good code, or you never learn it Oct 23 10:16:03 highly recommended read: http://www.thoughtcrime.org/blog/saudi-surveillance/ (semi-OT) Oct 23 10:38:38 intresting that they didn't FinFisher Oct 23 10:38:51 s/didn't/didn't turn to/ Oct 23 10:38:53 l4m3rx meant: intresting that they didn't turn to FinFisher Oct 23 10:39:48 brolin_empey: now looking at kernel-power Oct 23 10:40:34 brolin_empey: I have 4 patches: dm9601, asix, nokia-av and power_supply_sysfs Oct 23 10:41:31 dm9601 and asix only adding new usb ids... this will not break/change anything existing (only adding support for new hw) Oct 23 10:41:47 so somebody with that hw will need to check if work Oct 23 10:41:54 for me that two patches are ok Oct 23 10:42:25 next power_supply_sysfs patch which fixing dangling if statement Oct 23 10:43:45 and patch for nokia-av adding new (better) support for headset button. it reports headset button event immediately, but sometimes not correctly - so it report event as KEY_PROG1 (to not break any app which does not listen for that key) Oct 23 10:44:03 only this last patch needs more testing, all others are ok Oct 23 10:44:33 brolin_empey: I do not remember which patch is from you, so look at list ^^^ if there is that yours Oct 23 10:45:27 I will add info about testing new kernel packages to TMO kp52 thread Oct 23 10:59:18 Pali: Thank you. The dm9601 and asix patches are mine; I made and submitted them because I need them in order to use some models of USB ↔ Ethernet adapters I have without having to use a hexadecimal editor with the binary .ko files to manually replace the numeric identifiers of a model of adapter I do not have with the numeric identifiers of a model of adapter I do have. Oct 23 11:00:03 rigth, so you will test if that working with your HW Oct 23 11:03:07 Pali: Yes, I will test and report after installing the kp52. Oct 23 11:06:50 brolin_empey: now updating my phone to new version from git Oct 23 11:07:16 after some tests (in next minutes) I will publish deb packages for testing Oct 23 11:22:34 new kp for testing is here: http://talk.maemo.org/showthread.php?p=1381948#post1381948 Oct 23 11:22:36 brolin_empey ^^^^ Oct 23 11:22:53 freemangordon ^^^^ Oct 23 11:29:53 Pali: how much slack space does your uboot still have for the kernel? Oct 23 11:31:09 jon_y: max kernel size is 1832960 Oct 23 11:31:20 and max uboot size is 262144 Oct 23 11:32:04 (if you mean combined uboot+kernel image in n900 nand) Oct 23 11:32:11 ok Oct 23 11:32:42 also, I notice sometimes kp starts with console Oct 23 11:32:46 other times without Oct 23 11:33:13 not sure if I messed with the kernel by adding the console Oct 23 11:56:29 o.O what a gibberish!! a:3:{i:0;s:11:"programming";i:1;s:17:"test_et_bugreport";i:2;s:9:"ui_design";} Oct 23 11:57:03 lools almost like dbus parameters Oct 23 11:57:10 looks, even Oct 23 11:57:44 however it's most probably PHP, or some weird stuff midgard invented Oct 23 11:58:17 o.o looks like an unholy union between awk and c64 basic Oct 23 11:59:21 ;-) Oct 23 11:59:21 Ack. Oct 23 11:59:39 that's my "interests" as in http://maemo.org/profile/view/joerg_rw/ , as seen in administrative interface Oct 23 11:59:45 it looks like something php devs could come up with, yes, but i know a bit of php and never seen there any format like that Oct 23 12:00:27 looks sorta like bittorrent dht protocol.. distantly Oct 23 12:00:36 like, strings are s:length:"foobar" Oct 23 12:02:09 it could not be some sort of json? Oct 23 12:02:18 actually, yes. Oct 23 12:02:26 that's for sure not a valid json Oct 23 12:02:27 i haven't done web stuff lately, but Oct 23 12:02:35 well, not valid.. Oct 23 12:02:42 well, i have no idea.. Oct 23 12:03:36 and the only thing it shares with json are there three characters {}: Oct 23 12:03:57 these* Oct 23 12:04:13 some custom internal format i guess Oct 23 12:04:31 so let's face it: it's MIDGARD ;-P Oct 23 12:05:06 net.nehmer.account Oct 23 12:05:58 hi dos1 :-) Oct 23 12:06:06 hello! Oct 23 12:06:16 