**** BEGIN LOGGING AT Sun Oct 28 03:00:03 2012 Oct 28 03:04:54 DocScrutinizer06: got the USB charger open, but shorting random pins on the USB port seems to do nothing Oct 28 03:05:18 err, random pinds? Oct 28 03:05:27 well, dunno which ones are D+- Oct 28 03:05:36 you should short D+ to D-, aka the middle two Oct 28 03:05:44 there's no obvious correlation on the PCB Oct 28 03:06:01 ooh Oct 28 03:06:03 found it maybe Oct 28 03:06:04 there's not? how's that? Oct 28 03:06:20 USB port is horizontal, pins are vertical? Oct 28 03:06:23 or in a square Oct 28 03:06:27 not even sure which ones are which Oct 28 03:06:30 all USB receptacles I've seen had 4 pins in a row Oct 28 03:06:57 you see the USB receptacles? Oct 28 03:07:08 yes Oct 28 03:07:11 the solder points are right to the back Oct 28 03:07:34 ⁇? Oct 28 03:08:17 on backside of each receptacle there are 4 pins in a row Oct 28 03:08:45 if you consider the side where you insert the plug as fronstside Oct 28 03:09:02 ok Oct 28 03:09:03 I wouldn't know how to describe it any better Oct 28 03:09:43 is it possible the N900 got pissed at my shorting at random and decided to stop looking for a charger? <.< Oct 28 03:09:56 wait what Oct 28 03:09:58 I haven't ever seen any 4pin fullsize USB receptacles that were any ambiguous regarding pins Oct 28 03:10:09 I replugged it to reset any such logic… and it said Charger disconnected -.- Oct 28 03:10:31 looks like of the 4 in a row, the right two are D? Oct 28 03:11:08 the MIDDLE two!!! Oct 28 03:11:40 but only shorting the right two works.. :/ Oct 28 03:11:55 no matter from which side you look it's always the middle two that are data Oct 28 03:12:34 meh, great. Try&error on a $$$ device Oct 28 03:13:00 you're awatre that USB data is 3V3 though VBUS is 5V ? Oct 28 03:13:08 someone told me the worst that happens is I break the USB charger :| Oct 28 03:14:00 the one who told you that for sure didn't anticipate you're doing random insane stuff Oct 28 03:14:24 you may have a point there <.< Oct 28 03:14:30 data lines should be OVP for 5V Oct 28 03:15:07 hmm, maybe I'm just sucky at shorting the pins I intend to Oct 28 03:15:23 in any case, I think my 93% full is too full to test speed now :< Oct 28 03:15:25 connecting/shorting USB data to any point of mains input of charger might kill some stuff though Oct 28 03:15:35 and I lack any conductive material to make a perm connection with Oct 28 03:16:06 DocScrutinizer05: do you ever sleep? :) Oct 28 03:16:06 should have had wire glue around here, but it seems to be lost :< Oct 28 03:16:18 you should just happily assume it is all ok, and short the middle two pins of USB with a solder blob Oct 28 03:16:45 jon_y: sleep is overrated Oct 28 03:16:55 is your nick shared by multiple users? Oct 28 03:17:24 nope, multiple nicks are shared by me ;-P Oct 28 03:17:30 or doing the "overman" sleep Oct 28 03:17:39 I had to re-install the scratchbox-based SDK for Maemo 5. The crosscompiling stuff all works, and the proprietary Nokia stuff works, but dh)make is missing, so I can't build a .deb . Does anyone khow where I can get that from, and why it would not be present in the normal SDK installation? Oct 28 03:18:09 s/dh)make/db_make Oct 28 03:18:52 Ken-Young: I couldn't recall anybody complaining about some basic stuff like that missing in maemo SB. So maybe you're looking for the wrong thing? Oct 28 03:18:56 DocScrutinizer05: https://en.wikipedia.org/wiki/Polyphasic_sleep uberman sleep cycle? :) Oct 28 03:19:11 Six 20-minute naps (every 4 hours) Oct 28 03:19:19 2 hour sleep a day Oct 28 03:19:39 I sleep while not typing Oct 28 03:19:56 a few seconds between each two posts Oct 28 03:20:06 DocScrutinizer05, Well, I've done a search of every file under /scratchbox, and nothing called db_make is there. Oct 28 03:20:29 Ken-Young: maybe it's not called db_make then Oct 28 03:20:59 DocScrutinizer05, It sure was when I originally installed the SDK, a couple of years ago. Oct 28 03:21:07 Ken-Young: though you have to ask somebody with a tad more recent experience in using SB Oct 28 03:21:38 DocScrutinizer05: what build env do you use? Oct 28 03:21:47 I've been searching TMO threads. Some people mention this problem, but I haven't seen a solution yet. Oct 28 03:21:50 I'm not building stuff usually Oct 28 03:22:23 what is the most common system? SB? Oct 28 03:22:42 Ken-Young: you better wait for somebody using SB and not haviong the problem, then ask him how he's building debs Oct 28 03:22:52 maybe there's build_deb Oct 28 03:23:04 btw, is the N900 armv7? Oct 28 03:23:05 or whatever else Oct 28 03:23:17 jon_y: basically yes, I'd think Oct 28 03:23:25 but no hardfloat? Oct 28 03:23:56 yep, we have hardfp aka numbercruncher on cortex-A8 Oct 28 03:24:09 afaik Oct 28 03:24:25 oh ok, so it is possible to stick debian armhf into easychroot? Oct 28 03:24:38 should? Oct 28 03:24:57 iirc the normal armel platform is softfloat Oct 28 03:25:16 there's a normal armel platform? ;-) Oct 28 03:25:25 well, debian armel Oct 28 03:25:31 aaah Oct 28 03:25:33 from easydebian Oct 28 03:25:37 you meant the sw Oct 28 03:26:00 yeah Oct 28 03:26:51 softfp vs hardfp is just an API convention, where to pass floats, if in CPU domain or in float registers - afaik Oct 28 03:27:17 well, that and some hardware support for fp computations with hardfp Oct 28 03:27:22 basically both are lib stubs to the numbercruncher Oct 28 03:27:48 at least that's what they told me Oct 28 03:28:03 only hardfp takes advantage of the hardware support though Oct 28 03:28:17 and meego-CE as of stskeeps decided to go for hardfp on N900 Oct 28 03:28:18 softfp float computation is done in software, really really slow Oct 28 03:28:33 can confirm the armv7h Oct 28 03:28:39 nope, that's unrelated afaik Oct 28 03:28:52 meego-CE? Oct 28 03:29:04 even softfp may use numbercruncher Oct 28 03:29:18 DocScrutinizer05, Really, softfp does not mean software floating point (not using CPU float instructions)? Oct 28 03:29:20 the real meego, now mer Oct 28 03:29:49 I thought softfp meant emulation with integer math Oct 28 03:30:04 your mantissas and what not Oct 28 03:30:11 I also thought that, until stskeeps educated me Oct 28 03:31:05 ok, so softfp using numbercruncher depends on the math library? Oct 28 03:32:17 but as of now my knowledge says both are in a lib (probably libc), and the hardfp vs softfp only says how to invoke the lib functions for sin() div-float() etc. While the lib could use software emulation with softfp, but also could use numbercruncher with softfp Oct 28 03:32:28 yep Oct 28 03:32:42 ok, makes sense Oct 28 03:33:07 though I suppose float asm insn is only present in hardfp Oct 28 03:33:34 Hm, could we use GPU or DSP for some of the floating point calcs? Oct 28 03:33:56 I suppose you could if you can program them Oct 28 03:34:09 no, there's a dedicated numbercruncher afaik, called neon(?) Oct 28 03:34:34 I thought neon was some kind of SIMD thing? Oct 28 03:34:53 that's why there's a (?) behind it Oct 28 03:34:59 ok Oct 28 03:35:45 I'm not aure about the hw float support Oct 28 03:35:48 sure Oct 28 03:36:09 what I'm sure about is the diff between hardfp and softfp Oct 28 03:37:17 and since hardfp means "pass the floats in the float registers" ther have to be some float registers which wouldn't make sense if there's no float ALU to do something with those float registers Oct 28 03:38:55 ok, so N900 is using the softfp ABI, but has hardware support? Oct 28 03:39:23 no, maemo is using softfp ABI, mer/meego is using hardfp Oct 28 03:39:57 if this didn't change since I last heard about it Oct 28 03:40:05 ok, understood Oct 28 03:40:16 chroot isn't going to translate between ABIs Oct 28 03:40:26 nope Oct 28 03:40:40 I don't think debian armel is going to take advantage of the hw number cruncher Oct 28 03:40:59 but we recently talked about it and found that probably there's no FP gateways in the chroot borderline Oct 28 03:42:05 do you need the kernel to trap them? Oct 28 03:42:37 ypur whole system has to be compiled for either hardfp or softfp, but since all your 'system' lives inside chroot, a chroot hosting a hardfp could live on a host running softfp Oct 28 03:43:00 the float math is done in a lib, not in kernel Oct 28 03:43:17 floats are even deprecated or rather forbidden in kernel land Oct 28 03:43:40 the lib is part of your chroot Oct 28 03:43:56 guys, http://gcc.gnu.org/onlinedocs/gcc/ARM-Options.html Oct 28 03:44:12 so? Oct 28 03:44:21 ah right, I forgot the libc for debian also lives in chroot Oct 28 03:44:46 hoepfully there aren't any problems with context switching etc Oct 28 03:45:06 I still remember the win98 days and sse, there may only be 1 SSE program at a given time Oct 28 03:45:29 the OS wasn't sse aware and did not save/restore the XMM registers on context switching Oct 28 03:45:47 I'd guess context switching should take care of the float registers as well Oct 28 03:46:08 well, does the N900 kernel do that? :) Oct 28 03:46:37 does it need to be aware of the float registers? or are they shared with the gp registers? Oct 28 03:46:58 good question, but since I never heard stskeeps mentioning a hardfp kernel for mer/meego, I assume kernel usually cares about float regs as well Oct 28 03:47:08 in the sse case XMM