IRC log of #maemo-ssu for Thursday, 2013-09-05

*** FlameReaper has joined #maemo-ssu00:00
DocScrutinizer05FatPhil: maybe better pick option one for today, from >>Either it's late, or find_next_zero_bit() has a bug...<<00:00
freemangordonFatPhil: can't find any difference between KP and upstream in find_next_zero_bit00:00
freemangordoneven the commnts match00:01
freemangordon*comments00:01
* DocScrutinizer05 idly wonders whether we should ask the guy that hosts mxr to eventually move it to the maemo.org server infra00:05
*** NIN101 has quit IRC00:08
*** nox- has joined #maemo-ssu00:26
*** handaxe has quit IRC00:36
*** Vlad_on_the_road has quit IRC00:56
*** sunny_s has quit IRC00:57
*** Handaxe has joined #maemo-ssu01:02
*** Handaxe has left #maemo-ssu01:02
*** kolp has quit IRC01:34
*** dos1 has quit IRC01:35
ShadowJKI forget, did he go to palm/hp?01:43
DocScrutinizer05RIM?01:47
DocScrutinizer05tbh I even forgot who's hosting it, crashie or time less01:48
ShadowJKone of those01:49
DocScrutinizer05could do a whois, but too lazy01:49
ShadowJKI think HP was the one doing shotgun recruitment around elopcalypse time01:49
DocScrutinizer05prolly crashie then, since mxr.m.o for sure not hosted by a (ex-)Nokian01:51
DocScrutinizer05timeless worked on microB01:51
DocScrutinizer05also crash is hosting one of the mirrors, and I'd bet on IPs of the mirror and mxr are identical or at least close01:53
ShadowJKtimeless01:55
*** FlameReaper has quit IRC01:55
ShadowJKcrashanddie is, iirc, a very vocal and temperamental community member01:56
*** luf has quit IRC01:57
DocScrutinizer05hmm02:00
DocScrutinizer05ok02:00
ShadowJKI wonder if timeless is working on firefoxos now02:05
DocScrutinizer05ask him! he's over at #maemo02:12
*** aap has quit IRC02:37
*** aap has joined #maemo-ssu02:55
*** nox- has quit IRC02:58
*** M4rtinK has quit IRC03:00
T_Xdoes anyone know which kernel module is responsible for the phone/umts/etc.?03:24
T_Xfor the n900 that is03:24
T_Xand secondly, would unloading that module prevent any interaction with a phone tower?03:27
*** oldtopman has joined #maemo-ssu03:30
DocScrutinizer05a: there's not exactly a phone kernel module, there's phonet and cmt_speech03:35
DocScrutinizer05the one is for communication, the other for audio03:36
DocScrutinizer05b: the modem doesn't need any of those kernel modules neither libisi to communicate with the "phone tower"03:37
DocScrutinizer05as soon as you power up the modem, it already starts searching for cells, and it *might* even communicate with them03:38
DocScrutinizer05it usually _shouldn't_ transmit as long as no AT+CFUN=1 or the isi analogon gets sent from AP to cmt, but you can't guarantee that, the firmware might be buggy or compromised03:41
*** BCMM has quit IRC03:46
T_XDocScrutinizer05: ah, k, thx!03:55
*** Martix has quit IRC04:01
ShadowJKThere are direct lines from omap gpio to modem reset, though the modem is a blackbox and who knows if it can operate while being held in reset04:16
*** Pali has quit IRC04:21
*** dafox has quit IRC04:31
T_XI had some message about emergency numbers disabled when I once tried rmmodding random kernel modules. but yeah, you're right that doesn'- have to mean anything04:33
T_Xand I guess AT+CFUN=4 (disable transmit and receive RF circuits (flight mode) ) won't help either04:34
T_Xfunnily, for AT+CFUN=4 I don't get this message about disabled emergency numbers04:34
T_Xhm, rmmoding ssi_mcsaab_imp seems a little promising. that's the one disabling the emergency numbers04:53
T_Xand it also disables pnatd, so entering AT_CFUN=x can't be entered anymore :)04:54
*** freemangordon_ has joined #maemo-ssu05:10
*** freemangordon has quit IRC05:10
*** amiconn has quit IRC05:14
*** amiconn_ has joined #maemo-ssu05:14
*** amiconn_ is now known as amiconn05:14
DocScrutinizer05I wonder what you're after05:25
DocScrutinizer05http://siliconpr0n.org/wiki/doku.php?id=physical_protection :-D05:25
ShadowJKmcsaab is going to be the actual data lines07:07
T_Xhm, would be handy to have a base station in our hackerspace to test stuff like that07:13
T_X:)07:13
T_Xor a USRP07:14
*** ruskie has quit IRC07:19
ShadowJKTest what07:21
ShadowJKLoading and unloading kernel modules is kinda pointless :P07:21
ShadowJKThe modem is a fully functional computer with its own CPU, RAM and its own power management.07:23
