IRC log of #maemo-ssu for Sunday, 2012-12-09

*** Mihanizat0r has joined #maemo-ssu00:01
illuminatais ssu known to still have reboot loop issues? i upgraded to it on the advice that it should solve those problems but they're back00:15
*** nox- has joined #maemo-ssu00:15
kerioilluminata: which kind of bootloop?00:19
illuminatadidn't know there were multiple kinds, but the one where the carrier name and widgets flash in and out00:21
kerioi'm not sure why you think cssu is responsible for that00:23
illuminatawas told something similar before, that some watchdog type program wouldn't force a reboot when a widget wasn't functioning properly00:24
illuminatasomething to that effect, it was a couple/few months ago. your nick is familiar too00:24
illuminataor maybe this was geared more for #maemo proper00:25
*** freemangordon_ has quit IRC00:39
*** Guest73590 has quit IRC00:41
DocScrutinizer05freemangordon: for "stop current task" please consider to avoid deadlock when process table filled (forkbomb). It's an old boring pun in unix that you can't start a process to kill a process when there's too many processes. So usual workaround is to have a process already running (usually a shell) where you can `exec killall <forkbomb>`01:15
DocScrutinizer05dunno if all that makes sense, since maybe even pressing powerbutton to open the menu already needs to start a new process. Then obviously there's no use in later on trying to make the "stop current task" menubutton forkbomb-safe01:18
DocScrutinizer05but I guess you're using the osso-kill-task.sh script (or what's it's name) anyway, so probably all that doesn't make too much sense01:20
DocScrutinizer05or maybe there's no such script and you better forget my last remark01:23
DocScrutinizer05otoh I guess OOM starts killing processes as soon as ther are too many01:26
*** toxaris has joined #maemo-ssu01:44
*** ruskie has quit IRC02:14
freemangordonDocScrutinizer05: no, it works in a different way (killing), will elaborate tomorrow02:20
DocScrutinizer05probably via x-msg02:21
freemangordonyep, _NET_CLOSE_WINDOW02:22
freemangordonit tries to do a graceful close, killing a hang process is done by hildon-desktop02:22
*** toxaris has quit IRC02:25
*** NIN102 has quit IRC02:28
*** ruskie has joined #maemo-ssu02:33
*** joshgillies has joined #maemo-ssu02:55
*** jonwil has joined #maemo-ssu03:13
*** Martix_ has quit IRC03:29
*** joshgillies has quit IRC03:33
*** M4rtinK2 has quit IRC03:54
*** arcean has quit IRC04:00
Palimerlin1991, in /srv/apt/cssu-devel/incoming are again some dsc/tar.gz files04:17
Palianother bug?04:17
* merlin1991 investigates04:18
merlin1991hm one is an ages old h-d verison04:19
merlin1991the other one looks like the current one04:19
merlin1991regarding the current version, it seems like arcean uploaded a binary only .changes file04:22
merlin1991+ the sources04:22
merlin1991I'll contact him when he gets back online04:23
merlin1991Pali: ^^04:23
Paliok04:23
*** joshgillies has joined #maemo-ssu04:24
*** Pali has quit IRC04:27
*** nox- has quit IRC05:08
*** amiconn has quit IRC05:09
*** amiconn_ has joined #maemo-ssu05:09
*** amiconn_ is now known as amiconn05:09
*** DocScrutinizer05 has quit IRC06:04
*** DocScrutinizer05 has joined #maemo-ssu06:04
*** jonwil has quit IRC06:19
*** dhbiker has joined #maemo-ssu06:55
*** Mihanizat0r has quit IRC09:18
*** toxaris has joined #maemo-ssu10:06
*** Guest73590 has joined #maemo-ssu10:36
*** Martix_ has joined #maemo-ssu11:22
*** NIN101 has joined #maemo-ssu11:58
*** Martix_ has quit IRC12:29
*** arcean has joined #maemo-ssu12:34
freemangordonarcean: hi12:42
arceanfreemangordon: hi12:42
freemangordonarcean: do you have time/will to look at tklock replacement, it uses too much memory for my taste, but unfortunately my GTK/clutter/etc skills are not good enough to optimize it :)12:43
freemangordonit does not leak though12:43
arceanhmm sure, I can take a look :)12:43
*** Pali has joined #maemo-ssu12:44
freemangordonarcean: it uses ~500k, even when not active :(12:45
freemangordonof heap that is12:45
freemangordonPali: why? https://gitorious.org/community-ssu/pulseaudio/commit/2aa134d13fe2e4d1f560a18fa0b028cf0b66a99012:45
freemangordonarcean: is it possible to use clutter actors for tklock, instead of GdkPixmap. and does it make sense?12:47
