**** BEGIN LOGGING AT Wed Aug 25 02:59:57 2010 Aug 25 08:49:53 hi mickey|zzZZzz Aug 25 08:49:54 hmmm Aug 25 08:54:47 JaMa|GoNe, looks a piece of the previous patch is gone into this http://cgit.openembedded.org/cgit.cgi/openembedded/tree/recipes/linux/linux-openmoko-2.6.32/0004-ar6000_delay.patch.patch Aug 25 08:56:54 xrmx: right I noticed it later, but fixed it only in 2.6.34.. Aug 25 08:58:01 JaMa|GoNe, do you remember to have applied this patch to 2.6.29 kernel? http://docs.openmoko.org/trac/attachment/ticket/2277/wifi_ifupdown.patch Aug 25 08:58:47 xrmx: for 2.6.29 we were using only stuff already pushed to andy-tracking.. so check there Aug 25 08:59:07 JaMa|GoNe, ok thanks Aug 25 08:59:13 yw Aug 25 09:10:15 xrmx: usbhost patch fixed Aug 25 09:13:05 xrmx: do you mean you still have some odd problems with ar6000? Aug 25 09:14:56 PaulF, i can't get an ip from my router with debian 2.6.29 but worked with an old shr i have in nand Aug 25 09:16:19 xrmx: i'm not sure i'm following Aug 25 09:17:37 PaulF, i have an old version of shr (last autumn) on my nand, wifi was working. I have debian in my microsd and i can't get an ip address from my router Aug 25 09:17:54 JaMa|GoNe: wha't the error of unable to register miscdev on minor 130 while booting kernel?? Aug 25 09:18:19 xrmx: without kernel rev it's all pretty moot talk i'd say Aug 25 09:18:36 Q-Master: watchdog Aug 25 09:18:53 PaulF, debian have latest andy-tracking 2.6.29, shr can't remember but i think the same Aug 25 09:19:02 JaMa|GoNe: is it needed?? Aug 25 09:28:42 <[Rui]> hi all Aug 25 09:35:41 xrmx: debian is no-debug .29, while shr is debug. removeing debug brings few problems. one of this problems (device up intability) is fixed by my ar6000 delay patch. if it is absent (and it is absent in .29, both in debian and shr, but present in qtmoko), ifup fails frequently. with patch it fail because of some other reason in much fewer (5%-10% cases) here. Aug 25 09:38:05 xrmx: i mean old shr of course Aug 25 09:39:12 gena2x, the device goes up it's just not able to get an ip address, now i have a kernel with debug but haven't tried, let me check Aug 25 09:41:43 xrmx: i did fix looong time ago. can't recall detail, may be 'up but no address' is state of this bug. i had to enable some debugging inside driver to really know reason Aug 25 09:41:56 speaking about kernel, I read that lot of things were poushed mainstream. does this mean we could see a fresher kernel for our freerunner? or is it too much work and not enought labor? Aug 25 09:42:36 GarthPS: ThibG currently pushing more things. Aug 25 09:42:57 GarthPS: we have .32 and .34. what else do you want to have? Aug 25 09:43:59 gena2x: oh I did not know about .34 kernel :) and don't get it wrong : .32 is already awsome! :p Aug 25 09:44:55 GarthPS: some bugs still exist, like high power consumption in suspend, but nothing really critical. Aug 25 09:44:59 GarthPS: 2.6.34.5 http://build.shr-project.org/shr-kms/ Aug 25 09:45:58 GarthPS: i think all this will slowly be fixed as people will have time. so imo, kernel state on fr is 'very well'. many major parts upstream, kernel is very stable and in very good shape Aug 25 09:46:52 gena2x: in fact I am wondering about this because of some glitch with .32 kernel , like not sound throught HP, acceleromettre which are interpreted with an angle ofset , perhaps vibro not completely right Aug 25 09:47:16 diff shr-kernel and upstream: -rw-r--r-- 1 bitbake bitbake 1.7M Aug 25 11:45 gta02-2.6.34.diff Aug 25 09:47:22 gena2x: ahh about consumption I thought it was my battery :) Aug 25 09:47:56 gena2x: but yeah tour kernel are not so bad I agree too! Aug 25 09:48:10 xrmx: i advise no to turn general debug on of course, but (if you want to properly fix problem) enable ar debug and see why it is failing Aug 25 09:50:03 JaMa|GoNe: 95% of '1.7M' that is kms and accelerometers. Aug 25 09:50:26 and ar6000 Aug 25 09:50:57 JaMa|GoNe: and now glamo is going upstream. (without kms, but still diff be decreased). Aug 25 09:51:33 I just wanted to show that diff is not that small (code-wise - 1.7M) Aug 25 09:51:48 but you're ofc right that most important parts are there Aug 25 09:53:24 JaMa|GoNe: and that now (thanks to larsc, and whom else merge all this?) it is much, much better than andy-tracking. and most amazing - almost everything works and works relatively fast :) Aug 25 09:54:08 no wsods, partition corruptions, events/0 threads. Aug 25 09:57:45 guys where can I check known bugs in those new kernel: .32 and newer. to be sure to repport un-already-known bugs like I quote upper Aug 25 09:58:19 https://docs.openmoko.org/trac/ Aug 25 09:58:34 => no sound through HP, angle offset with accelero Aug 25 09:59:08 this sounds like shr-specific Aug 25 09:59:23 gena2x: I thought this one was deprecated.. Aug 25 10:01:39 JaMa|GoNe: no sound through HP is shr-specific? Aug 25 10:07:20 GarthPS: basically sound depends on mixer settings Aug 25 10:07:47 GarthPS: and only after that on kernel driver Aug 25 10:08:12 GarthPS: so, first place to check is distribution state files Aug 25 10:09:09 i don't know how shr guys recomment to report bugs against shr. Aug 25 10:09:40 gena2x: yep I just though that if they (shr-u) published a new image with a new kernel with this bug (just a stat file to right fill) that it was not just that simple otherwise they would already fix it Aug 25 10:10:30 uou are right i did not check if it was repported in their tracker Aug 25 10:12:50 gena2x, can you see any hint here? http://paste.debian.net/85688/ Aug 25 10:14:45 xrmx: i see that you didn't enabled debugging in ar6000 Aug 25 10:15:22 xrmx: but i am not true expert in ar. Aug 25 10:15:52 xrmx: also i do not see kernel module load log Aug 25 10:16:59 xrmx: to reload in-kernel ar stuff, you may use following: Aug 25 10:17:02 P=/sys/bus/platform/drivers/s3c2440-sdi Aug 25 10:17:02 echo s3c2440-sdi > $P/bind; dmesg | tail -n 20; echo s3c2440-sdi > $P/unbind Aug 25 10:17:17 without dmesg of course Aug 25 10:20:06 xrmx: to enable debug you should add 'EXTRA_CFLAGS += -DDEBUG' to drivers/ar6000/Makefile Aug 25 10:20:32 and rebuild module/kernel. Aug 25 10:21:25 xrmx: then you should do same thing with working kernel Aug 25 10:22:09 and find diff in debug messages. this way you probably will be able to fix problem without knowledge of ar internals. Aug 25 10:22:21 sorry it this is obvious. Aug 25 10:22:35 s/it/if/ Aug 25 10:22:35 gena2x meant: sorry if this is obvious. Aug 25 10:24:49 gena2x, ok, compiling Aug 25 10:27:43 xrmx: and enable printk timing information. Aug 25 10:28:25 gena2x, was already compiled i think Aug 25 10:35:23 gena2x: reported https://docs.openmoko.org/trac/ticket/2362 Aug 25 10:37:27 GarthPS: you have to include as a minimun your distribution name and version into reports. Aug 25 10:38:32 gena2x: I specified SHR-U but not