ShadowJKThe same modem can (and probably has been) used on its own in a featurephone or low-end symbian phone, with a display and keypad connected directly to it07:23
ShadowJKAfter Symbian gained realtime capabilities, the OS could be loaded into the modem's cpu and run concurrently with the realtime tasks of the modem :)07:25
ShadowJKSave a cpu+ram combo07:25
ShadowJKI'll note further that the GPS is hooked up to the modem ;)07:25
*** ruskie has joined #maemo-ssu07:46
*** Vlad_on_the_road has joined #maemo-ssu08:17
*** Vlad_on_the_road has quit IRC08:36
*** LauRoman has quit IRC08:42
*** oldtopman has quit IRC08:55
FatPhilOne would hope that from a power management perspective, once the last thing making use of the modem is removed, it would be powered down.09:31
FatPhilperhaps powertop will show the difference in power/clock domain activity in the two cases09:33
*** sunny_s has joined #maemo-ssu10:08
*** LauRoman has joined #maemo-ssu10:12
kerioFatPhil: it might be so10:30
*** freemangordon_ is now known as freemangordon10:39
*** kolp has joined #maemo-ssu11:01
*** Pali has joined #maemo-ssu11:04
*** sunny_s has quit IRC11:17
*** sunny_s has joined #maemo-ssu11:20
*** LaoLang_cool has joined #maemo-ssu11:22
*** LaoLang_cool has quit IRC11:33
*** piscodig has joined #maemo-ssu11:42
*** discopig has quit IRC11:42
DocScrutinizer05ShadowJK: BB5 is on many Nokia phones, indeed12:09
DocScrutinizer05FatPhil: how would modem know about "removing things" that use it?12:11
DocScrutinizer05it sounds like "after switching off speakers and monitor, i'd hope the PC powers down"12:12
DocScrutinizer05FatPhil: actually it's like that: you need to tell modem to power down, explicitly.12:18
DocScrutinizer05more often than not it's the AP that plays power-down, and it's modem to wake AP up12:19
keriomore like "i hope that if nothing is using the audio, then the audio card powers down"12:19
keriowhich is not ridiculous, to be honest12:19
DocScrutinizer05that's correct... for audiocard12:19
*** Martix has joined #maemo-ssu12:19
kerios/powers down/is powered down/12:19
DocScrutinizer05nonono12:19
DocScrutinizer05aiui the card auto-powerdowns12:20
DocScrutinizer05not on N90012:20
DocScrutinizer05but on PC aiui12:20
DocScrutinizer05well, somedtimes it might be the alsa soundcard driver module in kernelspace doing the powerdown12:21
DocScrutinizer05with some luck, and when not using braindamaged PA12:21
DocScrutinizer05pukeaudio12:22
DocScrutinizer05err PolypAudio12:22
DocScrutinizer05dang, is tmo dead? or just all its users?12:24
DocScrutinizer05tmo is more mindboggling than a pub, you never know when there's busy time so you don't find a seat and when it's dead12:25
DocScrutinizer05last ten days or so I seen ~2 new posts per hour, and ~100 thanks per day. Since 12h zilch12:25
jon_ytmo is dead?12:26
jon_y:)12:26
* jon_y tend to avoid forums and use mailing lists instad12:26
jon_y*instead12:26
DocScrutinizer05yeah, did same until I came to maemo12:27
DocScrutinizer05for OM we *had no* forum12:27
DocScrutinizer05God was that a good time12:27
jon_yless eternal September shenanigans12:28
DocScrutinizer05then came freeyourphone.de12:28
jon_ynot to say mailing lists were free of them12:28
DocScrutinizer05dang, I didn't have a look at freeyourphone since... years?12:29
jon_ywhat is that site about?12:29
DocScrutinizer05err about openmoko Neo Freerunner and other "free" phones12:45
jon_ycarriers don't want "free" phones12:51
jon_ythey want iphones :(12:51
*** dafox has joined #maemo-ssu12:57
DocScrutinizer05carriers want business and thus customers13:07
jon_ydo consumers want freedom?13:08
jon_ythat's a hard question13:08
jon_yprobably not, choices are hard13:09
*** BCMM has joined #maemo-ssu13:13
DocScrutinizer05well, quite obviously many (most?) of them prefer "free" services like fb/twitter/watsap/gmail crap over freedom13:15
keriowell13:16
kerioif you don't mind the privacy invasion, gmail is pretty fucking awesome, to be fair13:16
jon_ykerio: at least I get to use gnupg to pre-encrypt first13:17
jon_yI'm still waiting one day when a webservice will say "No encryption allowed"13:17
DocScrutinizer05a friend of mine used to attend each and every competition he could find, and then complained about all the spam snailmail and email he received13:17