arceannot really, only if you want to do some animations/transitions12:50
freemangordonarcean: ok. Toldya, all this stuff is unknown to me :D12:51
arceanhehe :P12:52
freemangordongoing to shoot valgrind against tklock, just to be sure it does not leak (too much :P )12:54
*** M4rtinK2 has joined #maemo-ssu12:57
*** jonwil has joined #maemo-ssu13:02
Palifreemangordon, audio over network (module-native-protocol-tcp) not working correctly13:10
Palifreemangordon, or we need to increase memory block size and decrease count of blocks13:16
PaliI'm getting these error massages: [pulsecore/memblock.c:333 pa_memblock_new_pool()] Memory block too large for pool: 49956 > 1638413:16
freemangordonPali: you mean this is used only when using PA over the network?13:17
PaliPA over network via native protocol not working correctly with that nokia patch13:17
freemangordonPali: BTW when do you expect to get those 50MB?13:17
freemangordon*where13:17
PaliI do not know if memblock is used only in network code13:18
freemangordonis that a real allocation?13:18
Palifreemangordon, you can decrease PA_MEMPOOL_SLOTS_MAX from 1024 to 256 and then max size of memory for memblock will be same13:21
Paliwith nokia patch slots_max is 1024 and slot_size is 16*102413:21
Palibut network code not working with lower slot_size, so I reverted it back to 64*1024 and if you decrease slots_max from 1024 to 256 then total memory for use will be same13:22
freemangordonPali: my concern is: how will that patch affect PA memory usage in normal usecases?13:23
PaliI can check it13:25
DocScrutinizer05my concerns are even more fuzzy and like a gut feeling that messing with PA is not worth the effort13:27
DocScrutinizer05and most probably not a good idea anyway13:27
freemangordonDocScrutinizer05: well, if it does not affect day-to-day usage of the device, then (the patch) should be OK.13:28
freemangordonThough that should be verified very carefully13:28
DocScrutinizer05and who's verifying that?13:28
freemangordonexactly ;)13:28
kerioPali is!13:29
DocScrutinizer05stutter, subtle deterioration of audio quality, complete bork in special usecases, whatnot else... a nightmare to verify13:30
freemangordonagree, that's why I started questioning13:30
DocScrutinizer05particularly since our audio isn't all that bugfree right now13:30
jonwilNot to mention that we have a bunch of binary blobs that make it very hard to do too much with PulseAudio13:31
freemangordonPali: Please, check the memory allocation/usage with and without your change, Nokia did that patch on purpose after all13:32
Palibtw, what to use for checking memory allocation?13:32
DocScrutinizer05so for any new bug report, it's up to anybody's guess whether it got introduced with latest patch, the patch 3 months ago, or been there from beginning even in stock PA and only rears its ugly head now. Or maybe is completely unrelated, or caused by some evil side-effect like changed race conditions that now trigger the sleeping bug in a unrelated subsystem13:33
freemangordonPali: well, use smaps13:33
freemangordon"/proc/$PA_PID/smaps"13:34
Paliheap --> size: 768kB13:36
freemangordonlooks ok to me. and what happens when use it over the network?13:38
freemangordonBTW I still not get it why systemui uses ~1600KB :(13:39
*** Jaded has quit IRC13:40
jonwilmemory leak in the systemui binary?13:40
freemangordonjonwil: no13:41
freemangordonit does not leak, for sure13:41
freemangordonthe first time you open powerkey menu or tklock, memory usage jumps with ~500k, some Gtk stuff, but I cannot get how to force Gtk to free that memory13:42
jonwilI do wish we knew just what the PulseAudio binary blobs actually DO, that would mean we could find replacements for each piece :)14:15
freemangordonjonwil: one of them is highpass filer14:16
freemangordon*filter14:16
freemangordonthe other is maybe echo cancelation14:17
jonwilalso speaker protection14:18
jonwiland also sidetone I think14:18
freemangordon(highpass filter that is ;) )14:18
freemangordonsidetone?14:18
jonwil]http://en.wikipedia.org/wiki/Sidetone14:19
jonwilhttp://en.wikipedia.org/wiki/Sidetone14:19
freemangordonaah, yeah14:19
jonwilyep, module-nokia-voice references sidetone14:19
freemangordonjonwil: I don;t think we'll be capable of replacing that14:20
jonwilheh yeah14:20