back to proper PC? Oct 23 12:06:31 yup Oct 23 12:07:03 I think THE page needs some improvements that cpuld predate THE announcement ;-) Oct 23 12:07:13 could* Oct 23 12:08:15 and to answer your question (7weeks?), see last post in Oct 23 12:08:20 http://talk.maemo.org/showthread.php?t=91142 Oct 23 12:08:22 DocScrutinizer05, it is php Oct 23 12:08:30 ooh, it IS? Oct 23 12:08:33 DocScrutinizer05, same format as php uses for storing session ata Oct 23 12:08:42 :nod: Oct 23 12:08:56 oh Oct 23 12:09:03 see http://php.net/manual/en/function.serialize.php Oct 23 12:09:49 thabks for the pointer, luckily I don't need to learn about it, just ran into it on midgard, while checking progress of karma updating Oct 23 12:10:08 what, that is a serialization format? Oct 23 12:10:28 bloody hell Oct 23 12:10:43 sono: that is php serialization format - adding "php" before it changes everything :D Oct 23 12:10:53 i suppose so Oct 23 12:11:37 DocScrutinizer05: some list of short bullet points about what could be added/changed on the website would be helpful Oct 23 12:11:39 looks like they were trying to make something like json, misnamed it serialization, and then went ahead and just built something Oct 23 12:12:15 sono: php learned how to deal with json in its standard library quite recently Oct 23 12:12:18 dos1: I think we can set up a sneaky wiki page, in some dark corner Oct 23 12:12:26 :nod: Oct 23 12:13:34 for now main concern is clear emphasis on distinction between the hw device and any OS that will not ship with it Oct 23 12:15:08 DocScrutinizer05: btw. I've just recieved second gta02, so my concerns about distinction between daily and devel devices regarding gta02/gta04 aren't valid anymore :) Oct 23 12:15:31 hmm, will adding "Shipped OS" row in comparition table be enough for now? Oct 23 12:15:52 they were apeing bencode format i guess Oct 23 12:17:04 I think we should mention that GolDeliCo is NOT developing the OS for the device. And while we gonna ship the device with errr one of the OS that Nikolaus mentioned, it's basically up to the user to find and install the OS of his choice. Then point to FPTF Oct 23 12:36:00 dos1: F5! Oct 23 12:39:01 dos1: (then point to FPTF) It's very important that customers understand that the device doesn't ship with fremantle, but that there OTOH is a community project that already takes care about porting fremantle Oct 23 12:40:25 and fremantle been ported before to other devices Oct 23 13:27:00 there any way to select twxt in the browser for copying? changing your keymap has the fun circular dependency Oct 23 13:27:44 where to have the characters to edit your keymap you must first edit your keymap to have such characters :P Oct 23 13:51:09 Tekk_, I'm not sure if you're after this, buy if you swipe the stylus from left of the screen to right (that is, start from outside the screen, swipe to right), you should be presented a button. clicking that should allow you to select text, after which you can use the menu or keyboard shortcut to copy it to clipboard. Oct 23 13:52:31 thanks, I'll look at that Oct 23 13:54:13 ~gestures Oct 23 13:54:14 ftp://ftp.handhelds.org/pub/projects/xstroke/release-0.1/alphabet.ps Oct 23 13:54:18 hmm, no Oct 23 13:54:54 https://wiki.maemo.org/N900_shortcuts_and_gestures Oct 23 13:55:00 :p Oct 23 13:57:39 btw DocScrutinizer05, maybe a stupid Q , but what do the both 'browserd' processes do? (I mean there's few of them.... 2 with few forks) Oct 23 13:59:51 scratch that .... Oct 23 14:52:31 really? Oct 23 14:53:41 browserd does a lot of jobs: conversations, iirc call history, whatnot else Oct 23 14:54:02 even ovi-map afaik Oct 23 14:58:16 okay .. another Q.... witch is more linux kernel behavior relayed ... does chnice effect the swap IO? Oct 23 14:58:34 of a process Oct 23 15:08:42 chnice? Oct 23 15:09:01 my manpages don't know chnice Oct 23 15:09:25 Pali: How do I install kp52 with the files from http://atrey.karlin.mff.cuni.cz/~pali/kp52+git/ ? Oct 23 15:09:52 install