DocScrutinizer05jon_y: hah, dream on! I heard they support encryption now - *serverside* via web frontend X-P13:19
jon_ykind of defeats the point entirely13:19
DocScrutinizer05actually e.g web.de does that since ~10 years13:19
jon_ydoesn't that mean they have your private key?13:20
keriojon_y: why does that defeat the purpose? what purpose is it defeating?13:20
DocScrutinizer05yes, they create it for you ;-P13:20
kerioassuming you trust the service, that is13:20
jon_ykerio: the whole idea of offering encryption when they can read it too13:21
*** LauRoman has quit IRC13:22
jon_yalso, unencrypted over the air transmission13:23
DocScrutinizer05yup13:23
kerioi assume you have https to your webmail13:23
DocScrutinizer05or poorly encrypted by SSL13:23
jon_ygovernment has certs13:24
jon_ywhy do you need to encrypt it, comrade?13:24
DocScrutinizer05the snoop-now-decrypt-later approach is pretty effective for SSL13:25
jon_ydefinitely need httpss13:25
jon_yasymetrical rather than symmetrical13:26
jon_yeach "session" uses a different set of keys13:26
jon_ythen again, it sounds like ballroom dancing13:26
DocScrutinizer05there's a means for that in SSL, but nobody uses it13:26
DocScrutinizer05except google ;-P13:27
jon_ythey do?13:27
DocScrutinizer05Perfect Forward Secrecy13:30
jon_yis it possible to do this with standard apache setup?13:33
DocScrutinizer05http://wstaw.org/m/2013/09/05/plasma-desktopZj3743.png13:34
DocScrutinizer05openssl s_client -connect wiki.maemo.org:8013:36
DocScrutinizer05openssl s_client -connect wiki.maemo.org:44313:36
DocScrutinizer05Cipher    : .*DHE.*13:37
jon_yok, encryption cipher selection?13:40
DocScrutinizer05http://wstaw.org/m/2013/09/05/plasma-desktopqi3743.png  <- list of browsers vs "supports PFS"13:40
DocScrutinizer05fb, twitter, yahoo, ebay, paypal -> no DiffieHellman13:44
DocScrutinizer05aka PFS13:44
DocScrutinizer05MS as well not really. Google though full support13:44
kerioCipher    : DHE-RSA-AES256-SHA13:45
keriodefault nginx config :D13:45
keriodem russians13:45
DocScrutinizer05DH == DiffieHellman == PFS13:45
jon_yok13:46
DocScrutinizer05chrome shows encryption used, search for *DH*13:46
jon_yhere's my own server ECDHE-RSA-AES128-SHA25613:46
jon_yis that good?13:46
kerioeyup13:46
DocScrutinizer05DH is good13:46
jon_yok13:47
keriowhat does the E stand for?13:47
kerioephemeral?13:47
jon_ybasically no change needed13:47
jon_yno idea :)13:47
DocScrutinizer05nfc13:47
DocScrutinizer05aah eliptical13:47
DocScrutinizer05elyptical?13:47
kerionah, that's "EC"13:48
keriofor ellyptic curve13:48
DocScrutinizer05*shrug*13:48
DocScrutinizer05yep13:48
kerioyeah, E stands for ephemeral13:48
jon_ybtw, windows still hate DAV+SSL on win713:48
*** BCMM_ has joined #maemo-ssu13:49
DocScrutinizer05ephemeral13:50
DocScrutinizer05meh he beats me13:50
DocScrutinizer05NB you can check all services using SSL with the above commandline13:52
DocScrutinizer05just pick the port of your choice13:52
DocScrutinizer05http://wstaw.org/m/2013/09/05/plasma-desktopkO3743.png <- check of some popular service providers (German focus)13:54
FatPhilDocScrutinizer05: I meant the last kernel driver using the modem resource turns off the lights. Some of that's automatic, such as the clock tree framework. The last thing using the McSAAB bus will reduce the usage count to 0 as it exits, and the bus' clock will be disabled, and its parents if possible.14:05
DocScrutinizer05would be nice, but doesn't work this way with modem afaik14:06
DocScrutinizer05layering14:07
DocScrutinizer05th eprotocol to shutdown modem is 2 layers above the phy14:07
DocScrutinizer05after all it's "far end", the kernel driver could only shut down the OMAP interface block14:09
DocScrutinizer05kernel doesn't know shit about "AT"14:10
DocScrutinizer05AT+POFF, AT+CFUN=014:10
keriono fun huh14:11
DocScrutinizer05sure, BB5 uses ISI, but the idea stays the same14:11
DocScrutinizer05you can't teach the tty kernel driver to display a "I'm offline" message on the vt10014:12
DocScrutinizer05your shell could do that, on receiving SIGHUP, before shutting down and closing dev/tty14:13
DocScrutinizer05for modem that's libisi14:14
DocScrutinizer05or even higher14:14