was a different register than the general purpose ones, so OS needs to be awre of them Oct 28 03:47:23 ok Oct 28 03:47:23 DocScrutinizer05: http://luke.dashjr.org/tmp/pics/20121027_010.jpg Oct 28 03:48:05 looks good Oct 28 03:49:21 there are no floats in arm kernel so the hardfp kernel people keep asking about is irrelevant and you need to hack Makefiles to even achieve that. Oct 28 03:49:41 though I'm worried about those resistors which are next to /under the USB receptacles. They look like they implement another charger standard, compatible to apple etc Oct 28 03:50:28 Skry: please rephrase Oct 28 03:51:09 Luke-Jr: I'd suggest you don't short all 4 USB receptacles like that, keeping some of them for apple products Oct 28 03:51:17 Skry: the kernel does not need to be aware of the registers? Oct 28 03:51:40 eg hardfp and softfp registers are the same, maybe placement is different? Oct 28 03:52:00 actually are there any difference in register use? Oct 28 03:57:23 >> The public ARM Cortex-A8 processor incorporates the technologies available in the ARM7 architecture. These technologies include Neon for media and signal processing and Jazelle RCT for acceleration of realtime compilers, Thumb-2 technology for code density and the VFPv3 floating-point architecture. << Oct 28 03:58:50 meh, just try to pass CFLAGS="-mfloat-abi=hard" to compiler when building kernel and look if it's actually used. Not. Anyways, there is no gain in hardfp kernel. Oct 28 04:01:16 and yes you can run hfp in chroot on sfp system, incompatibilites come with linking libraries of different abi Oct 28 04:02:00 ok, kernel support not needed Oct 28 04:02:36 and if I recall correctly, kernel does not have that much floats so vfp is sufficient Oct 28 04:03:22 Skry: I mean does the kernel actually need to know about vfp specifically for the libc to use hardfloat ABI? Oct 28 04:04:21 Skry: there's obviously no sense in passing float-abi=hard to kernel, since nothing is passing any floats to kernel. Nevertheless that's not been the question in the end Oct 28 04:05:37 the hardfp ABI is to libc, which is not used by kernel afaik Oct 28 04:06:43 still the question is if kernel (particularly scheduler) will take care about VFPv3 registers on task switching Oct 28 04:06:57 yeah, will it foul up like the SSE case in win98? Oct 28 04:08:54 afaik it does take care of Oct 28 04:09:12 and sorry, read backlog very briefly, kinda busy here Oct 28 04:09:24 jon_y: IroN900:~# cat /proc/cpu ->Features : swp half fastmult vfp edsp neon vfpv3 Oct 28 04:09:47 ok, cpu does know about vfp Oct 28 04:10:12 and Skry says the kernel task switcher takes care of the vfp regs Oct 28 04:10:29 It would be more than insane if it doesn't Oct 28 04:10:42 looks like no trouble at all in chroot hardfp as long as you don't mix the libc Oct 28 04:10:53 :nod: Oct 28 04:11:01 eg don't call easydebian stuff from outside the chroot Oct 28 04:11:38 DocScrutinizer05: Apple products are broken? Oct 28 04:12:04 Luke-Jr: they just use another equally ligitimate charger spec Oct 28 04:12:32 something like "D+ = 0.66*VBUS; D-=0.33*VBUS" Oct 28 04:13:18 which looks exactly like what those resistors are implementing Oct 28 04:14:12 a 2R/R devider for one D line, a R/2R devider for the other Oct 28 04:14:33 wtf? two incompatible standards? :/ Oct 28 04:15:13 make sure to get BIG resistors, so load don't particularly affect the voltage :) Oct 28 04:15:33 yeah, and both introduced resp certified by USB cert Oct 28 04:15:34 maybe I forgot my basic electronics Oct 28 04:15:52 ugh Oct 28 04:16:10 afaik Oct 28 04:17:06 at least D+- short *is* certified by USB cert foundation Oct 28 04:17:20 or whatever the name of those xxxxx Oct 28 04:17:45 not entirely sure about the "apple standard" Oct 28 04:18:45 Luke-Jr: actually you should unsolder/remove the 4 resistors on the shorted USB port Oct 28 04:19:10 DocScrutinizer05: does it really matter? Oct 28 04:19:13 they easily might spoil the D+- short detection Oct 28 04:19:14 already put it back together.. Oct 28 04:19:22 iirc "Apple standard" was formulated before the d+- short thing Oct 28 04:19:31 they needed more power than the standard USB provides Oct 28 04:19:53 some sort of signaling mechanism to say it's ok to draw more than the normal current Oct 28 04:20:11 that's what D+- short does as well Oct 28 04:21:00 Apple did it for their iPods Oct 28 04:21:09 before all this was standardized Oct 28 04:22:42 actually N900 and other similar devices are supposed to detect D+- short by attaching a pullup to one D-line, a weak pulldown to the other, and then check if the line with weak pulldown is actually at high level. So somehow the both standards might be partially compatible under certain preconditions Oct 28 04:23:54 (or maybe swap "up" and "down" in my last statement, I can't remember) Oct 28 04:25:01 Luke-Jr: ^^^ that's why the resistors might interfere with N900's internal detection of D+- short Oct 28 04:25:50 DocScrutinizer05: in practice, it doesn't Oct 28 04:26:01 Luke-Jr: the D+- standard requires none of the D-lines conected to any other voltage level, not via resistors nor directly Oct 28 08:39:21 yay, chanserv is back ^_^ Oct 28 08:39:24 and you're all back too! Oct 28 11:24:55 yoh, moinmoin Oct 28 11:25:40 moin Oct 28 11:26:51 morning Oct 28 11:31:03 WTF? PA in -thumb uses more CPU than stock. Something is wrong here Oct 28 11:35:22 ...wat Oct 28 11:35:38 yeah Oct 28 11:35:49 freemangordon: try disabling thumb and compiling with gcc4.7 Oct 28 11:35:55 will do Oct 28 11:36:20 but first want to see why is that Oct 28 11:39:38 DocScrutinizer05: D: Oct 28 11:43:45 is cpufreq/governour a usual process like any other? I.E. will it get scheduled and switching of cpufreq only happens when governor is the active&running process? Oct 28 11:44:38 anyway, freemangordon: use powertop Oct 28 11:45:18 DocScrutinizer05: why? the difference is clearly visible. frequency locked on 250 Oct 28 11:45:27 top etc are basically useless for in-depth and really meaningful investigations of cpu usage of a process Oct 28 11:45:33 ooh Oct 28 11:45:39 I bet it is because -O3 -ffast-math Oct 28 11:45:50 with locked freq that's a tad different indeed Oct 28 11:46:07 what's fast-math? Oct 28 11:46:10 gcc is still not good enough doing NEON optimizations Oct 28 11:46:27 enable auto-vectorizer Oct 28 11:46:34 mhm Oct 28 11:46:37 i'll recompile with -o@ Oct 28 11:46:42 -O2 Oct 28 11:47:20 trying to do math in vector processor might be a poor idea Oct 28 11:47:26 for audio Oct 28 11:47:32 and -ffast-math is often slower Oct 28 11:47:46 that's what I meant Oct 28 11:47:56 in theory that should speed up floating point a lot Oct 28 11:48:06 the math in audio isn't exactly vector math usually Oct 28 11:48:07 BTW I am using stock compiler flags Oct 28 11:48:47 "Disabling ‘fast-math’ not only solved the consistency problem, it made Avida faster in certain cases and had no effect in others." Oct 28 11:49:05 which gcc? Oct 28 11:49:13 it's a 2009 article, so idk Oct 28 11:49:23 that is old Oct 28 11:49:30 though I saw the same with openssl Oct 28 11:49:56 enabling tree-vectorizer and ffast-math made openssl slower Oct 28 11:50:16 by a 2-3% but still Oct 28 11:53:54 freemangordon: are you sure you're not confusing the logs, btw? Oct 28 11:54:08 which logs? Oct 28 11:54:32 well, the results Oct 28 11:54:48 anyway, -ffast-math does some *nasty* things Oct 28 11:58:10 kerio: -O3 along with -ffast-math allows gcc to use NEON intrinstics Oct 28 11:58:30 or better said tells gcc to use them Oct 28 11:59:04 well it's clearly not working as intended Oct 28 11:59:24 wel,, it works as inteneded, just a bit slower Oct 28 11:59:44 stock PA is compiled with the same flags Oct 28 12:00:22 but as 4.7 has way better support for NEON, it "optimizes" much more code to use NEON Oct 28 12:00:27 everyone says that gcc is crap at using NEON intrinsics :s Oct 28 12:00:40 and the result seems to be slower code Oct 28 12:00:41 yah Oct 28 12:00:45 *yeah Oct 28 12:00:57 i say asm dat shit Oct 28 12:01:40 naah, i'll disable fast-math and will -O2 instead of -O3 for start Oct 28 12:01:49 after having something to eat :D Oct 28 12:18:28 how about some decent profiling tool? Oct 28 12:18:46 not that I could name any specific one right away Oct 28 12:35:00 * ShadowJK 'd cry more about pa if it used floating point for processing audio Oct 28 12:36:49 also try -Os :D Oct 28 12:37:22 ShadowJK: wait, it doesn't? Oct 28 12:38:36 it's written by lunatics so it probably does Oct 28 12:38:40 but i dunno Oct 28 12:43:39 found something mentioning resampling is fixed point atleast Oct 28 12:44:08 ShadowJK: wouldn't floats make it faster? Oct 28 12:44:16 no? Oct 28 12:44:49 apparently it's neon optimised version of speex's integer resample :) Oct 28 12:46:59 loosely related: http://tech.groups.yahoo.com/group/twinklephone/message/4013 Oct 28 12:47:14 at least a confirmation od Oct 28 12:47:26 So if the neon code didnt get compiled with newer gcc4.7, or if the neon code was written with intrinsics and new gcc does something random in a different way now Oct 28 12:47:41 of "shadowJK: it's written by lunatics" Oct 28 12:48:03 can't we write a PA emulation layer for alsa? Oct 28 12:48:04 :D Oct 28 12:48:08 