which one you are right.. Aug 25 10:38:57 GarthPS: ah, sorry Aug 25 10:39:03 GarthPS: missed title :) Aug 25 10:39:07 gena2x: in the titel :) Aug 25 10:39:20 gena2x: lol ok so it is ok no ? Aug 25 10:39:31 Gabrys: ok of course Aug 25 10:39:58 gena2x: it was for me this last sentence ? Aug 25 10:40:13 sure :) Aug 25 10:40:17 ;) Aug 25 10:40:19 Gabrys: sorry Aug 25 10:49:15 hello all, sorry to bug you but I'm looking at libpohone-ui-shr code and trying to work out where all the g_debug() messages go to. I've been searching Internet and any info I get is on the various calls but not how you configure it. I can't see any of the logged info in /var/log so assume it's switched off. Aug 25 11:14:37 gena2x, i don't have any debug message with CONFIG_AR6000_WLAN_DEBUG set http://pastebin.com/aJLdbS0x Aug 25 11:37:44 Arigead: the code is here http://git.shr-project.org/ and there is deug level setting in /etc/phonefsod.conf log_level=INFO Aug 25 11:39:16 xrmx: did i told to use CONFIG_AR6000_WLAN_DEBUG? Aug 25 11:39:34 gena2x, nope, compiling with -DDEBUG Aug 25 11:41:49 xrmx: i helped as much as i can Aug 25 12:02:08 hi mickey|zzZZzz Aug 25 12:02:14 again....sigh Aug 25 12:02:46 does someone knows frameworkd's ogpsd subsytem here? does someone knows the internal Aug 25 12:03:07 I've to implement a NMEA thing for a device which uses its own resource activator Aug 25 12:03:31 I mean the "own resource activator" registers Enable() and Disable() Aug 25 12:05:21 should I revert that: http://git.freesmartphone.org/?p=framework.git;a=commitdiff;h=a355a0f8c6374ded40fe5446cadfb0646c3a3172;hp=77b887da0d72a9983e6f214d2208fe39ae8f90d7 Aug 25 12:05:22 ? Aug 25 12:05:50 how do I prevent registering the resource? Aug 25 12:09:27 basically maybe someone else than mickeyl could help Aug 25 12:09:36 because there is also a python "issue" Aug 25 12:09:49 I copied the eten class Aug 25 12:09:54 which inherit the nmea class Aug 25 12:10:14 which inherit the gpsdevice class Aug 25 12:10:23 which inherit the resource class Aug 25 12:10:24 so... Aug 25 12:10:32 if I put that in my eten class: Aug 25 12:10:38 def _enable( self, on_ok, on_error ): Aug 25 12:10:41 pass Aug 25 12:10:46 what will it do? Aug 25 12:13:45 will the enable of the cloned eten class be called? Aug 25 12:14:17 vanous: thanks for that I'll try it out. I already have the code locally and I'm building it and messing with it for investigative purposes ;-) Aug 25 12:14:18 instead of someone of an inherited class Aug 25 12:14:25 when someone does: Aug 25 12:14:28 self._enable Aug 25 12:14:33 in an inherited class Aug 25 12:47:13 it's ultra-unproductive that way Aug 25 12:47:18 does someone knows ogpsd? Aug 25 12:48:45 mickey|patio, wow nice thanks a lot for comming Aug 25 12:49:07 mickey|patio, I've tried to accelerate a bit things by trying to have gps working out of the box but I failed Aug 25 12:49:24 I've again spent countless hours hitting my head against the walls Aug 25 12:49:34 by work out of the box I mean: Aug 25 12:49:42 * I didn't do any autotools work Aug 25 12:49:48 * I tried to do that: Aug 25 12:50:14 ./MsmGpsResourceActivator Aug 25 12:50:24 and to register the resource Aug 25 12:50:31 allong with the already used gps system Aug 25 12:50:43 I mean if I disable fsotldt and frameworkd's ogpsd Aug 25 12:50:46 it works Aug 25 12:50:53 but if I keep ogpsd it doesn't Aug 25 12:51:13 I've tried to make a msm.py gps thing