DocScrutinizer05SHR for example literally suspends APE and thus shuts down tty to modem completely. Modem lives on happily and sends a signal via a special GPIO line when some message pending, to wake up APE14:15
DocScrutinizer05you even have that concept in age old RS232: RI14:16
DocScrutinizer05RingIndicator14:17
DocScrutinizer05the ancient days' WOL14:17
jon_ywhat's wrong with RS232? simple 2 wire protocol was pretty nice14:25
*** arcean has joined #maemo-ssu14:27
merlin1991argh rs23214:33
merlin1991kernel interface for that is driving me mental14:33
DocScrutinizer05who said sth wrong?14:35
DocScrutinizer05and RS232 is not 2 wire at all14:36
DocScrutinizer05more like 9 wire14:36
DocScrutinizer057 maybe14:36
DocScrutinizer05or 1914:36
DocScrutinizer0513 actually14:38
DocScrutinizer05http://en.wikipedia.org/wiki/Rs232#Signals14:38
DocScrutinizer05ok, you could argue that virtually nothing which is working anymore supports secondary channel14:40
DocScrutinizer05then it's 8-wire14:40
DocScrutinizer05jon_y: however what I refered to been simply this: >>Ring Indicator' (RI), is a signal sent from the modem to the terminal device. It indicates to the terminal device that the phone line is ringing. In many computer serial ports, a hardware interrupt is generated when the RI signal changes state.<<14:41
merlin1991DocScrutinizer05: I've seen a lot of 2 wire "rs232"14:47
merlin1991ie garmin gps recievers send there data over that14:47
merlin1991s/there/their/14:47
infobotmerlin1991 meant: ie garmin gps recievers send their data over that14:47
DocScrutinizer05that's not RS232 though, that's *using* rs23214:50
DocScrutinizer05ok, jon_y never said RS232 *is* 2-wire, actually14:51
DocScrutinizer05you could call 2-wire low-voltage rs232 as well 20mA-tty14:53
DocScrutinizer05actually it resembles 20mA more than it's similar to RS23214:54
jon_yDocScrutinizer05: iirc rs232 refers to the old 12V signaling protocol14:55
DocScrutinizer05since 20mA-tty actually is a 2-wire, and voltage levels are not up to spec for both lv-rs232 and lv-20mA14:56
jon_yI can't remember what the others are called14:56
DocScrutinizer05rs485 for example14:56
DocScrutinizer05rs422?14:56
DocScrutinizer05there are several14:56
jon_yI mean they could talk to each other provided you have some voltage buffer in between14:56
jon_yall have start/stop bits14:57
jon_ywork over 2 wire rx/tx14:57
DocScrutinizer05http://en.wikipedia.org/wiki/Rs232#Related_standards14:58
FatPhilDocScrutinizer05: In that case the last user-space app to use the modem functionality should turn out the lights on the modem, and then turn out the lights on the kernel modules.14:58
DocScrutinizer05yep14:58
jon_yof course, this assume both side talk at about the same clock freq14:58
*** kolp has quit IRC15:03
*** amiconn_ has joined #maemo-ssu15:05
*** amiconn has quit IRC15:05
*** amiconn_ is now known as amiconn15:05
*** discopig has joined #maemo-ssu15:16
*** piscodig has quit IRC15:18
*** piscodig has joined #maemo-ssu15:19
*** discopig has quit IRC15:23
ShadowJKI remember MeegoCE on N900 used significants amount of power also *after* shutting down the phone :)15:26
jon_yafter? how?15:26
ShadowJKit didn't turn off everything before switching off cpu15:29
jon_ymaybe ARM should have ACPI states instead15:29
*** raccoon_ has quit IRC15:38
*** raccoon_ has joined #maemo-ssu15:39
*** piscodig has quit IRC15:47
*** discopig has joined #maemo-ssu15:52
*** piscodig has joined #maemo-ssu15:55
*** dhbiker has quit IRC15:58
*** discopig has quit IRC15:59
*** arcean has quit IRC16:21
*** BCMM_ has quit IRC16:23
*** Martix has quit IRC16:23
*** BCMM_ has joined #maemo-ssu16:46
*** oldtopman has joined #maemo-ssu16:51
*** sixwheeledbeast has quit IRC17:04
*** jade has joined #maemo-ssu17:11
*** jade has quit IRC17:11
*** jade has joined #maemo-ssu17:11
*** sixwheeledbeast has joined #maemo-ssu17:11
*** piscodig has quit IRC17:15
DocScrutinizer05ShadowJK: yeah, but stskeeps always thought they could take care about power management "later", first get the system working. Seems i'm right this concept doesn't fly17:19
DocScrutinizer05much like openmoko thought they could take care about proper user account management later, and just got for root-only openembedded shit for a start. Also didn't pan out17:21
DocScrutinizer05untill this very day most openmoko distros have root-only config17:22