kerio; was it you with the xplosm cartoon edit of PA? Oct 28 12:48:15 ShadowJK: it's not mine Oct 28 12:48:19 i got it on #redeclipse Oct 28 12:48:28 well do you still have link Oct 28 12:48:39 ShadowJK: /topic :) Oct 28 12:48:43 oh :) Oct 28 12:48:54 also wtf, this is not #maemo-ssu Oct 28 12:49:20 not in topics? Oct 28 12:49:24 ShadowJK: http://i.imgur.com/SAp0D.png Oct 28 12:49:34 well, we're discussing the cssu-thumb recompile of pulseaudio... Oct 28 12:49:40 :D Oct 28 12:53:32 * ShadowJK bookmarks it Oct 28 12:54:11 freemangordon: benchmarked PA without -ffast-math yet? Oct 28 12:57:39 author: fakjujahuu X-P Oct 28 12:57:57 why didn't I think about that? Oct 28 12:59:43 DocScrutinizer05: huh? Oct 28 12:59:45 (( also wtf, this is not #maemo-ssu)) no worries, some users wouldn't ever know about the awesome stuff we do at CSSU, otherwise ;-) Oct 28 13:00:02 nobody's complaining Oct 28 13:00:45 kerio: http://tech.groups.yahoo.com/group/twinklephone/message/4013 author: fakjujahuu Oct 28 13:01:17 it makes sense once you say tehe author's name out loud Oct 28 13:02:26 c: Oct 28 13:05:19 [general notice] If you could be bothered to contribute to maemo in this very particular way: Council Elections nomination period is closing today - I'd encourage you to run for a seat in maemo council! Oct 28 13:05:38 DocScrutinizer05: should i run for council? Oct 28 13:05:44 yes!! Oct 28 13:05:53 well too bad, i'm not gonna Oct 28 13:06:01 :-( Oct 28 13:06:01 mwahaha Oct 28 13:06:23 i don't even know what the council is supposed to do Oct 28 13:06:55 that's pretty much up for a fundamental redefinition right now Oct 28 13:07:50 I'd think it's basically useless now that we got hildon foundation, and probably should get discontinued / merged with HF Oct 28 13:08:12 what is hildon foundation? Oct 28 13:09:25 the non profit foundation that you were supposed to elect for, some 2 weeks ago. It's for taking responsibility regarding e.g continuation of servers for *.maemo.org, owning the domain name from Nokia, etc Oct 28 13:10:34 pretty much a lot of legalese and organizational stuff that only a legal entity could do Oct 28 13:11:55 while council been established to play ambassador between community and Nokia, so now between community and HF, which doesn't make much sense in my book Oct 28 13:12:53 since, unlike Nokia, HF is a non profit organization and got elected by community just like council Oct 28 13:13:44 A new version of the orrery app for the N900 has been uploaded to the Extras Testing repository. This new version will display the two comets which may be bright in 2013. Oct 28 13:14:06 so council are community's proxies to negotiate between community and community's proxies Oct 28 13:14:46 Ken-Young: \o/ great Oct 28 13:15:52 kerio: I'd think you pretty much could help on that daunting task (community proxy) Oct 28 13:21:52 but i hate the community! Oct 28 13:34:26 kewl Ken-Young :) Oct 28 13:34:46 kerio, you sure? ;) Oct 28 13:34:57 yes! Oct 28 13:35:00 especially you! Oct 28 13:35:03 :) Oct 28 13:37:39 btw, have you ever used jSolun? it's java-based.. but has some really nice features. it's one of those applications i miss on N900 Oct 28 13:37:42 hey gry :P Oct 28 13:38:14 hello, Sicelo! Oct 28 13:38:17 and night :) Oct 28 13:39:11 still have your N900? Oct 28 13:39:19 Sicelo, I had never heard of that program. I'll have to take a look at it. Thanks! Oct 28 13:44:41 * SpeedEvil goes to reinatall orrery Oct 28 13:56:28 (( also wtf, this is not #maemo-ssu)) no worries, some users wouldn't ever know about the awesome stuff we do at CSSU, otherwise ;-) Oct 28 13:56:43 wtf, there is some awesome stuff I don't know about? Oct 28 13:56:50 akls: yep Oct 28 13:57:02 i don't know if you know about them, though Oct 28 13:57:07 like what? Oct 28 13:57:34 like CSSU Oct 28 13:57:39 and CSSU-thumb Oct 28 13:58:14 essential stuff Oct 28 13:58:23 and a slower pulseaudio! Oct 28 13:58:25 no, wait Oct 28 14:00:15 oh, by the way Oct 28 14:00:19 about pulseaudio... Oct 28 14:01:06 when listening music through cmus player music stops playing when I hit power button twice to lock the screen Oct 28 14:01:23 and it seems like with native player there's no such problem Oct 28 14:01:35 is it some problem with cmus process priority or what? Oct 28 14:01:42 akls: probably the policy daemon Oct 28 14:01:49 so... tough luck Oct 28 14:02:57 it pauses for a moment with mplayer too Oct 28 14:08:29 ShadowJK: is it ok to you if I distribute your charge21.sh which I modified to some extent in my Arch packages? Oct 28 14:24:27 Skry: as long as some (c) is kept I'd guess it's ok Oct 28 14:24:38 ofc Oct 28 14:28:03 oh my, the latest WTF from canonical, they'll be pushing out SDK for ubuntu Oct 28 14:31:16 Skry: SDK = ? are they releaseing their own toolchain or is it something ubuntu distro related? Oct 28 14:31:23 for distro development that is Oct 28 14:33:00 apparently for "application development" Oct 28 14:33:36 http://www.phoronix.com/scan.php?page=news_item&px=MTIxNjU Oct 28 14:34:42 like... gcc? Oct 28 14:35:23 probably enforcing some specific versions of compiler and libs Oct 28 14:35:40 and ofc everything buntu-patched Oct 28 14:36:24 but WHY? Oct 28 14:36:25 ... Oct 28 14:36:29 they are breaking linux Oct 28 14:36:45 been doing that for awhile Oct 28 14:37:10 ms applenical Oct 28 14:43:30 Skry: it better be running on Windows :) Oct 28 14:43:39 like android emulators Oct 28 14:43:59 wouldnt surprise me :) Oct 28 14:44:01 heh, maybe it's really ubuntu in a vm image Oct 28 14:44:19 oh, and MSVC integration Oct 28 14:44:36 corporate drones love MSVC Oct 28 14:44:44 yeah thats a must Oct 28 14:45:27 so, they build their "solutions" and upload the image to the vm Oct 28 14:45:41 sounds so complicated when written that way :) Oct 28 14:46:31 fuck ubuntu Oct 28 14:48:12 ubuntu used to be nice, until they dissed all the standard stuff and doing things the MS way Oct 28 14:48:26 indeed, fuck ubuntu Oct 28 14:48:31 my favourite distro is mint, now :3 Oct 28 14:48:37 well, if they want to take the android route so be it, good riddance. Oct 28 14:48:39 for "user-friendly" things Oct 28 14:48:43 * jon_y uses Debian Oct 28 14:48:46 erm, mint is ubuntu Oct 28 14:49:05 except without the stupid annoying UI, so one step above ubuntu Oct 28 14:49:12 true Oct 28 14:49:12 Based on is not "is" Oct 28 14:49:24 please Oct 28 14:49:45 Just because they use Ubuntu repositories doesn't mean they're beholden to the same architecture decisions Ubuntu makes. Oct 28 14:49:59 GeneralAntilles: it uses upstart, though Oct 28 14:50:11 mint debian edition is probably better in that regard Oct 28 14:50:24 what is upstart? Oct 28 14:50:31 I've not really had issues with Upstart. Oct 28 14:50:34 what so different than the normal sysvinit? Oct 28 14:50:56 jon_y: it's one step closer to systemd Oct 28 14:51:14 totally forgot debian edition exists, well, that's approved :) Oct 28 14:51:15 sorry, I haven't been keeping up with all the startup systems Oct 28 14:51:29 so, sysvinit, upstart, systemd Oct 28 14:53:19 what is the difference between them? Oct 28 14:53:57 anyways, mint is pretty usable, excellent beginner distro, hopefully they take some distance to ubuntu Oct 28 14:54:11 hmm, PA with -O2 and -fno-fast-math seems to be at least as fast as stock Oct 28 14:55:08 jon_y: sysvinit is the old monolithic beast that the new monolithic beast systemd is going to deprecate Oct 28 14:55:23 what does it do differently? Oct 28 14:55:40 jon_y: pretty much everything Oct 28 14:56:50 jon_y: ie. it does not have runlevels but targets which are similar concepts though Oct 28 14:57:18 it also has socket based activation of services and stuff Oct 28 15:00:16 besides doing shit the non-traditional way, making some scary changes, when you get into it it's really great Oct 28 15:01:29 obviously one can disagree and take it as an abomination, and I can understand why, but I can also see why it is being adopted so widely already Oct 28 15:01:44 ok, does it have the old runlevel-like functionality? Oct 28 15:02:06 eg, I want to shutdown networking capability temporary Oct 28 15:02:14 yes, you can switch targets and select default targets Oct 28 15:02:38 but if you want to shutdown network temporarily you dont need to switch targets, you just stop the service Oct 28 15:02:53 run levels are now known as targets? Oct 28 15:02:58 so the basic functionally resembles sysvinit very much Oct 28 15:03:34 well, you have targets like sound, graphical desktop and so on, which can depend on others and change accordingly Oct 28 15:03:54 interesting Oct 28 15:04:02 ie. multiuser.target, graphical.target etc Oct 28 15:04:17 you should read about it Oct 28 15:04:22 it's totally possible to up the graphical desktop without sound etc? Oct 28 15:04:28 I'm not that good explaining anything :P Oct 28 15:04:40 yes Oct 28 15:04:58 wikipedia has a really brief explanation on systemd Oct 28 15:05:46 check systemd site, and this is also a good read https://wiki.archlinux.org/index.php/Systemd Oct 28 15:06:06 freemangordon: what about PA with -O3 and -fno-fast-math? Oct 28 15:07:09 (( they are breaking linux)) they did from first minute they hijacked it Oct 28 15:11:14 DocScrutinizer05: you're just jealous