deb packages Oct 23 15:11:16 Pali: All 4 packages? Oct 23 15:11:28 depends on what you need Oct 23 15:12:48 l4m3rx: man ionice Oct 23 15:13:46 Pali: Can I install only the kernel-power-modules package? Oct 23 15:14:18 if you want to flash kernel into nand you need kernel-power kernel-power-modules and kernel-power-flasher Oct 23 15:14:32 if you want to boot it from uboot you need also kernel-power-bootimg Oct 23 15:14:46 l4m3rx: though kernel isn't controlled by nice or ionice Oct 23 15:14:58 kernel version is kp52+git Oct 23 15:15:18 kswapd seems to run at "top prio" Oct 23 15:15:22 and kernel zimage and modules version must match Oct 23 15:16:05 s/also// Oct 23 15:16:13 s/also/instead/ Oct 23 15:17:01 you probably don't want to install kp to kernel partition, just to replace it with uBoot some minutes later Oct 23 15:18:13 so aiui for uboot, kernel-power-bootimg and kernel-power-modules should suffice Oct 23 15:19:04 pending confirmation by pali Oct 23 15:19:39 no, kernel-power-flasher is needed too for dependences Oct 23 15:19:57 uboot and kernel-power-flasher depends on fiasco-image-update-ask Oct 23 15:20:13 maybe some unfortunate dependencies in installation procedure make kernel-power and *-flasher mandatory even with uBoot Oct 23 15:20:14 and that package cause that you will be asked if you want to flash something into nand Oct 23 15:20:56 * DocScrutinizer05 been too slow Oct 23 15:21:21 l4m3rx my bad, i ment ionice , witch is like chnice but not for IO Oct 23 15:22:24 And the Q here is if you change the IO prio of a process , will that affect the swapping IO of him (the process) Oct 23 15:22:28 chnice: unknown Oct 23 15:23:20 and answer as of above: afaik kswapd doesn't care about any ionice or other prio of the process "owning" the pages to swap Oct 23 15:23:38 so - no :/ Oct 23 15:24:10 kswapd is a kernel process and runs with kernel prio and scheduling Oct 23 15:24:34 that's why system comes to a grinding halt when swap hell hits Oct 23 15:26:19 all disclaimer: AFAIK Oct 23 15:26:30 no expert here Oct 23 15:26:49 ShadowJK might know details Oct 23 15:28:48 but when a process has lower prio then it stays in sleep longer before it gets scheduled (gets a time slot). But once it got a timeslot is has same prio as any other process that tries to run, and kernel tries to provide the resources to allow it running, incl any swapped out pages Oct 23 15:30:20 actually I seem to recall that all processes ahve a real live prio and a niceness. The real live prio increases the longer the process stays in sleep Oct 23 15:31:47 once the real prio of a process is high enough, it will get next timeslice, and it doesn't make sense to preempt it just because some pages need to get swapped to make it run Oct 23 15:32:48 maybe other lower prio processes can run while the process waits for pages swapping in, but once that io-wait finished, the process again is highest prio and again claims cpu Oct 23 15:33:43 and kswapd will not swap out or abot swap in of pages that just been requested for swap in Oct 23 15:33:58 that would lead to a total deadlock of whole system Oct 23 15:41:45 DocScrutinizer05.. Oct 23 15:44:22 after a couple of reboots, I noticed that something weird is going on, cpumem applet shows high cpu and memory usage, Oct 23 15:45:28 and now it wont allow me to open anything, says "too many applications open, unable to open additional apps without closing some." Oct 23 15:47:41 eeeek Oct 23 15:47:53 what the heck? Oct 23 15:48:55 you should immediately restore original state of /etc/init.d/rcS Oct 23 15:49:00 i dont know whats wrong.. sometimes I am not able to even open up xterminal Oct 23 15:49:46 pls tell me how to restore.. original state. Oct 23 15:49:57 you got mc installed? Oct 23 15:50:08 what? Oct 23 15:50:09 or what#s your fav editor on xterm? Oct 23 15:50:37 currently nothing.. Oct 23 15:50:57 well, then type following line into xterm Oct 23 15:51:21 wait, icant open xterm now. its showing the same error as above Oct 23 15:51:48 sed -i 's@.