DocScrutinizer05and nobody ever will do the fancy to fix all the permission issues that would make it break *epically* when introducing real non-root users17:23
DocScrutinizer05apps accessing /sys/* just like pfff17:23
DocScrutinizer05all sorts of files either world readable or only by root17:24
DocScrutinizer05an abomination17:25
DocScrutinizer05and pretty much the same happened to meego re PM17:25
DocScrutinizer05"but! but! we DO have only ONE user!" BWAHAHA17:28
DocScrutinizer05I can see ~50 distinct "users" on my PC and that has no interactive user account but mine17:29
DocScrutinizer05likewise you never get the whole peripherals PM sorted when you start with a flawed API17:31
DocScrutinizer05and flawed paradigms at large17:31
DocScrutinizer05like "last one shuts down the light" (decrement use counter) won't really meet embedded needs when the sensor or whatever peripheral you are talking to is properly event/IRQ driven17:34
DocScrutinizer05you need a whole new API paradigm to account for that17:34
DocScrutinizer05so *first* define proper PM-compliant APIs, only *then* even _start_ implementing processes/apps using those APIs17:36
kerioDocScrutinizer05: fwiw, a lot of processes don't need explicit power management17:36
DocScrutinizer05yeah fwiw17:36
DocScrutinizer05it's never processes that need PM, it's the hardware that does17:37
keriono, i mean17:37
kerioif you do async I/O and your UI toolkit is sane, you wouldn't drain a handheld's battery17:38
keriowhen not in use, at least17:38
DocScrutinizer05and how do you handle e.g. GPS?17:38
kerioqt mobility :P17:38
DocScrutinizer05ROTFL17:38
kerioi'm just saying17:38
*** obironbo has joined #maemo-ssu17:38
kerioif your code is littered with spin locks, the problem is not that you're going to drain my battery17:39
keriothe problem is that you're breathing my air17:39
DocScrutinizer05qt mobility has exactly ZILCH concept to define how exact, how frequent, how urgent your "fix" shall be17:39
keriois there a standard gps api in linux?17:39
DocScrutinizer05no17:39
keriobut gpsd is developed by ESR17:40
DocScrutinizer05you need a proper middleware to handle such stuff17:40
DocScrutinizer05userland apps request certain services of a certain quality (or whatever other properties/attributes) at the middleware. Middleware's job is to aggregate/consolidate thos requests and transform it into proper management of resources17:41
DocScrutinizer05QtM completely ignored that, initially (and for many sensors still does)17:42
DocScrutinizer05on maemo for example you got liblocation that does a rather decent job as middleware17:43
DocScrutinizer05of course I prefer a more general concept like FSO17:43
DocScrutinizer05http://www.freesmartphone.org//17:44
kerio"It offers DBus-APIs all over the place"17:45
*** jade has quit IRC17:49
*** discopig has joined #maemo-ssu17:51
DocScrutinizer05yeah, unlike LIBlocation17:52
DocScrutinizer05btw: http://gypsy.freedesktop.org/wiki/17:55
DocScrutinizer05and http://gypsy.freedesktop.org/why-not-gpsd.html18:08
*** dafox has quit IRC18:19
DocScrutinizer05I can't state honestly that I been involved in development of FSO (only some peer review and design discussions), but I consider it a worthwhile thing to look into and use whenever possible18:20
DocScrutinizer05phonet folks in their hybris and arrogance rejected any cooperation, blaming FSO for allegedly just being a glorified AT interface to GSM18:21
DocScrutinizer05in the end I think this attribute fits perfectly to phonet itself18:22
DocScrutinizer05see BT tethering, for example18:23
DocScrutinizer05kerio: who or what is "ESR"?18:25
kerioEric S. Raymond18:32
DocScrutinizer05whoever is that18:41
DocScrutinizer05should I know him?18:41
kerioa university dropout who wrote The Cathedral and the Bazaar18:41
DocScrutinizer05umm18:42
DocScrutinizer05http://gypsy.freedesktop.org/why-not-gpsd.html is a nice read18:42
kerioyeah, i read it18:42
keriomakes sense18:42
DocScrutinizer05NB FSO is using org.freedesktop.gypsy18:43
kerioi still don't like dbus, but i guess that it's the least worse IPC18:43
DocScrutinizer05ack18:43
*** dos1 has joined #maemo-ssu18:43
DocScrutinizer05dbus is an abomination, but alas the only  thing we got18:43
DocScrutinizer05did you know that timeout is hardcoded in dbus? :-O18:44
freemangordonDocScrutinizer05: hmm? what timeout18:45