of poettering's awesoem system architecturing skillz Oct 28 15:12:33 appropriately enough, poettering's blog is called "I'll Break Your Audio" Oct 28 15:12:36 no, i'm jalous of his complete ignorance regarding situations where he can make himself a fool Oct 28 15:13:22 his God-alike self-esteem Oct 28 15:13:56 what I'm definitely NOT jealous about are his actual skills Oct 28 15:14:52 kerio: I can't think of a way to benchmark it precisely Oct 28 15:15:15 play song with known player Oct 28 15:15:21 and for the love of god, *not mafw* Oct 28 15:15:32 why not mafw? Oct 28 15:15:33 try high-volume speaker too Oct 28 15:15:37 because mafw sucks balls Oct 28 15:15:47 how is that related to PA? Oct 28 15:16:00 you want something that doesn't use a lot of cpu Oct 28 15:16:18 kerio: I want to measure gow much CPU PA uses Oct 28 15:16:27 *how Oct 28 15:16:28 mplayer -ao pulse Oct 28 15:16:31 top Oct 28 15:16:33 ie Oct 28 15:16:41 that is what I am doing Oct 28 15:16:47 besides I use mafw Oct 28 15:16:56 with frequency lokced to 250 Oct 28 15:17:09 but PA CPU usage varies Oct 28 15:17:36 any idea how to get only one process CPU usage and to write it to a file? Oct 28 15:18:00 so I can put that later in a table, for comparison Oct 28 15:23:21 powertop, maybe? Oct 28 15:23:37 top -d 1| grep -v grep | grep `pidof pulseaudio` | cut ... Oct 28 15:23:50 cat /proc/self/stat Oct 28 15:23:59 self? Oct 28 15:24:18 well, <`pidof yourprocess`> Oct 28 15:25:45 hmm, which column is CPU usage? Oct 28 15:25:54 you have to calculate the ratio of jiffies for that process to jiffies expired in total since last query Oct 28 15:26:03 yeah :D Oct 28 15:26:10 i'll stick with top Oct 28 15:26:12 and on ubuntu topic.. I've been using Lubuntu for a year or so and it's something a human being would expect from a linux distro Oct 28 15:26:59 freemangordon: good luck with top -d 1 Oct 28 15:27:10 why? Oct 28 15:27:29 try it Oct 28 15:27:37 it seems work Oct 28 15:27:42 i've tried it Oct 28 15:27:58 never worked for me to pipe / redirect top output to anything Oct 28 15:28:40 maybe for messybox top that's different Oct 28 15:29:38 maybe you meant top -n 1 Oct 28 15:29:53 d should be for interval iirc Oct 28 15:29:57 which actually seems to somwhat work Oct 28 15:31:01 anyway top output is ncurses, rather ugly and indigestible Oct 28 15:31:17 top|less Oct 28 15:31:26 I just need CPU usage percent with 1 s sampling rate Oct 28 15:31:31 freemangordon: you could loop this ps -eo pid,user,%cpu,args --sort %cpu | grep pulseaudio | tail -n1 | awk '{print $3}' | awk -F \. '{print $1}' Oct 28 15:31:32 which is a nice cmdline in itself ;-P Oct 28 15:33:00 ok, have the results for O2 4.7, lets see how is stock Oct 28 15:33:29 yeah, this would actually work, for procps, though not for busybox ps Oct 28 15:33:47 ah, I always forget the busybox :S Oct 28 15:34:04 fscking hate it Oct 28 15:34:15 erm, dislike Oct 28 15:34:43 don't say "I hate it" or somebody here will tell you you're silly as hating a software is impossible Oct 28 15:35:34 ;) Oct 28 15:36:12 ~messybox Oct 28 15:36:12 messy... err busybox is meant for lean scripting. Regarding all the missing options and immanent limitations (see su, passwd) it's not really the interactive shell of choice. A lot of people hate busybox because a lot of system integrators don't understand the difference between busybox and a decent user interactive shell plus unix utils Oct 28 15:38:14 Skry: btw many thanks for the nice ps options cmdline Oct 28 15:38:19 mad useful Oct 28 15:40:34 DocScrutinizer05: np Oct 28 15:41:45 what is awk -F \. '{print $1}' doing though? Oct 28 15:42:17 aiui it just prints the digits before decimal? Oct 28 15:43:37 (btw I got procps on my iroN900 ;-D ) Oct 28 15:43:38 DocScrutinizer05: matter of output formatting Oct 28 15:43:59 if you want to see the .? you can omit that Oct 28 15:44:09 already did Oct 28 15:44:23 Xorg 0 vs Xorg 0.2 Oct 28 15:44:30 yeah Oct 28 15:44:56 guess it would be more sane to round it though Oct 28 15:45:17 DocScrutinizer05: but you have to uninstall the metapackage to install procps D: Oct 28 15:45:17 maybe freemangordon can use % with decimals Oct 28 15:45:36 aaah, so it's been THAT Oct 28 15:46:10 and installing stock camera-ui already had no problem then Oct 28 15:46:32 now I recall I wrote exactly that when nswering the question in CSSU tmo thread Oct 28 15:46:54 no, no, the stock camera-ui is still problematic Oct 28 15:47:18 well, if you say so Oct 28 15:47:27 i do say so Oct 28 15:47:34 I think the MP at large isproblematic Oct 28 15:48:05 and CSSU should get rid of it, sonner rather than later Oct 28 15:48:10 sooner* Oct 28 15:48:37 it's locking in users Oct 28 15:48:45 camera-ui (>= 1.1.29.2+0cssu13) Oct 28 15:49:03 that's a dependency of the latest mp-fremantle-community-pr outside of thumb Oct 28 15:49:22 (the thumby one depends on +0cssu14+thumb0) Oct 28 15:49:49 DocScrutinizer05: HAM has a way of enabling a "magic" package that upgrades everything Oct 28 15:50:16 I'm listening Oct 28 15:50:26 no, it ends here :) Oct 28 15:50:36 there's still no way to cherrypick updates for non-user packages Oct 28 15:50:40 didn't help me out Oct 28 15:50:41 unless you show all packages Oct 28 15:52:09 hmm, does not look good Oct 28 15:52:26 seems that thumb PA uses more CPU than stock Oct 28 15:52:45 how much more? Oct 28 15:52:47 wtf Oct 28 15:52:48 why? Oct 28 15:52:50 it makes no sense Oct 28 15:53:00 well, thumb never claimed to be more CPU efficient, only more compact Oct 28 15:53:13 well, what i did is to get 95 samples, for one and the same song Oct 28 15:53:37 then substracted sample by sample and sum Oct 28 15:53:59 the difference is 80.5 Oct 28 15:54:05 ~80.5/94 Oct 28 15:54:05 0.856382978723 Oct 28 15:54:11 80.5 what? Oct 28 15:54:40 kerio: read again Oct 28 15:55:05 * DocScrutinizer05 does, scratches head Oct 28 15:55:09 i'm still missing the unit of measurement Oct 28 15:55:43 SUM(thumb_cpu_usage(i)-stock_cpu_usage(i)), i=1-94 Oct 28 15:56:00 WHAT'S THE UNIT OF MEASUREMENT Oct 28 15:56:05 % Oct 28 15:56:09 oic Oct 28 15:56:17 btw a tricky method to calculate cpu usage of a process is to run a busy counting loop in another process and see how much it slows down Oct 28 15:56:18 ...you summed percentages? :s Oct 28 15:56:29 kerio: yes Oct 28 15:57:03 and according to that -thumb uses .85 % more CPU per secon Oct 28 15:57:07 *second Oct 28 15:57:50 But I just don't understand why Oct 28 15:57:52 :( Oct 28 15:58:16 0.85% per second? wtf is that? an inflation magnitude? Oct 28 15:58:22 going to try 4.7 without thumb Oct 28 15:58:34 DocScrutinizer05: see the formula above Oct 28 15:58:46 after 120s it will use 110$ of CPU? Oct 28 15:58:50 freemangordon: how do we call the result, though? Oct 28 15:58:54 % Oct 28 15:59:05 +nonthumb0 Oct 28 15:59:10 and according to that -thumb uses .85 % more CPU per second Oct 28 15:59:19 DocScrutinizer05: more per second Oct 28 15:59:19 I'm 35 years per minute old Oct 28 15:59:37 DocScrutinizer05: i don;t understand your point Oct 28 15:59:50 wtf is % per second? Oct 28 16:00:01 or better said you don;t understand what i did Oct 28 16:00:10 cpu load Oct 28 16:00:15 no Oct 28 16:00:29 no what? Oct 28 16:00:44 since when something is using 1% per second, it uses 3% after 3 s Oct 28 16:00:46 thats an average for the period divided by the number of sample Oct 28 16:00:58 *samples Oct 28 16:01:18 or even the average of the error for the period Oct 28 16:02:10 your savings my give 3% per year of interest Oct 28 16:02:41 a process can hardly use % / s of CPU Oct 28 16:02:52 DocScrutinizer05: I am not exactly famous with m English skills, that is why I hope you math skills are in place, so you can grok the above formula Oct 28 16:03:17 DocScrutinizer05: samples are taken every second Oct 28 16:04:04 so if you sum them and divide by the number of samples you'll get an average CPU load per second Oct 28 16:05:34 if "per" is not the correct english word I am pretty ok Oct 28 16:05:46 but that does not change the math Oct 28 16:06:08 your average CPU usage for case A over 95 samples of 1s duration been 80.5%, for case B 94%. The *difference* are 13.5% Oct 28 16:07:05 what? Oct 28 16:07:37 which is an absolute factor of 0.856382978723 slowdown Oct 28 16:08:01 it is not absolute, as the original unit is CPU load % Oct 28 16:08:20 that is why unit is %/s Oct 28 16:08:27 the ratio between the two percentages though is absolute Oct 28 16:08:35 it is not ratio Oct 28 16:08:49 [2012-10-28 16:54:05] ~80.5/94 Oct 28 16:08:53 ^^^ ratio Oct 28 16:09:03 94 is seconds Oct 28 16:09:11 ummm Oct 28 16:09:11 and 80.5 in in % Oct 28 16:09:23 *is in Oct 28 16:09:28 dividing % by seconds is BS Oct 28 16:10:02 no, it is not, that is how you find the average fo a given period Oct 28 16:10:14 the same as S=V.t :P Oct 28 16:10:28 V=S/t is the average speed Oct 28 16:10:37 nonsense Oct 28 16:11:08 yeah, sure Oct 28 16:11:35 freemangordon: the average for a given period is gotten by dividing the sum of all sampled values by the number of samples. Oct 28 16:11:40 you don't tell me you added PA CPU % of 94 samples and got 80.5 as result Oct 28 16:11:43 exactky Oct 28 16:11:47 exactly Oct 28 16:12:13 but % is not a real value Oct 28 