*/prox.*@@' /etc/init.d/rcS Oct 23 15:52:03 well the phone menu works, thats all. Oct 23 15:52:08 reboot and immediately open xterm Oct 23 15:52:15 pls wait Oct 23 15:52:33 rebooting Oct 23 15:52:58 you also might try to umount /prox Oct 23 15:53:09 both needs to get done as root Oct 23 15:54:06 maybe this hack I suggested has unexpected side effects Oct 23 15:54:24 okay.. in xterm now Oct 23 15:54:26 s/maybe/obviously/. Oct 23 15:54:33 sudo gainroot Oct 23 15:54:58 umount /prox Oct 23 15:55:20 sed -i 's@.*/prox.*@@' /etc/init.d/rcS Oct 23 15:56:11 after those three lines your system should stop acting up Oct 23 15:56:54 check system status with command Oct 23 15:56:55 top Oct 23 15:56:57 then Oct 23 15:57:24 cannot umount , device or resource busy Oct 23 15:57:30 watch for a lot of similar processes, tell me what they look like Oct 23 15:57:50 continue with sed command Oct 23 15:58:38 then check output of Oct 23 15:58:58 tail -n 4 /etc/init.d/rcS Oct 23 15:59:26 it shouldn't have any " mount " lines anymore Oct 23 15:59:51 if that's ok, reboot again Oct 23 16:00:09 but before you reboot, please check output of top command Oct 23 16:00:10 no mount lines.. just exit 0 Oct 23 16:00:16 excellent Oct 23 16:00:19 top Oct 23 16:00:38 (to stop top, you press "q") Oct 23 16:01:23 you probably see a lot of similar looking processes in top Oct 23 16:01:33 1, cpu 72.4 --force syslog Oct 23 16:01:54 2, cpu 23.8 hald-addon gpio Oct 23 16:01:56 looks weird Oct 23 16:02:30 these two are the top cpu usage ones Oct 23 16:02:42 there must be a command name before "--force syslog" Oct 23 16:03:02 yes /sbin/mce --force-syslog Oct 23 16:03:05 and I'm interested in a lot of similar looking processes Oct 23 16:03:19 yeah, mce blowing chunks Oct 23 16:03:23 damn Oct 23 16:04:12 there is no other /sbin/mce others are /usr/bin with less than 10% cpu usage Oct 23 16:04:19 for unknown reasons hald-addon-gpio doesn't like the bindmount on prox sensor Oct 23 16:04:33 exept top and hald-addon-gpio Oct 23 16:04:54 cpu usage is irrelevant, you said the error message been "too many apps started" or sth like that Oct 23 16:05:35 yes.. and some times. it said some thing like (cannot continue due to low memory) Oct 23 16:05:48 which means there seems to be too many concurrent processes, which are like 999 or sth, on maemo Oct 23 16:06:15 low memory is another issue, but that is also not related directly to cpu usage Oct 23 16:06:28 you can sort by memory usage in top Oct 23 16:06:43 press "F" Oct 23 16:06:54 press "n" Oct 23 16:07:08 press enter Oct 23 16:08:10 what does top say under "Tasks: 152 total," what's is your number? Oct 23 16:08:27 (top second row) Oct 23 16:09:10 its sortecant find Oct 23 16:09:16 cant find Oct 23 16:09:35 sorry? Oct 23 16:10:08 there is no where saying tasks : xx total Oct 23 16:10:24 wait press f, n enter ??? Oct 23 16:10:26 top, the 2nd row from top Oct 23 16:11:04 2nd row is load avreage: 1.00 1.00 1.00 2/210 2510 ? (or smthng like that?) Oct 23 16:11:06 F, not f Oct 23 16:11:19 oops Oct 23 16:12:13 http://privatepaste.com/d612e0a3a3 Oct 23 16:13:34 well, there is only mem and cpu above the list.. Oct 23 16:13:44 what am i doing wrong Oct 23 16:13:54 sorting doesnt really work Oct 23 16:13:57 too large font? Oct 23 16:14:22 use volume rocker to get smaller font Oct 23 16:15:02 nope, not the font. its not there Oct 23 16:15:21 weird Oct 23 16:15:42 sorry, I can't help much Oct 23 16:16:17 hmm.. Oct 23 16:16:58 and btw with f, n, enter you probably disabled display of MEM% col Oct 23 16:17:17 doing same again will re-enable it Oct 23 16:17:22 no, memory % colum is still there.. Oct 23 16:17:46 it sorts by cpu% Oct 23 16:18:16 you can change this to %MEM by pressing F,n, enter Oct 23 16:19:29 it was never sorted by mem usage, Oct 23 16:20:16 pressing F should show a list of fields, with chars to select ome of them Oct 23 16:20:18 one* Oct 23 16:20:27 F n enter sorts cpu usage in descenting