freemangordon?18:45
DocScrutinizer05some timeout when using sync calls, I guess18:45
*** M4rtinK has joined #maemo-ssu18:45
freemangordonwhere did you get that from? afaik you pass the timeout as a parameter18:46
* freemangordon is going to check18:46
FatPhil"Don't Use Malloc"? WTF18:46
DocScrutinizer05well, it seems I've seen that being mentioned some time ago, and that for embedded it had to get tweked from 30s to 2 min, or the other way around18:46
DocScrutinizer05ROTFL, yeah18:47
DocScrutinizer05ESR18:47
kerioto be fair, if you only need a static amount of memory...18:48
FatPhilI know the javaloons say "memory allocation is far too important to risk letting the programmer have control", and greybeard C-ites say "memory allocation is far too important to let it all be handled automatically", but I think I'm going to invent a new twist18:48
DocScrutinizer05ome years ago there even was a *nasty* bug in all async calls of dbus, that seems was rather hard to fix18:48
DocScrutinizer05I for one think some of the bugs in sbus still hide there and show every now and then18:48
FatPhil"memory allocation is far too important to let fucknut programmers take control of it". And I thikn I've now found those fucknuts.18:49
DocScrutinizer05dbus*18:49
freemangordonDocScrutinizer05: hmm, http://dbus.freedesktop.org/doc/api/html/group__DBusConnection.html#ga8d6431f17a9e53c9446d87c2ba8409f018:49
FatPhildbus had some horrific O(n) or worse behaviour a few years ago18:49
freemangordonFatPhil: where, you made me curious18:49
FatPhilDocS's gpsd link18:49
kerioO(n) is fairly good18:49
kerioi like O(n)18:50
DocScrutinizer05freemangordon: nfc then, maybe it's in some libdbusglib or whatever18:50
*** sunny_s has quit IRC18:50
FatPhilO(n) effort to achieve O(1) work is fucking awful18:50
keriophp has a library function to compute a greedy estimation of the distance between two strings in Om(n^3)18:51
keriothe textbook exact calculation is O(n^2)18:51
FatPhilDbus was *clearly* (from an algorithmist's perspective) designed from the ground up to not be scalable.18:51
FatPhilYet its selling point was that it was supposed to be scalable18:52
FatPhilIpso facto, it's designed and implemented by people who don't know what they're talking about.18:52
FatPhilNot a good sign.18:52
*** NIN101 has joined #maemo-ssu18:53
DocScrutinizer05Mickey Lauer (architect of FSO) more than one time moaned "I'm going to write by own IPC now, nuke d-bus from the surface of God's great earth"18:53
DocScrutinizer05due to the aforemantioned bug, fso was not using async anywhere for quite some time, which of course made life a lot easier ;-)18:54
freemangordonPali: hmm, why is AVS enabled on your linux-n900 tree?18:56
*** arcean has joined #maemo-ssu18:57
freemangordonhmm, mayde it is time to move CSSU to github, gitorious is simply unuisable :(18:58
freemangordon*maybe18:58
*** sunny_s has joined #maemo-ssu19:02
freemangordonPali: https://gitorious.org/linux-n900/linux-n900/source/d5da1f5357345ad270a491ce285c78dcf901a7ff:arch/arm/configs/rx51_defconfig#L31819:04
DocScrutinizer05hmm, can't spot it anymore, and coldn't a year ago either: http://irclog.netripper.com/openmoko-cdevel/2012/8/4/19:08
*** DrCode has quit IRC19:09
*** DrCode has joined #maemo-ssu19:21
dos1I remember that bug19:21
*** M4rtinK has quit IRC19:23
ShadowJKwhat's wrong with using malloc? I even use it on systems with 2k ram :s19:26
FatPhilShadowJK: my comment was in the context of http://gypsy.freedesktop.org/why-not-gpsd.html19:35
dos1DocScrutinizer05: https://bugs.freedesktop.org/show_bug.cgi?id=1979619:36
povbot_Bug 19796: was not found.19:36
ShadowJKIn some ways they've got a point, in that over time for long running processes, heap fragmentation can get pretty bad19:37
*** mkaindl has quit IRC19:39
*** Martix_ has joined #maemo-ssu19:39
DocScrutinizer05dos1: EXACTLY, how the heck did you manage to find this one?19:39
DocScrutinizer05ShadowJK: solution: allocate your *static* vars during _init_19:41
*** dafox has joined #maemo-ssu19:41
DocScrutinizer05the best of both concepts19:41
dos1DocScrutinizer05: I recalled reading it on freedesktop bugzilla, so I just grepped my irc logs for freedesktop ;)19:42
DocScrutinizer05hah!19:42
*** i_a_delta_GB has joined #maemo-ssu19:42
DocScrutinizer05NICE; still nit "fixed"19:48