16:12:21 oh sorry, my brain went out the window Oct 28 16:12:26 well then nevermind, since a absolute CPU usage of <1% by PA is negligible, and also highly unlikely Oct 28 16:12:43 SUM(thumb_cpu_usage(i)-stock_cpu_usage(i)), i=1-94 Oct 28 16:12:57 and the result of that is 80.5 Oct 28 16:13:06 meh Oct 28 16:13:36 and IIRC this is called error Oct 28 16:13:42 OMFG, obfuscated math formula contest Oct 28 16:13:43 or sigma Oct 28 16:14:06 DocScrutinizer05: help him get it right then, rather than telling he's doing it wrong... Oct 28 16:14:13 indeed Oct 28 16:14:16 (the formula syntax) Oct 28 16:14:27 nah, he's doing it "right", just completely weird Oct 28 16:14:28 btw, I asked yesterday about what resampler pulse uses in maemo, didnt get the answers and still dont know what it uses but still.. pulseaudio 0.99.2 cpu usage according to top: trivial:~12%, speex-float-0:~33%, ffmpeg:~40% Oct 28 16:14:42 DocScrutinizer05: that might be, but saying it's weird/odd/wrong is not exactly helping Oct 28 16:14:45 :) Oct 28 16:15:00 and 94 is not seconds but samples, a basically unitless value Oct 28 16:15:14 Skry: PA uses speex resampler , which is closed source Oct 28 16:15:38 DocScrutinizer05: it is in secconds if you have 94 samples taken for 94 seconds Oct 28 16:15:55 freemangordon: he meant that your formula is not telling what 94 is Oct 28 16:16:00 just that it should be 94 in that example Oct 28 16:16:04 freemangordon: I'd calculate average for CPU usage of thumb_cpu_usage(i) for i:1..94, then do same for stock_cpu_usage(i) Oct 28 16:16:18 makes no difference Oct 28 16:16:23 then you got average CPU usage of case A and B, then you can compare the two Oct 28 16:16:37 sure, makes no differnece in math, but in legibility Oct 28 16:17:15 DocScrutinizer05: but it makes difference if you have to deal with large numbers, you ca get overflow ;) Oct 28 16:17:24 freemangordon: whatwhat? is it some nokia hack version or what? I mean, speex is free. Oct 28 16:17:26 modulo you'd probably notice that your 94 are not of unit 'seconds' if you'd do a well established simple average Oct 28 16:17:42 freemangordon: ack for overflow Oct 28 16:18:03 Skry: no, it is BSD according to the bug report on bugs.maemo.org Oct 28 16:18:37 DocScrutinizer05: that is why it looks weird, it is just the way I am used to write the code Oct 28 16:18:42 freemangordon: 94 is the number of samples, no matter if every sample is 1 or 33.8 seconds in duration Oct 28 16:19:11 so your result are % differnece, not %/s Oct 28 16:19:24 DocScrutinizer05: it matters if you want to know by what % is CPU load increased every second Oct 28 16:19:36 DocScrutinizer05: yes Oct 28 16:19:54 18:00 and 80.5 in in % Oct 28 16:19:59 *is Oct 28 16:20:34 yes, it's the sigma-difference for i from 1 to 95 Oct 28 16:20:49 so you divide by 95 to get true average Oct 28 16:20:49 Skry: there is an open bug to open the source of speex Oct 28 16:21:14 DocScrutinizer05: finally :P Oct 28 16:21:43 freemangordon: http://downloads.xiph.org/releases/speex/speex-1.2rc1.tar.gz Oct 28 16:21:45 95 is Oct 28 16:21:50 ta-da Oct 28 16:22:07 Skry: we have NEON-optimized resampler here Oct 28 16:22:15 seconds is not even showing up in hte whole equation Oct 28 16:22:20 look in debian/changelog Oct 28 16:22:54 since i is a enum, not a seconds integer Oct 28 16:23:42 DocScrutinizer05: well, it could be said that way too, but i is seconds too, as the samples were taken every second Oct 28 16:23:56 doesn't matter Oct 28 16:24:19 if your samples were 2 seconds long, the i still is 94 for 94 samples Oct 28 16:24:21 so the functions could be f(t) as well Oct 28 16:24:43 but I want to average over the time, not the number of samples Oct 28 16:24:50 freemangordon: maybe the NEON resampler sucks Oct 28 16:24:52 *over the number Oct 28 16:25:31 freemangordon: alas that's not how a simple average works Oct 28 16:25:36 freemangordon: libspeex/resample_neon.h Oct 28 16:25:40 kerio: according to changelog it speeds the things a lot Oct 28 16:25:51 Skry: in speex-dev? Oct 28 16:26:02 yes, since 2011 Oct 28 16:26:10 hahaha Oct 28 16:26:14 upstream? Oct 28 16:26:41 who is the copyright holder, Nokia? Oct 28 16:27:23 patches are from jyri sarha Oct 28 16:27:39 @ti.com Oct 28 16:28:23 dont know of the upstream status but patches are in speex-dev ml and used by ubuntu for example Oct 28 16:28:44 Skry: sounds too risky to put that in -thumb Oct 28 16:29:10 mm, might be Oct 28 16:29:35 freemangordon: it's probably what nokia has, actually Oct 28 16:29:35 i'll check the upstream status.. Oct 28 16:29:43 anyway, gcc sucks at NEON Oct 28 16:29:58 DocScrutinizer05: that is exactly how simple average over the time works Oct 28 16:30:14 freemangordon: if you'd want to do an integral average, you have to multiply each sample value by the sample duration (in seconds) and then divide the whole thing by the sigma(duration) over all samples. (sigma((thumb_cpu_usage(sample-i) * thumb_sample_duration(sample-i)) / sigma(thumb_sample_duration(sample-i)) Oct 28 16:30:51 so you get %*s/s = % Oct 28 16:31:37 the fact that all your thumb_sample_duration(sample-i) = 1second doesn't change the formula, so doesn't result in %/s Oct 28 16:34:49 DocScrutinizer05: the way I did it fits for the purpose: thumb uses more CPU than stock Oct 28 16:35:19 yes, your calculation been correct, the units though were not Oct 28 16:35:40 meh Oct 28 16:36:17 it is correct too, as if it is the average for the period, it could be average for 1 second period Oct 28 16:36:20 and the difference is actually <1% if your calculation actually been accurate Oct 28 16:36:25 thus negligible Oct 28 16:36:35 well, it was excel to do the math Oct 28 16:36:37 or rather, below tolerance Oct 28 16:37:02 well, I was expecting -thumb to behave much better than stock Oct 28 16:37:16 maybe I am missing something Oct 28 16:37:39 in fact they behave more identical than anybody would expect for two runs of very same software Oct 28 16:37:42 freemangordon: is this the thumby? Oct 28 16:37:52 what about the army? Oct 28 16:38:03 though if we have FP math all the way... Oct 28 16:38:09 kerio: going to try it Oct 28 16:38:17 kerio: join the army, man! Oct 28 16:41:54 freemangordon: honestly, compare arm to arm (two consecutive runs of same binary) and you'll likely find larger difference than this 0.8% Oct 28 16:42:23 DocScrutinizer05: the point is that they shouldn't be the same Oct 28 16:42:32 yes Oct 28 16:42:36 the thumb one should pwn the stock one Oct 28 16:42:51 freemangordon: wasn't that you I did some hardfp vs softfp simple test binary runs, and we noticed effectively 0 difference between both? Oct 28 16:43:03 DocScrutinizer05: no Oct 28 16:43:33 not in upstream, but learned that speex has C64x support, whoa. Oct 28 16:43:50 though, if most of the stuff is done in FP, there will be no difference Oct 28 16:44:10 Skry: yeah, speexdsp is closed source too in maemo5 :D:D:D Oct 28 16:45:06 kerio: I don't agree on the assumption that thumb always has to be faster than arm, since it is meant to yield compact code, not faster execution (the faster execution is a side effect under certain narrow preconditions). What I agree though is that we should see way larger differences, in whatever direction Oct 28 16:45:30 DocScrutinizer05: less of a thumb matter, more of a gcc4.2 vs gcc4.7-linaro matter Oct 28 16:46:25 the smaller the program, the smaller the opportunities for compiler to optimize Oct 28 16:47:04 fmg: arm pls Oct 28 16:47:26 one very interesting detail would be size of resulting binary Oct 28 16:47:41 ok, build is ready, lets see Oct 28 16:48:33 by the way, are you sure you're using the correct pulseaudio? Oct 28 16:48:58 the one on CSSU gitorious Oct 28 16:49:04 *in Oct 28 16:49:58 no, i mean, are you sure the pulseaudio you're running is the recompiled one? :) Oct 28 16:49:58 I like to implement any arbitrary marker into the stuff under test, so I can tell for sure it's been this one binary that got used in the test Oct 28 16:50:16 or what kerio says ^^^ Oct 28 16:51:08 kerio: well stop/start pulseaudio should do the trick I think Oct 28 16:51:23 i'd still reboot to be sure :3 Oct 28 16:51:28 lets see what arm build will give as results Oct 28 16:51:30 a printf to stdout, listing e.g the number of processed samples/frames is usually just fine Oct 28 16:51:50 or stderr Oct 28 16:52:12 in "destructor" Oct 28 16:52:25 or just before exit Oct 28 16:53:47 or even to /dev/console or whatever Oct 28 16:54:00 instead of to stderr Oct 28 16:54:11 hmm, i'll reboot and retry, just for sure Oct 28 16:55:04 as now 80.5 become 33 :D Oct 28 16:55:41 freemangordon: if you had a welcome msg printf at beginning of your new code, and a "processed %d samples. This was Version th-33" at exit, you'd notice immediately what's going on Oct 28 16:56:20 there is no "my new code" I am just doing dpkg-buildpackage Oct 28 16:56:32 yes I'm aware of that Oct 28 16:56:35 and changing flags in debian/rules Oct 28 16:56:51 too lazy to place "your new code" into ticks Oct 28 16:57:51 you could pass the verson string via compile parameter -D Oct 28 16:58:38 -rwx------ 1 root root 72784 Oct 28 2012 /usr/bin/pulseaudio Oct 28 16:58:43 do a logger printout to syslog Oct 28 16:58:49 still