order Oct 23 16:20:54 have you looked at the screen after pressing F ? Oct 23 16:21:25 http://privatepaste.com/c0b11788a8 Oct 23 16:21:51 nope nope nothing like that... Oct 23 16:23:06 no idea what's going on with your shell or top Oct 23 16:23:49 does it even update at all? Oct 23 16:23:58 ?? Oct 23 16:24:05 I.E. does the values change? Oct 23 16:24:39 yes when i press F screen just refreshes Oct 23 16:24:47 o.O Oct 23 16:25:01 press Q, reboot Oct 23 16:25:06 I'm out of ideas Oct 23 16:25:24 rebooting. Oct 23 16:26:46 sorry Oct 23 16:27:27 still top doesnt respond to F Oct 23 16:27:31 but.. Oct 23 16:28:19 cpu 7.2% mem 60kk free Oct 23 16:28:23 OHDAMN! busybox top Oct 23 16:28:53 ~lart busybox Oct 23 16:28:54 * infobot hits busybox with an anvil and laughs with a contralto voice ... Haha Ha HA Ha Oct 23 16:28:59 wait.. it just raised.. and now falled down Oct 23 16:29:13 cpu% Oct 23 16:29:29 raised again.. trackerd..? Oct 23 16:29:43 whatever, irrelevant range Oct 23 16:30:16 no sbin/mce in top 50 haha Oct 23 16:30:17 might be tracker indeed Oct 23 16:30:55 sorry, I gave a poor advice. It obviously isn't helpful Oct 23 16:31:57 seems hald-addon-gpio blows chunks with that bindmount on proximity/state Oct 23 16:32:07 well i found sbin/mce in somewhere 5th or 6th position from top Oct 23 16:32:27 but total cpu% is 7% Oct 23 16:32:43 don't worry, everything back to normal Oct 23 16:32:54 incl dialer misbehaviour Oct 23 16:32:54 sbin/mce gone now Oct 23 16:33:26 to properly clean up you may delete that file Oct 23 16:33:30 rm /prox Oct 23 16:34:52 sorry but where was prox located? I am dumb Oct 23 16:35:07 sorry to ask Oct 23 16:35:12 dumb again Oct 23 16:35:18 rm /prox Oct 23 16:35:23 ^^ got it Oct 23 16:36:06 so i suppose if i use it without the automatic configure at startup, it would be fine..? right? Oct 23 16:37:29 thanks a lot doc.. you saved a reflash. Oct 23 16:41:07 I'm really wondering what happens when you do the `` mount -o bind ´´ late after boot Oct 23 16:42:11 either it acts upp exactly like it did now, or it will magically work since maemo-dialer and hald-addon-gpio are weird and inconsistent design Oct 23 16:42:34 SAiF: I almost *caused* a reflash Oct 23 16:42:38 :-/ Oct 23 16:42:41 :) Oct 23 16:42:55 you saved a couple earlier.. Oct 23 16:43:10 so i still owe you a couple of reflashes . :-) Oct 23 16:44:41 you're free to experiment further with the original interactive cmdline I gave you. It will revert to normal anyway, as soon as you reboot device Oct 23 16:45:38 echo open >/prox; mount -o bind /prox /sys/...blablabla Oct 23 16:45:46 dunno what top you got (htop?) but top has not #tasks Oct 23 16:47:16 ^^ doc Oct 23 16:47:25 my top is the REAL top, not any messybox crap Oct 23 16:47:32 =# Oct 23 16:47:42 gtop then Oct 23 16:47:49 on normal maemo Oct 23 16:47:58 if installed Oct 23 16:48:02 well, if you wanna call it that way, it will obey ;-) Oct 23 16:49:36 IroN900:/etc/init.d# ls -l `which top` Oct 23 16:49:37 -rwxr-xr-x 1 root root 80976 2009-11-16 16:02 /usr/bin/top Oct 23 16:49:39 IroN900:/etc/init.d# top --version Oct 23 16:49:40 top: procps version 3.2.7 Oct 23 16:49:43 so i gather you have coreutils without the silly prefix Oct 23 16:50:08 procps, which nukes my mp-community-pr Oct 23 16:50:15 ok Oct 23 16:50:28 ... so much work so little time Oct 23 16:50:41 I recommend htop Oct 23 16:50:45 at least you guys all went the way before Oct 23 16:51:36 * sono is on the bus enjoying his phone =) Oct 23 17:01:44 bye good night (UGT), time to go home. Oct 23 17:18:36 DocScrutinizer05; kswapd swaps out, apps mostly swap in themselves Oct 23 17:18:51 ooh Oct 23 17:19:15 how's an app going to swap in autonomously? Oct 23 17:20:12 how's the app going to know where to in swap the swapped-out pages got moved? Oct 23 17:20:33 "themselves" in quotes perhaps. Page fault happens when app's memory is missing, kernel schedules swapin Oct 23 17:20:43 yes Oct 23 17:20:54 I thought that's what kswapd is doing