DocScrutinizer05not*19:48
DocScrutinizer05I think it's a WONTFIX since evrybody got used to d-bus blowing chunks every now and then19:50
DocScrutinizer05I frequently notice effects in KDE where it _clearly_ lost some message or event. Simply doesn't react on an event in the way it's supposed to. Not now and not after 30 minutes19:52
dos1that's ridiculous...19:52
DocScrutinizer05I'm also rather sure d-bus is memleaking19:53
dos1init systems on top of dbus starts to sound now like really great idea...19:56
dos1I wonder how kdbus project is going on19:56
merlin1991you mean the inkernel messaging system?19:58
dos1yep19:58
*** mkaindl has joined #maemo-ssu19:59
*** mkaindl has left #maemo-ssu20:00
merlin1991I pretty much heard nothing about it after the initial "we have a system for this already in the kernel" response20:00
merlin1991didn't look for it though20:00
*** Martix_ has joined #maemo-ssu20:02
*** BCMM_ has quit IRC20:04
dos1I remember that systemd devs had big hopes for it20:04
FatPhilsystemd devs had high hopes about kitchen sinks too20:05
FatPhilEventually, the answer to the question "What OS do you run?" would be "systemd"20:05
*** Martix_ has quit IRC20:07
*** luf has joined #maemo-ssu20:07
keriofucking systemd20:09
FatPhilShadowJK: one way around fragmentation is to give out handles to memory, rather than pointers. That way, a sufficiently intelligent allocator could optimise their location. (Which of course requires devs to lock and unlock around accesses, which history tells us is expecting too much from some developers)20:09
kerioiirc linus is a bit tired of poettering's shit20:09
kerioanyway, mkfs.ext4 -c is at 828 errors at 58% on this netbook's hd :s20:10
FatPhilEveryone's tired of it. However, he's no idiot. Plenty of what he says makes good sense.20:10
ShadowJKI remember when I had bash running inside xchat. My OS was IRC20:10
kerioi'm looking forward to the end of these bad blocks :s20:10
*** i_a_delta_GB has quit IRC20:10
kerioShadowJK: pffff20:10
keriowhat a noob20:10
FatPhilMy OS was emacs for a while20:11
kerioinit=/usr/bin/emacs20:11
freemangordonFatPhil: couldn't that lock/unlock be done automatically for those who are lazy? or just sigsegv?20:11
kerioand then run bash inside ERC20:11
*** Martix_ has joined #maemo-ssu20:11
*** arcean has quit IRC20:11
*** mkaindl has joined #maemo-ssu20:11
FatPhilfreemangordon: I wouldn't want to design such a system, and then roll it out in a high availability environment, put it that way.20:11
*** mkaindl has left #maemo-ssu20:12
FatPhilfreemangordon: of course, it can be done in a language like C++, with polymorphism.20:12
freemangordonFatPhil: well, I always think C++. or rather C with classes :D20:12
FatPhilbut then you have to auto-lock/unlock around every individual access which is teh suck20:13
freemangordonI am old enough to thing anything but C/C++ should be forbidden by the law inside kernel/low level libs/middleware20:13
FatPhilI have many years of being a C++ evangelist on my CV... Never touch the language mowadays.20:13
*** iDont has joined #maemo-ssu20:14
merlin1991freemangordon: but, but python? ;)20:15
dos1C++ sucks :P20:16
dos1"let's take C, add some nice things to it, but screw up many more" ;)20:17
FatPhilTotally agreed.20:17
FatPhilI was thinking of writing some performance-oriented highly-OO code, and looked around for an alternative to C++, and found Digital Mars D.20:18
FatPhilHowever, I looked away for about a year, and when I turned back, I noticed that the same fucking idiot who had introduced a megaton of shit into C++ had now started fucking up Digital Mars D too.20:19
FatPhilSo that project is still on hold...20:20
*** Martix_ has quit IRC20:20
FatPhilIf you're introducing class schemas into a language, that you are deprecating only a couple of years later, then you are *dangerous* to the language.20:21
freemangordonmerlin1991: what about python?20:24
freemangordondon't tell me you like the way you are forced to use tabs(or spaces), come on, this is ridiculous20:25
merlin1991well the whitespace thingy is odd20:25
merlin1991but my other likes outweight that20:25
freemangordonand is enough to make me don;t want to touch anything python with a stick20:26
FatPhilI've never looked at python closely enough to evaluate its other features20:26