jonwilwell in theory we could look at what MeeGo/Mer/Nemo has since they seem to have newer PulseAudio running on N900 without any binary blobs at all14:20
freemangordonBTW did anyone tried those modules with newer PA?14:20
freemangordon*try14:21
*** toxarisswe has joined #maemo-ssu14:28
jonwilI bet the modules are tied to specific layout of PulseAudio structures that have changed between versions14:28
jonwilor PulseAudio API calls14:29
*** toxaris has quit IRC14:30
*** _rd has joined #maemo-ssu14:38
jonwilso the answer is I suspect we are stuck on current PulseAudio unless we can figure out all algorithms in the binary blobs and find replacements where needed :)14:38
*** RST38h has quit IRC14:41
*** Guest73590 has quit IRC14:41
*** RST38h has joined #maemo-ssu14:42
Palijonwil, some nokia pulseaudio blobs was opened see meego git14:47
Palionly algorithms are still closed14:47
jonwilyeah MeeGo has open bits14:48
Palifrom info what I have pulse audio nokia algorithms doing some magic for better voice support in gsm calls14:48
jonwilwe still dont know what the closed-bits that weren't opened do or which of those bits actually matter for functioning phone :)14:48
Paliso we need something which can fix buggy gsm codec14:49
Paliit doing some voice correction or something like that14:49
Paliand some speaker protection14:50
arceanfreemangordon: what about http://gitorious.org/community-ssu/osso-systemui-tklock/blobs/master/osso-systemui-tklock.c#line801 ?14:53
arceando we need to copy the style or only the window's background?14:54
freemangordonarcean: this comment is mine, btu I really don't know :)15:20
freemangordonarcean: it was done that way in the stock tklock, I googled a bit and find the commented code15:22
freemangordon"gdk_window_set_back_pixmap( GDK_WINDOW (window), pixMap, TRUE);"15:23
freemangordonarcean: try it15:23
arceanIIRC the correct way is to create a new style and then apply the background image15:26
freemangordonarcean: could be, I don't know, thats why I kept the stock way15:27
*** NIN102 has joined #maemo-ssu15:45
*** NIN101 has quit IRC15:45
*** joshgillies has quit IRC15:50
*** Pali has quit IRC15:59
*** Pali has joined #maemo-ssu16:01
*** freemangordon__ has joined #maemo-ssu16:11
*** freemangordon has quit IRC16:11
*** freemangordon__ is now known as freemangordon16:11
*** _rd has quit IRC16:24
*** ekze-nyan has quit IRC16:49
*** jonwil has quit IRC17:13
*** _rd has joined #maemo-ssu17:32
merlin1991arcean: ping17:41
arceanpong merlin199117:43
merlin1991when you uploaded the latest h-d to -devel it looks like the .changes file was from a binary only build17:43
merlin1991the sources were left behind in the incoming dir and not imported17:44
arceanso I've uploaded the wrong file?17:44
merlin1991I have no idea because the .changes file is gone by now, but if you look here: http://maemo.merlin1991.at/cssu/community-devel/pool/free/h/hildon-desktop/17:45
merlin1991you'll notice the sources are of an older version, and there were no errors in the import log17:45
arceanuhh17:46
merlin1991the .tar.gz and .dsc is there so I should be able to just import those, but I don't even know if the debs there were build of exactly that source :/17:47
arceanif it's on the server, then it's exactly the source we're looking for17:49
merlin1991I've imported them by hand now17:53
arceanthanks :)17:54
*** rd_ has joined #maemo-ssu17:55
*** rd_ is now known as Guest2883317:55
*** _rd has quit IRC17:58
*** Guest28833 has quit IRC18:02
*** Guest28833 has joined #maemo-ssu18:02
*** dhbiker has quit IRC18:29
*** dhbiker has joined #maemo-ssu18:43
*** freemangordon has quit IRC20:14
*** freemangordon__ has joined #maemo-ssu20:14
*** Martix_ has joined #maemo-ssu20:22
*** Jade has joined #maemo-ssu21:56
*** Jade has joined #maemo-ssu21:56
*** arcean has quit IRC22:07
*** arcean has joined #maemo-ssu22:08
*** _rd has joined #maemo-ssu22:13
*** Guest28833 has quit IRC22:15
*** NIN101 has joined #maemo-ssu22:35
*** nox- has joined #maemo-ssu22:35
*** NIN102 has quit IRC22:36
*** NIN102 has joined #maemo-ssu23:12
*** NIN102 has quit IRC23:12
*** NIN101 has quit IRC23:20
*** NIN102 has joined #maemo-ssu23:20
*** NIN101 has joined #maemo-ssu23:26
*** NIN102 has quit IRC23:27
*** NIN101 has quit IRC23:28
*** NIN102 has joined #maemo-ssu23:28
*** joshgillies has joined #maemo-ssu23:40

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