Oct 23 17:21:35 page fault, exception, kernel, scheduling a swap task in kswapd queue ? Oct 23 17:22:13 dunno, never looked into that Oct 23 17:22:13 From kernel's point of view, the swapin is accounted as done by app, and all swapouts accounted as done by kswapd Oct 23 17:22:25 mhm Oct 23 17:22:31 interesting Oct 23 17:23:03 so the swap-in is basically scheduled as a normal (mmap) IO? Oct 23 17:23:11 iirc Oct 23 17:23:19 in app context Oct 23 17:23:23 makes sense Oct 23 17:24:07 so swap-in is also susceptible to ionice? Oct 23 17:24:15 there's little distinction made in a high level sense between disk cache, executable code, libraries and "anonymous memory" Oct 23 17:24:42 No idea, but in practice it's hard to make ionice work for mapped stuff, I think.. Oct 23 17:25:10 interesting. Thanks for sharing Oct 23 17:25:15 Also, the tweaks prioritizing executable/library memory over disk cache was added later than 2.6.28, iirc Oct 23 17:25:54 In any case, swapout hurts us significantly more than swapin Oct 23 17:28:20 that's for sure Oct 23 17:28:48 since damn swap-out is a write to a media that's exceptionally slow on write, compared to read Oct 23 17:29:12 yep Oct 23 17:29:37 in my experience, lowering page-cluster after things start swapping heavily helps Oct 23 17:30:14 As it'll preread in 16k or something instead of full 128k Oct 23 17:32:33 ShadowJK: Given that flash pages are 128k, how is it going to help? Oct 23 17:32:54 One way or another it will still have to switch a full page in Oct 23 17:35:22 Not entirely Oct 23 17:36:35 erase units are bigger than 128k, but you can read them in pages of 4k on the emmc, and the latency hit in switching to another "128k" on read is insignificant enough Oct 23 17:39:37 While a whole "128k" has to be erased for writing, once erased you can also write it 4k at a time :-) Oct 23 17:39:54 But the emmc's ability to keep track of partially written things is extremely limiteed. Oct 23 17:40:09 Most kingston stuff, for example, can keep track of 1. :-) Oct 23 17:40:40 So as soon as you venture outside your current 128k block, the next time you come back it's full erase-modify-write again. Oct 23 17:41:17 Better stuff is able to track 8-12 open erase blocks Oct 23 17:43:38 well, I suppose it will clean out pages whenever it can, but pretty soon there's no "empty" pages left over. As long as there's an empty page the controlled can read-modfy-write_elswhere without prior page erase Oct 23 17:44:24 and in fact it's the erase that takes ages, while writing to a erased page is comparably fast Oct 23 17:45:39 thus an optimized API would allow to mark blocks as empty Oct 23 17:46:11 some controllers try to deduce this from interpreting the fs, by looking at the FAT Oct 23 17:46:37 thus they could erase unused pages async Oct 23 17:49:22 trim? Oct 23 17:49:36 anyway if I had to design the controller, I'd use virging pages to do a readA-modify-writeB-eraseA cycle, instead of a readA-modify-eraseA-writeA Oct 23 17:50:10 yeah, trim goes this direction Oct 23 17:51:55 anyway a mkfs tends to spam the whole device with inodes, so no virgin pages left over Oct 23 17:52:50 except maybe for the hidden spare pages if the card has any Oct 23 17:53:46 you could test that with two virgin cards, creating a 100% partition on one and a 90% partition on the other, then compare performance Oct 23 17:56:36 the more free empty pages the card has, the longer the burst you can write before deferred page-erase kicks in and needs to finish clearing a page for the next chunk to write Oct 23 17:57:54 I think equally spread inodes across the whole storage range are the worst thing you can do to a flash storage Oct 23 17:59:47 I'm not so sure the SD protocol lends itself to the device doing background cleaning :/ Oct 23 18:00:26 New kernels and new SD cards (or emmc), and with driver support, there's actually a trim available Oct 23 18:00:30 "The MultiMediaCard and SD ERASE (CMD38) command provides similar functionality