the same after the reboot Oct 28 17:00:34 could it be that PA does not actually do much job? Oct 28 17:00:51 72768 here Oct 28 17:01:51 well, the whole stuff is piped thru gstreamer afaik, by mafw Oct 28 17:02:17 so we have only 16 bytes difference between 4.2.1 and 4.7.2? WTF? Oct 28 17:03:19 IroN900:~# ls -l /usr/bin/pulseaudio Oct 28 17:03:21 -rwx------ 1 root root 72768 2010-06-14 15:32 /usr/bin/pulseaudio Oct 28 17:03:28 going to try -Os Oct 28 17:03:35 hahaha Oct 28 17:03:59 freemangordon: this is the ARM one, right? Oct 28 17:04:03 yep Oct 28 17:04:16 so... compiling in ARM gave us a giant reduction? Oct 28 17:04:30 yeah, 16 bytes more Oct 28 17:04:35 :D:D:D Oct 28 17:04:43 ok, gimme that package :£ Oct 28 17:04:44 :3 Oct 28 17:07:07 freemangordon: 795 pulse 12 -8 83192 6912 2424 S 21.6 2.8 5:16.93 pulseaudio Oct 28 17:07:16 21.6% Oct 28 17:07:50 what is it? Oct 28 17:08:03 total CPU usage? Oct 28 17:08:14 top output for >> while play-sound /home/opt/usr/share/osso-lmarbles/sounds/alarm.wav; do :; done << Oct 28 17:08:26 yep, CPU usage Oct 28 17:09:08 no other process used significant CPU Oct 28 17:09:09 i wonder... what's the CPU usage when on speaker and when using BT headphones? Oct 28 17:12:51 hmm, makes sense to check that Oct 28 17:29:40 hmm, -Os produced binary which is about 68k ARM build, but not much of a difference in CPU usage Oct 28 17:30:16 so it seems it is not PA itself that uses the CPU , most probably it is the closed packages that do :( Oct 28 17:39:02 kerio: connecting BT HF drops CPU usage by a couple of % Oct 28 17:39:15 5-10 Oct 28 17:39:24 hehe Oct 28 17:39:31 hm, is that HFP/HSP or A2DP? Oct 28 17:39:38 HFP Oct 28 17:39:49 sucks Oct 28 17:40:05 actually it streams to my PC, but I am almost sure it is not a2dp Oct 28 17:40:12 is it stereo? Oct 28 17:40:21 actually, easier question: Oct 28 17:40:24 hmm, it is a2dp Oct 28 17:40:25 does it sound like shit? Oct 28 17:40:34 no, it sound ok Oct 28 17:40:39 *sounds Oct 28 17:40:39 then it's a2dp :) Oct 28 17:40:45 yes it is, i checked Oct 28 17:40:46 stupid lowpass filter Oct 28 17:41:20 hmm, actually lowpass filter seems to take only 5 or so % , which is not that much Oct 28 17:41:37 on 250 is that Oct 28 17:42:11 it seems it is the resampler to eat the cpu Oct 28 17:42:32 or better said I am out of ideas :D Oct 28 17:44:40 though in upstream PA there are some NEON optimizations, maybe I can backport them to check if it will be better Oct 28 17:51:06 Skry; sure Oct 28 17:51:17 freemangordon: anyway, the arm-compiled PA is faster than the stock one, right? Oct 28 17:59:57 kerio: no Oct 28 18:00:03 it is the same Oct 28 18:00:14 dafuq Oct 28 18:00:18 BTW changing resample-method = trivial drops CPU usage by 10% Oct 28 18:00:24 what had the giant advantage then? Oct 28 18:00:33 in /etc/pulse/daemon.conf Oct 28 18:00:47 freemangordon: that's cheating Oct 28 18:01:42 I know Oct 28 18:02:00 But at least gives clue who is using the CPU Oct 28 18:02:53 freemangordon: it's audibly worsse Oct 28 18:02:55 *worse Oct 28 18:02:58 much, much worse Oct 28 18:03:09 i wonder why it needs to resample to A2DP Oct 28 18:04:04 hrmpf, is there a reason 48kHz is the default samplerate? Oct 28 18:09:30 setting it to 44.1kHz has no adverse effects, it seems Oct 28 18:14:18 hm, why are we using pulseaudio 0.9 when the current release is 2.1? Oct 28 18:19:18 because nobody dared to update stuff since maemo5 birth? Oct 28 18:19:51 becuae the closed blobs wouldn't fit into any newer version, since we can't compile them? Oct 28 18:20:00 i wonder if it's really thumb that's giving the benefits or just the updated gcc 4.7.2-linaro (no proof though) Oct 28 18:20:13 kerio: because the closed plugins depend on the internal pulse api and this you can't pimp its version Oct 28 18:20:31 or that ;-) Oct 28 18:20:33 why do we need the closed plugins? Oct 28 18:20:36 MrPingu: there are no benefits! Oct 28 18:20:57 merlin1991: though we can backport what worths, i saw some NEON optimized conversion routines Oct 28 18:21:01 because we love our speakers and the decent phone audio quality Oct 28 18:21:57 ShadowJK: ok thanks, that's a great script. I added led control, logging to syslog, isp1704 charger detection and adapted it to be used with udev. Does the job quite nicely. Oct 28 18:22:03 Kerio: With benefits I mean the smaller binarysize ;) Oct 28 18:22:10 though it seems to me cmtspeech is actually open, just XPROT is closed Oct 28 18:22:10 http://www.mail-archive.com/pulseaudio-discuss@lists.freedesktop.org/msg04777.html Oct 28 18:22:17 what's XPROT? Oct 28 18:22:26 speaker protection Oct 28 18:23:25 Skry: isp1704 charger detection might mess up things as soon as you're running it with established USB session Oct 28 18:23:36 which can be done *in hardware* via the equalizer Oct 28 18:23:43 going to try to backport that Oct 28 18:24:09 "main" CSSU will benefit from it too if I succeed Oct 28 18:24:30 DocScrutinizer05: it only does detecting if charger is present and gets charger type for logging purposes Oct 28 18:24:54 mhm Oct 28 18:25:29 kerio: nemo has 2.1, modules from meego/mer are ported, not sure if they have call audio with it yet. Oct 28 18:25:58 Skry: sorry when "charger detection runs only when charger detected" doesn't make sense to me Oct 28 18:27:05 Kerio: http://talk.maemo.org/showpost.php?p=1285892&postcount=971 Oct 28 18:27:21 xD Oct 28 18:30:16 DocScrutinizer05: uhm? udev launches script if it detects power supply event, script checks if isp1407 reports charger and continues accordingly. Perhaps it should be done more properly via more exact udev rule but i have no interest digging into udev insides more than I already know. Oct 28 18:47:01 we sold the n9, then he left the n950 on a train Oct 28 18:47:04 WHAT THE FUCKING FUCK Oct 28 19:02:36 How can I configure which accounts or messages lead to vibration? Oct 28 19:08:38 invariant: I don't think there's such config option Oct 28 19:18:29 What is the best irc client for n900 which allows setting port number? Oct 28 19:19:10 quassel2go Oct 28 19:20:46 invariant: xchat Oct 28 19:21:00 Installation corrupted. Yes/no Oct 28 19:21:18 Interesting message Oct 28 19:21:56 What should I pick? Oct 28 19:22:04 'no'? Oct 28 19:22:25 invariant: installation of what? Oct 28 19:24:39 Qutim Oct 28 20:00:33 ~seen qwazix Oct 28 20:00:40 qwazix <~qwazix@athedsl-353742.home.otenet.gr> was last seen on IRC in channel #maemo, 1d 7h 49m 10s ago, saying: 'who's got a Nexus 7? Native ubuntu on it since yesterday, with hildon on it could be almost maemo.'. Oct 28 20:16:11 WTF, Ducky died? Oct 28 20:16:38 Dr. Mellard :-o Oct 28 20:17:29 who died? :o Oct 28 20:18:56 DocScrutinizer05: your fucking kiding me?! Oct 28 20:19:56 >>While walking on a beach, Ducky suffered a heart attack in the season nine finale, "Till Death Do Us Part", after hearing about the bomb blast at NCIS headquarters and then is seen lying, motionless on a beach. In "Extreme Prejudice", it is shown that he's alive and receiving treatment thanks to Jimmy Palmer finding him. << Oct 28 20:20:32 .... Oct 28 20:20:41 thats old news :p Oct 28 20:20:56 not here ;-) Oct 28 20:21:04 haha :p Oct 28 20:21:13 thats last sesson! Oct 28 20:23:37 I thought you spoil :p Oct 28 20:25:29 nah, i'm not a partypooper Oct 28 20:26:26 the bad news: next Sunday evening I won't watch next episode of NCIS Oct 28 20:27:15 since I bet they introduce another 3 months minimum pause before starting next (10th) season Oct 28 20:28:13 hmm.. that sux.. :/ Oct 28 20:29:06 judging by rating, the 10th season must be even better: http://en.wikipedia.org/wiki/NCIS_(TV_series)#Ratings Oct 28 20:31:32 haha nice :D Oct 28 20:35:29 think we are at s09 e05 or something in swe :p Oct 28 22:24:41 ~botsnack Oct 28 22:24:41 thanks, DocScrutinizer51 Oct 28 22:32:20 hahaha aww Oct 28 22:32:44 your bot DocScrutinizer05? Oct 28 22:40:21 not really Oct 28 22:40:36 Tim's Oct 28 22:41:32 ~owner Oct 28 22:41:33 TimRiker is my owner Oct 28 22:42:12 I'm only one of the botmasters Oct 28 22:43:41 means I can do some admin tasks Oct 28 22:46:05 and I try to keep the knowledge database clean and up to date, for a 0.001% part of it Oct 28 22:46:13 aka factoids Oct 28 22:46:21 like Oct 28 22:46:26 ~power Oct 28 22:46:27 power is probably http://wiki.maemo.org/N900_Hardware_Power_Consumption Oct 28 22:50:57 DocScrutinizer05: ahh, I see, well it's a cutie :) Oct 28 22:54:01 indeed Oct 28 22:54:13 ~status Oct 28 22:54:30 eh? Oct 28 22:54:31 Since Tue Oct 23 14:06:35 2012, there have been 3 modifications, 184 questions, 0 dunnos, 0 morons and 113 commands. I have been awake for 5d 8h 47m 56s this session, and currently reference 118762 factoids. I'm using about 23856 kB of memory. With 0 active forks. Process time user/system 1491.16/27.45 child 0/0 Oct 28 22:54:46 ~useless Oct 28 22:54:46 * infobot starts crying and hides from docscrutinizer51 in the darkest corner of the room. :( Oct 28 22:55:03 aww Oct 28 22:55:10 ~hug Oct 28 22:55:16 :( Oct 28 22:55:25 * infobot hugs cor-ai Oct 28 22:55:26 she's slow Oct 28 22:55:33 or busy Oct 28 22:55:42 haha or both Oct 28 22:56:15 well, shes smarter then mine! hehe Oct 28 22:56:23 alas I have no access to logs and CPU load Oct 28 22:57:24 well, time to