freemangordonsame here :)20:26
FatPhilSemantic meaning to lengths of runs of whitespace is such an absolute no-no, it should never have been written in the first place.20:27
ShadowJKRight now I'm happy when I get anything less silly than Siemen's "beginner's ladder" language. You graphically draw logic gates and connect them together.20:28
FatPhilI remember looking at the archives of the python programming yahoogroup once - it died out very quickly, as none of the members could post any code - it would all get mangled!20:28
FatPhildataflow languages can be awesome20:29
*** discopig has quit IRC20:29
ShadowJKI often end up "I wish I had a clock", then take a nand gate and connect the output to input, and then start constructing a cpu-like monstrosity but run out of gates before getting where I wanr20:30
FatPhilThe best thing about that yahoogroup was that some people had replaced runs of initial whitespace with runs of periods in order to preserve the indentation.20:30
ShadowJKthe hardware does 150 cycles per second!20:30
FatPhilstockholm syndrome20:30
ShadowJKrofl20:30
ShadowJKThe only good thing about that siemens gear is that each unit has a beancounter friendly price of <200 euro, which is amazing for industrial automation kit20:32
FatPhilcan't you get some Xylinx or Altera boards for less than that?20:33
ShadowJKboards, yeah20:34
ShadowJKbut not with 10A 12-240V I/O drive strength and foolproofing against "hook up random wires to random terminals and see what happens"-electricians20:35
FatPhilGood point.20:35
ShadowJKthis seems to be the common troubleshooting strategy when things stop working20:36
ShadowJKcomes after hitting it wth a hammer and turning it off and then on again.20:36
ShadowJKsorry, I mean "percussive maintenance"20:36
*** DrCode has quit IRC20:45
*** DrCode has joined #maemo-ssu20:49
*** BCMM_ has joined #maemo-ssu20:56
*** BCMM_ has quit IRC20:56
*** BCMM_ has joined #maemo-ssu20:56
*** NIN101 has quit IRC21:01
*** NIN101 has joined #maemo-ssu21:11
*** BCMM_ has quit IRC21:14
*** BCMM_ has joined #maemo-ssu21:16
*** NIN101 has quit IRC21:27
*** nox- has joined #maemo-ssu21:27
kerioc++ is just a bad idea overall21:52
*** arcean has joined #maemo-ssu22:03
*** FlameReaper has joined #maemo-ssu22:10
*** arcean has quit IRC22:10
freemangordonkerio: yeah, python, java and .NET are waaay better :P22:12
keriopython is really good for smallish-to-medium-sized projects22:12
keriojava is excessively enterprisey22:12
kerioand .net is java22:12
keriobut made by microsoft22:13
freemangordonyou mean it is bytecode?22:13
kerioit fits the same niche as java22:13
kerioaka enterprise programs22:13
freemangordonyou mean "take the next moron from the street and make him developer" paradigm?22:14
kerios/the street/India/22:14
freemangordonbut why, there are streets all over the world, and I doubt morons can only be found in India22:15
kerioi mean, java is kinda made to avoid too many issues caused by bad programmers22:15
freemangordonactually I am sure there are lots of the inside India22:15
kerioeverything is strictly divided in small parts22:16
freemangordon*them outside22:16
freemangordonkerio: ever seen a very big project coded in java?22:16
kerioa bad programmer working in python or ruby will happily change builtins because they have copypasted a cool trick from stockoverflow22:17
keriono, i have not22:17
kerioand i feel happier because of it22:17
freemangordonit is a nightmare to deal with the codebase. at least for me. It takes ages to find all the parent classes, all the method overwrites, etc22:18
keriooh absolutely22:18
*** NIN101 has joined #maemo-ssu22:50
merlin1991nah .net isn't take the next moron23:13
merlin1991but at the same time it is in some way23:14
FatPhilMuch of it is laziness on the part of educators - they're just teaching whatever the current trend is, and the trend is driven my marketting.23:16
FatPhilAnd I say that as someone who's been on the teaching side of things23:17
*** iDont has quit IRC23:17
*** _rd has joined #maemo-ssu23:20
merlin1991well tu vienna teaches java23:20
merlin1991not exactly the latest marketing trend :D23:20
*** BCMM_ has quit IRC23:33
*** LauRoman has joined #maemo-ssu23:43
*** luf has left #maemo-ssu23:58

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!