to the ATA TRIM command, although it requires that erased blocks be overwritten with either zeroes or ones. eMMC 4.5 further defines a "discard" sub-operation that more closely matches ATA TRIM in that the contents of discarded blocks can be considered indeterminate (i.e., "don't care")." Oct 23 18:00:54 via wiki Oct 23 18:01:12 But whether cards actually do anything useful with it is unknown. Full device trim runs so fast on the card I've got in my raspberry pi, I suspect it's a no-op on the card :-) Oct 23 18:02:02 Oh the a-data card in my N800 actually shows improved performance after dd of zeroes :-) Oct 23 18:03:37 But after a full erase, it slows down far sooner than writing its capacity Oct 23 18:03:51 at 10% mark or so.. Oct 23 18:04:26 ShadowJK: (so fast) or deferred Oct 23 18:05:00 dos1: thanks Oct 23 18:05:50 Well the bestest best cards now allegedly run a SLC mode log-structured cache, to help making FAT update faster on their new fancy 12 megabyte blocksize TLC nand Oct 23 18:08:12 there should be an option on MLC/TLC cards to run in SLC mode at the cost of halving or 1/3 the memory... I wonder how long the card would last in such a mode... Oct 23 18:09:04 or I should say "slc" mode - using each cell for one bit instead of 2 or 3. Oct 23 18:12:03 I have a (fullsize) SD card SLC 16GB. Oct 23 18:12:09 It's not that great Oct 23 18:12:31 Would be more worthwhile spending silicon on improving the FTL Oct 23 18:14:29 Hey guys, my phone was shut down most likely because watchdog tried to reboot it, now I cannot boot it at all. When I press power button I can see NOKIA logo on the white screen but then nothing happens Oct 23 18:15:07 I even had some boot menu installed, but even if I open my keyboard it wont get to it Oct 23 18:15:20 simply turns off after showing nokia logo Oct 23 18:32:20 anybody? :) Oct 23 18:33:13 No idea, other than that bootloader throws up the logo Oct 23 18:33:25 so it's not proceeding from that Oct 23 18:33:58 So.. storage died, it thinks battery is dead, os is missing, other fault? Oct 23 18:35:53 could it be that it has no free space? Oct 23 18:36:24 battery is not dead for sure Oct 23 18:38:20 if I plug in my USB cable then it shows the logo, turns off backlight, then reboots and does it again and again Oct 23 18:41:30 I've tried taking out sd card but that doesn't seem to change anything Oct 23 18:41:49 so... reflashing is the only way? Oct 23 18:43:04 give 10 min without battery and try again? Oct 23 18:45:23 that's actually a good idea, but I'm afraid that my device doesn't need whole 10 minutes Oct 23 18:45:55 even 10 seconds without the battery will result in a date reset Oct 23 18:46:09 but I will try it, thanks Oct 23 18:46:36 is there anything that I can use to debug current situationL Oct 23 18:46:37 ? Oct 23 18:53:54 I'm wondering if flasher could upload a kernel+initram without flashing it, and boot that for some sort of minimalist debugos Oct 23 18:54:09 dunno if such an image exists though.. Oct 23 18:54:20 Never really looked up this rescueos thing or how it works Oct 23 19:00:32 no, battery trick did not work Oct 23 19:00:43 :( Oct 23 19:00:51 noooooo Oct 23 19:01:07 well, at least my files are backed up Oct 23 19:01:31 but still, bootmenu gave me false feeling that I'm safe even if something bad happens Oct 23 19:18:04 hello all Oct 23 19:29:49 it's called rescueOS Oct 23 19:30:15 ~rescueOS Oct 23 19:30:15 well, rescueos is http://n900.quitesimple.org/rescueOS/ Oct 23 19:30:40 Sysaxed: ^^^ Oct 23 19:37:51 ehh Oct 23 19:38:05 I wish there were more DocScrutinizers in the world Oct 23 19:38:18 DocScrutinizer05: do you have any kids? :) Oct 23 19:40:47 if yes - good job! :) Oct 23 19:45:23 please excuse my momentary oblivion Oct 23 19:54:57 haha Oct 23 19:55:21 "to disable this message type 'stfu'" Oct 23 19:55:55 so Oct 23 19:56:13 I get Attention: Charge the battery bla-bla Current: 0% Oct 23 19:56:51 does it mean that something is wrong with the battery? Oct 23 