sleep! the alarm sounds in 5 hour! nn Oct 28 22:57:46 3 commands you should know: ~help, ~listkeys, ~listvalues Oct 28 22:58:02 night! Oct 28 22:58:05 orly? Oct 28 22:58:25 ~liskeys maemo Oct 28 22:58:57 damn she must be busy Oct 28 22:59:05 haha I think she needs some sleep too! Oct 28 22:59:18 she needs one t Oct 28 22:59:21 meh Oct 28 22:59:29 ~listkeys maemo Oct 28 22:59:31 Factoid search of 'maemo' by key (16 of 28): maemo-lists ;; #maemo multiboot ;; mxr.maemo.org ;; maemo-version ;; maemo-man ;; /.maemo-mapper/paths.db ;; maemo-qt ;; #maemo-ssu ot ;; mxr.maemo.org deb ;; maemo-brand ;; maemo-sdk ;; #maemo power ;; maemo-logs ;; repository.maemo.org deb ;; chromium on maemo ;; maemo-down. Oct 28 22:59:53 ohh Oct 28 23:00:19 ~maemo-lists Oct 28 23:00:20 i heard maemo-lists is http://maemo.org/community/mailing-lists.html Oct 28 23:00:52 ~listvalues community Oct 28 23:00:55 Factoid search of 'community' by value (20 of 109): broken-maemo ;; council ;; cssu ;; cssu-optional ;; dway ;; freerunner ;; jargon bonk/oif ;; jargon core ;; jargon gnu ;; jargon phreaking ;; jargon rl ;; jargon tex ;; jargon twenex ;; ltp ;; maemo-lists ;; mailing-lists ;; pcp ;; pigtk ;; polycom ;; progeny+debian. Oct 28 23:01:54 damn I miss my n900 :/ realy have to get my thumb out and order a new usb socket! Oct 28 23:01:55 wtf? Oct 28 23:02:24 ~listkeys jargon Oct 28 23:02:26 Factoid search of 'jargon' by key (11 of 2335): jargon aluminum book2 ;; jargon baud barf2 ;; jargon block transfer computations ;; jargon bondage-and-discipline language2 ;; jargon book titles8 ;; jargon buffer overflow ;; jargon chrome ;; jargon classic c ;; jargon cokebottle ;; jargon dwim ;; jargon eat flaming death. Oct 28 23:03:46 ~jargon cokebottle Oct 28 23:03:47 somebody said jargon cokebottle was /kohk'bot-l/ n. Any very unusual character, particularly one you can't type because it it isn't on your keyboard. MIT people used to complain about the `control-meta-cokebottle' commands at SAIL, and SAIL people complained right back about the `altmode-altmode-cokebottle' commands at MIT. After the demise of the {space-cadet keyboard}, `cokebottle' faded away as serious usage, but was often invoked humorously to ... Oct 28 23:04:23 mm, I found eggdrop.tar.gz in my backups few weeks back, looked at it and found there was all the scripts, logs, stats and stuff for a bot I ran last time something like 10 years ago on ircnet. Made me feel nostalgic, then old. Oct 28 23:04:56 hehe Oct 28 23:05:14 not old! just not young any more! :D Oct 28 23:05:29 thats right! :) Oct 28 23:06:09 ~dict old Oct 28 23:06:11 Dictionary 'old' (4 of 13): past times (especially in the phrase `in days of old') ;; (used for emphasis) very familiar; "good old boy"; "same old story" ;; of long duration; not new; "old tradition"; "old house"; "old wine"; "old country"; "old friendships"; "old money" ;; \Old\ ([=o]ld), n. Open country. [Obs.] See {World}. --Shak. [1913 Webster]. Oct 28 23:07:06 ~wiki old Oct 28 23:07:10 At http://en.wikipedia.org/wiki/Old (URL), Wikipedia explains: "{{wiktionary|old|olde}} 'Old' or 'OLD' may refer to: *Old age or, by extension, a description or nickname for someone or something that has endured and become comfortable or widely familiar. *Old, Baranya, Hungary *Old, Northamptonshire, England *Old Town Municipal Airport and Seaplane Base in Old Town, Maine, United States (IATA code) *Old Boys and Old Girls, former pupils of schools in ... Oct 28 23:08:19 haha must try this! Oct 28 23:08:28 ~wiki cor-ai Oct 28 23:08:33 At http://en.wikipedia.org/wiki/Cor-ai (URL), Wikipedia explains: "{{Infobox television season |season_name=Stargate SG-1" Season 1 |image= |caption=Region 1 DVD cover art | show_name = Stargate SG-1 |dvd_release_date='Region 1': May 22, 2001'Region 2': October 21, 2002'Region 4': March 1, 2004 |country= |network=Showtime |first_aired=July 27, 1997 |last_aired=March 6, 1998 |num_episodes=22 |next_season=Season 2 |}} 'First season' of the military ... Oct 28 23:08:43 haha yay Oct 28 23:10:54 yeah.. bed.. nn, again! Oct 28 23:12:45 ok, me too :D Oct 28 23:12:52 ~wiki skry Oct 28 23:12:56 At http://en.wikipedia.org/wiki/Skry (URL), Wikipedia explains: "{{Citations missing|article|date=January 2011}} {{redirect-distinguish|Scry|Scrye}} (1902, oil on canvas)]] 'Scrying' (also called 'seeing' or 'peeping') is a magic practice that involves seeing things psychically in a medium, usually for purposes of obtaining spiritual visions and less often for purposes of divination or fortune-telling. The most common media used are reflective, ... Oct 28 23:14:02 ~skrymir Oct 28 23:14:08 ~wiki skrymir Oct 28 23:14:11 :S Oct 28 23:14:17 At http://en.wikipedia.org/wiki/Skrymir (URL), Wikipedia explains: "{{For|the Marvel Comics character|Utgard-Loki (Marvel Comics)}} [[Image:Louis Huard - Giant Skrymir and Thor.jpg|thumb|200px|[The] Giant Skrymir and Thor (c. 1891), by Louis Huard.]] In Norse mythology, 'Útgarða-Loki' (Anglicized as 'Utgarda-Loki' and in other ways) was the ruler of the castle Útgarðr in Jötunheimr. He was one of the Jötnar and his name means literally "Loki o Oct 28 23:20:56 pff, it appears i cant do the charging thingy with udev after all. funny thing is that i can do it with udev AND systemd.. now, someone is apparently forcing me to do stuff with systemd :) Oct 28 23:23:46 the charging thingie shield work under any root permissions. If however you plan to mess around with indicator LED via DBUS msg to MCE, you'll need run-standalone.sh Oct 28 23:24:04 s/shield/should/ Oct 28 23:24:05 DocScrutinizer05 meant: the charging thingie should work under any root permissions. If however you plan to mess around with indicator LED via DBUS msg to MCE, you'll need run-standalone.sh Oct 28 23:27:32 I'm doing the led indicator stuff via sysfs. I just want script to run when charger is plugged in, somehow it sounds stupid that to achieve this I need to use both udev and systemd for the job. Oct 28 23:28:04 I wouldn't see why Oct 28 23:28:05 there is a timer in udev which kills process if it takes too long Oct 28 23:28:11 that's needed Oct 28 23:28:39 hmm, then run the charger script in a background shell Oct 28 23:28:42 & Oct 28 23:28:48 ! Oct 28 23:28:57 oh the fuck I feel stupid Oct 28 23:29:04 ;-D Oct 28 23:29:27 seriously :D Oct 28 23:30:12 some distros have a command like start-daemon or whatever for that Oct 28 23:30:46 run-service Oct 28 23:30:54 you get the idea Oct 28 23:31:38 yeah. hmm, i suppose i should start the script via flock to prevent it spawning Oct 28 23:33:46 if someone decides to charger rape his n900 Oct 28 23:34:47 or bad contact or something so it gets detected multiple times in a row, i have no idea if there is any safety mechanism in udev itself for stuff like that Oct 28 23:51:36 start_daemon: Usage: Oct 28 23:51:37 start_daemon [-f] [-L] [-n +/-] [-u uid] [-g gid] [-v] [-e] \ Oct 28 23:51:38 [-l log|-q|-d] [-p pid_file] [-i ignore_file] [-c root] /path/to/executable [args] Oct 28 23:52:30 won't start a daemon twice, that's what pid_file is for Oct 28 23:52:50 iirc Oct 28 23:53:45 I'd nevertheless suggest to have a look at pali's bq24150.ko module Oct 28 23:54:37 yeah. flock -n /var/run/charger.lock /bla/bla/charger.sh should do the trick too. Oct 28 23:57:33 trouble with using palis module is that last time I tried, it did not work with my current kernel, hooks get called and so on, module reports correct stuff via sysfs but bq27200 reports that battery is discharging Oct 28 23:58:01 while charge21.sh is based on what I found out and designed in a few weeks, pali's bq24150.ko is based on what I thought about same topic after 2+ years of pondering it Oct 28 23:59:43 of course it's not yet as mature as charge21.sh Oct 29 00:00:06 but it's definitely "the future" Oct 29 00:00:19 anyways, I dont know what the fault was, there has been some problems with musb stack in 3.6.x so it might have been to cause to this behaviour. Oct 29 00:00:35 dunno Oct 29 00:01:08 also, pali checked my code so that should not be the issue :) Oct 29 00:01:12 yeah, it's a kernel module, so likely needs some plumbing to work with a perticular kernel Oct 29 00:01:56 and since that's related/linked to musb_hdrc, it's particularly nasty Oct 29 00:02:46 I can tell since I did a lot of diving into this mess by mentorgrafix called musb_hdrc, for usb hostmode Oct 29 00:04:46 yeah i believe Oct 29 00:04:47 in the end you'll come to the conclusion that a battery charger/maintenance-daemon doesn't match too nicely to userland Oct 29 00:05:08 already have :) Oct 29 00:05:17 maemo4 has a story to tell about it Oct 29 00:06:28 I'm feeling quite comfortable with all this stuff, since it's system architecture, my primary competence Oct 29 00:06:35 I'm trying to do this as cleanly as possible so I could have at least some working solution to offer. And once pali gets his module upstreamed I will switch to it immediately Oct 29 00:07:22 you do seem to possess quite admirable level of knowledge about this stuff Oct 29 00:07:28 hoping for upstrem is often just a delusion Oct 29 00:08:41 well, at least he is trying, sadly i happen to know what is required for it get accepted.. -> lot of work. Oct 29 00:08:51 in embedded world you have to accept you