19:57:27 Does it mention a voltage? Otherwise the charge meter could've been reset when you had the battery out Oct 23 19:57:45 nope Oct 23 19:59:21 umm, now when rescueos is running, what can I try? Oct 23 20:02:43 for example, how can I check the free space? Oct 23 20:04:47 uhhhmm Oct 23 20:05:30 /rescueOS/umount-maemo-root.sh says umount: can't umount /mnt/maemo: Invalid argument Oct 23 20:05:59 Couldn't unmount. Most likely your fault. Are you in /mnt/maemo ? Oct 23 20:06:09 ah shit Oct 23 20:06:13 I'm dumb Oct 23 20:06:23 got it working, thanks Oct 23 20:11:29 holy shit Oct 23 20:11:42 now it booted up correctly Oct 23 20:11:49 how could this happen Oct 23 20:18:29 Well, I cannot believe it, but seems like simply booting rescue os fixed the problem Oct 23 20:22:20 Sysaxed: I'm scared by the idea there could be more grumpy old men just like me Oct 23 20:23:40 anyway I consider defect battery a plausible cause of your problems Oct 23 20:25:29 Sysaxed: I'm doing my share to help extinct mankind - so no, no kids here Oct 23 20:26:49 s/extinct/exterminate/ Oct 23 20:26:50 DocScrutinizer05 meant: Sysaxed: I'm doing my share to help exterminate mankind - so no, no kids here Oct 23 20:28:35 maybe NIN101 has kids - would be the way better elternative to me getting scared by my alter egos Oct 23 20:28:52 alternative even Oct 23 20:30:44 for testers: pre release version of kp53: http://talk.maemo.org/showthread.php?p=1381948#post1381948 Oct 23 20:31:57 me? Better not :-). Oct 23 20:33:00 DocScrutinizer05: noooooooo :( are you kidding me Oct 23 20:38:31 * DocScrutinizer05 just thought about a world with 95% of human beings being an exact clone of DocScrutinizer05. :-o A terrible place to be, for ~218 minutes. Then maybe evolution gets a second chance after the slaughter Oct 23 22:00:10 DocScrutinizer05: Then 95% of the human population would be male. Oct 23 22:08:46 brolin_empey: not after the cleansing. Oct 24 01:40:20 anyone here besides the developers seriously considering the new MB option for the N900? Oct 24 01:40:58 seriously considering how? Oct 24 01:41:16 as in ready to write a check if the kickstarter funds Oct 24 01:41:36 personally I'm not interested in it since it seems like a really puny upgrade for the extra price Oct 24 01:42:17 but is there really anything better that is useable? Oct 24 01:42:29 I'd just continue using the old board Oct 24 01:43:07 until something else(hopefully mer/nemo based) comes along Oct 24 01:44:00 what would make it worthwhile for you? Oct 24 01:44:18 the motherboard upgrade? Oct 24 01:45:22 probably at least 1 gb ram, cpu either with more cores(might be helpful since maemo is so multitask friendly) or at least clocked higher than people overclock the original processor to Oct 24 01:45:24 it can't get any cheaper if the production quantities are low Oct 24 01:45:35 it's 1GB RAM already Oct 24 01:45:42 I thought it was 512 Oct 24 01:45:44 yes Oct 24 01:45:48 okay, that just leaves the cpu then :) Oct 24 01:45:58 http://neo900.org/ Oct 24 01:46:12 hmm Oct 24 01:46:16 isn't multicore going to slam battery life? Oct 24 01:46:19 actually does the new modem support at&t freqs? Oct 24 01:46:47 wish there was some clear indicator on if they will add BLE or not though Oct 24 01:46:59 there are three modem options, but I don't know about at&t freqs Oct 24 01:47:56 lte modem does Oct 24 01:47:57 cool Oct 24 01:48:07 and the 3g Oct 24 01:48:22 I do wonder if they are going to get IMEI blocked by ATT since ATT likes to certify their devices Oct 24 01:48:48 might be worth it between being able to use better than edge and more ram Oct 24 01:48:55 but I haven't any money to pledge Oct 24 01:49:16 a big advantage will be that the hardware is much better supported by the kernel, meaning you won't be stuck with 2.6 and other operating systems will run on it much better Oct 24 02:01:24 yea, hopefully the nokia closed bits won't affect the kernel in anyway **** ENDING LOGGING AT Thu Oct 24 02:59:58 2013