sometimes need a partial 'fork' for proper management of your device's peripherals Oct 29 00:10:27 true. biggest issue for me with the whole Arch project so far has been the sgx chip. Oct 29 00:10:38 upstream either accepts braindamaged crippled retarded drivers and you never can replace the established fsckdup API by something sane (see lis302dl), or you simply can't get your stuff upstream as it's not generic enough Oct 29 00:11:14 yeah, also true at times Oct 29 00:12:49 and currently there aro quite some big changes going at linux-omap at least, which requires developers to adapt to new ways of doing things, and it looks like everything unmaintained is going to become unusable Oct 29 00:13:30 and in example omapdss seems to change almost daily :) Oct 29 00:13:59 :-\ Oct 29 00:18:37 kernel lunatics messing up stuff to fubar stae every other week is a long known notorious phenomenon Oct 29 00:18:47 state* Oct 29 00:18:53 there is nothing wrong with progress and improving everything. It's just sad how much devices are left out of it and those abominable kernel forks are becoming more and more common Oct 29 00:19:19 :nod: Oct 29 00:25:26 I've actually bugged quite alot of people around since I first started with N900 and 3.x kernel, I've talked to people from TI, Nokia, Mer, I even harrassed freemangordon to forward port the modem shits. Sad outcome of this is, that there are about a (small) handfull of people with interest of the matter. Oct 29 00:26:34 And from those people, there are maybe two or three who could actually do something, but lack the time Oct 29 00:27:52 and people like me, who can utter wise comments but never implement any code Oct 29 00:28:24 I know how you feel, I've seen that, been thare Oct 29 00:28:26 there* Oct 29 00:29:41 after all SHR (#openmoko-cdevel) is a distro meant to run on N900 too Oct 29 00:30:17 I've done some mad packaging magic to get all closed bits working, and done compromises I would never even considered making. I've spent _shitloads_ of hours trying to figure out how to make this and that so everything does not break, compiling kernels, automating packaging, compiling some more, writing udev rules and shit like that. Oct 29 00:30:28 And all in all, it's all for nothing Oct 29 00:31:09 I cant port those missing/not working drivers, and every kernel release breaks more stuff and there is no-one to fix it. Oct 29 00:31:17 you should talk to SHR folks Oct 29 00:31:33 over at #openmoko-cdevel Oct 29 00:32:47 gnutoo, jama, mrmoku... Oct 29 00:32:52 well, this unknown obvious person already contacted me about shr related stuff, I handed him my kernel config and link to my github page. Oct 29 00:33:29 i guess the most fundamental problem is that people capable are all spread over different projects Oct 29 00:34:15 that Oct 29 00:34:30 can't blame them, I too have no interest in any other distribution than Arch. Oct 29 00:34:44 at some point, spending money actually works Oct 29 00:34:51 ofc I'm glad to share anything I do, and help and such Oct 29 00:35:07 money is sadly something I dont have any to spare :\ Oct 29 00:35:37 I mean to get people who have to concentrate on what you tell them to do Oct 29 00:36:33 otherwise, you either need to energise them in a community, or give it up. Oct 29 00:36:40 or do it yourself. Oct 29 00:37:06 by which time you finish all but the smallest project, nobody cares. Oct 29 00:37:18 yeah, very true. Oct 29 00:37:31 can I compile a new openssh for maemo? Oct 29 00:37:35 * SpeedEvil hopes in a couple of years nobody will care about maemo. Oct 29 00:37:38 the old version doesn't have some support that I need Oct 29 00:37:40 yeah, I "wasted" like 3 years of 50%..75% part time "job" to maemo (and openmoko, after the Inc gone bonkers), but since a year I can't do that anymore, my monetary reserves are used up and I need to do 'real' work for my bagels Oct 29 00:37:45 because Jolla will be awesome Oct 29 00:38:35 at some point even diehards lose focus, or even die. Oct 29 00:38:43 internetishard: I can't see any obstacles Oct 29 00:39:04 I did that Oct 29 00:39:17 ecdsa asks for a password when using the key even thought it is passwordlessss Oct 29 00:39:22 it was annoying for a reason I forget Oct 29 00:39:34 internetishard: permissions problem? Oct 29 00:39:43 I've got the impression that Jolla is going to just grab Mer as a base and just pour their closed source stuff all over it.. :\ Oct 29 00:39:59 it depends. Oct 29 00:40:13 actually closed source = bad Oct 29 00:40:41 closed source, with well documented APIs - hmm. Oct 29 00:40:50 SpeedEvil: no idea Oct 29 00:40:53 latest rumour has it they will jump to tizen Oct 29 00:41:09 chmod 600 on they key.. Oct 29 00:41:21 closed - from a buisness sense of licencing, but with source available, awesome Oct 29 00:41:31 internetishard: and the parent dir? Oct 29 00:42:09 anyways, it's Finnish company so I will support it nevertheless. Oct 29 00:42:42 non free licence with the ability of third parties to buy licences for other hardware - possibly awesome Oct 29 00:42:50 i have no trouble with closed source if it is done properly, and maintained properly Oct 29 00:43:49 memo style - sorta meh Oct 29 00:49:30 meh Oct 29 00:49:33 :) Oct 29 00:51:44 SpeedEvil: bah, it still thinks it needs a password too Oct 29 00:51:50 ecdsa is pretty new to ssh... Oct 29 00:52:09 version on here is 5.1, but on the server 6.1... Oct 29 00:52:15 Can I compile a new version on maemo? Oct 29 00:52:34 I'd like to do the same with sshfs Oct 29 00:55:10 anyways, after opening up, I would like to ask any of you who could actually do something to the kernel modules, to consider doing some forward porting. Most of the drivers that are more or less broken (sensors, audio) should be pretty easy to fix. Also, bluetooth driver needs forward porting. Oct 29 00:55:47 Ok, apparently the version of ssh on maemo doesn't support ecdsa, so I need to compile a new version of ssh and sshfs - how do I do this? Oct 29 00:56:40 Also, few of us are on #maemo-alternatives, if some feel they could answer stupid questions of me _trying_ to do that shit, feel free to join. Oct 29 00:57:06 I just copied / into a chroot, installed build-essentials from tools in that chroot, and went from there Oct 29 00:57:14 there are probably better ways Oct 29 00:57:43 SpeedEvil: is that to me? Oct 29 00:57:55 yes Oct 29 00:58:06 I can just download the regular openssh source and it will build on maemo? Oct 29 00:58:40 from memory, yes Oct 29 00:58:49 it's been a while thoyg Oct 29 00:59:27 interesting Oct 29 01:00:06 like, from the busybox repo? Oct 29 01:02:38 SpeedEvil: what would you choose? Oct 29 01:02:58 no Oct 29 01:03:11 I think I just grabbed openssl latest Oct 29 01:03:21 I'm talking about ssh here Oct 29 01:03:51 and SSH Oct 29 01:03:58 from??? Oct 29 01:04:10 why do you want SSH Oct 29 01:04:26 like I said, I need new ssupport Oct 29 01:04:32 in order to connect to servers using ecdsa keys Oct 29 01:05:54 SpeedEvil: soooo where do I get teh source? Oct 29 01:08:23 DocScrutinizer? Oct 29 01:11:32 the SSH source repo Oct 29 01:11:49 how about openssh.org? Oct 29 01:12:04 shouldnt really be that hard to find Oct 29 01:12:30 I didn't think you could just get any linux package and compile it on arm Oct 29 01:15:37 most of the time you can, sometimes it's not that easy Oct 29 01:15:46 like in this case :) Oct 29 01:16:19 oh, you're saying it won't work... Oct 29 01:16:42 if the required libraries in maemo are new enough for latest openssh, you should have no trouble Oct 29 01:16:50 :( Oct 29 01:16:52 This is no fun Oct 29 01:16:55 just try it Oct 29 01:17:27 maybe its "easy" using a chroot? Oct 29 01:17:31 since you only need a client Oct 29 01:17:33 I'm not saying it wont work, I'm saying there is a possibility it wont work Oct 29 01:17:52 chroot for what? Oct 29 01:18:33 a debian release with a newer version of ssh Oct 29 01:18:42 ohhh Oct 29 01:18:48 "use fkn debian, son" Oct 29 01:18:54 I never do that :X Oct 29 01:18:55 I should Oct 29 01:25:37 I hate when that happens. Usually I just grunt a simple "No, I don't want to." Oct 29 01:25:49 (use debian) Oct 29 01:26:23 why not? Oct 29 01:26:40 why should I? Oct 29 01:27:11 i only used debian based OSs and don't dislike them yet Oct 29 01:27:25 what do you use? Oct 29 01:28:09 Arch Oct 29 01:28:50 I use arch too Oct 29 01:29:19 I guess I could use arch in the maemo chroot instead! Oct 29 01:29:28 Sure Oct 29 01:29:42 there a downside of using a different distro from debian on maemo? Oct 29 01:29:52 or rather "for the maemo chroot" Oct 29 01:30:03 depends on what you're doing with it Oct 29 01:30:17 commandline stuff should be a breeze Oct 29 01:31:23 gry: I'm not saying I dislike debian, I just find few other distros to be more appealing to me personally. I've used debian in a past for quite some time. Oct 29 01:33:48 nod Oct 29 01:35:45 in the end it's just the same what you use, but I do tend to recommend the more minimalistic approach usually, and Mint for beginners (going to change that habit though) Oct 29 01:42:39 hmm, off to take care of the snakes, bbl Oct 29 01:46:04 :) Oct 29 02:03:34 pissed off blood python == no fun **** ENDING LOGGING AT Mon Oct 29 03:00:01 2012