Aug 25 12:51:17 copying eten.py Aug 25 12:51:20 but I failed again Aug 25 12:51:27 I rebased that: Aug 25 12:51:38 http://git.freesmartphone.org/?p=framework.git;a=commitdiff;h=a355a0f8c6374ded40fe5446cadfb0646c3a3172;hp=77b887da0d72a9983e6f214d2208fe39ae8f90d7 Aug 25 12:51:42 reverting it Aug 25 12:51:47 s/rebased/reverted Aug 25 12:51:58 still no go: it doesn't work anymore Aug 25 12:52:05 because the dbus paths are not there anymore Aug 25 12:52:10 there is only / Aug 25 12:52:11 on ogpsd Aug 25 12:52:13 so... Aug 25 12:52:20 what should I do? Aug 25 12:52:44 I don't feel capable or would have time for rewriting ogpsd in vala btw Aug 25 12:53:03 mickey|patio, what do you think? Aug 25 12:54:06 there are several ways to proceed from here. i guess the major question is which apps do you want to use GPS with? new ones or existing ones? Aug 25 12:54:17 if existing ones, which protocol do they speak? Aug 25 12:54:29 I want to do that: Aug 25 12:54:57 settings->position->manual->on Aug 25 12:55:05 and have the posision etc....displayed under Aug 25 12:55:25 btw what do you mean by new-ones and existing ones Aug 25 12:55:40 I want fso-gpsd compatibility Aug 25 12:55:44 like for instance: Aug 25 12:56:10 hi, is it right place to ask kernel-related questions? Aug 25 12:56:12 I launch some program like OpenBMap Aug 25 12:56:30 anarsoul, it depends on your question Aug 25 12:56:43 anarsoul, but there is a #kernelnewbies channel not on freenode Aug 25 12:56:55 it's on oftc Aug 25 12:57:08 GNUtoo|laptop: I need some testing for my s3c24xx-related patches on s3c2410 (gta01) and s3c2442 (gta02) devices :) Aug 25 12:57:10 but if it has something to do with free phones or an openmoko driver.... Aug 25 12:57:16 ah ok Aug 25 12:57:20 so it's the right place Aug 25 12:57:43 here're patches: http://www.spinics.net/lists/arm-kernel/msg96212.html Aug 25 12:58:17 with this patchset it's possible to use DMA for s3cmci driver Aug 25 12:58:45 according to the git log it brings speedups which is nice Aug 25 12:59:16 actually, it brings some CPU time for other things Aug 25 12:59:28 mickey|patio, so ? what should I do? what are older/newer apps? Aug 25 12:59:29 than reading/writing from/to FIFO Aug 25 12:59:39 anarsoul, have you cc'ed ben dooks? Aug 25 12:59:45 xrmx: of course Aug 25 12:59:52 the situation is quite ugly Aug 25 13:00:01 we have an obsolete API Aug 25 13:00:11 ahh ok Aug 25 13:00:17 and we have an adapter that converts an obsolete API into the "standard" Aug 25 13:00:26 ideally we would have a) a new API Aug 25 13:00:29 ok that adapter is fso-gpsd I bet Aug 25 13:00:31 and b) talking the standard Aug 25 13:00:34 correct Aug 25 13:00:39 hmmm Aug 25 13:00:46 and using something like fso-gpsd? Aug 25 13:00:47 this is what i want to do with fsotdld Aug 25 13:00:49 oops Aug 25 13:00:50 sorry Aug 25 13:00:56 fsotdld should be integrating fso-gpsd Aug 25 13:01:01 but not converting gypsy into gspd API Aug 25 13:01:02 and using something like gpsd under it Aug 25 13:01:03 but rather our new one Aug 25 13:01:19 anarsoul, so you are the guy who ported the GENERIC_TIME patch right? thanks for doing so Aug 25 13:01:39 xrmx: yep, you're welcome Aug 25 13:01:40 until this happens, there are two options Aug 25 13:01:55 a) write an ogspd device plugin for external NMEA resources Aug 25 13:02:00 anarsoul, are you actually using it with no_hz? Aug 25 13:02:06 I have version of this patch that _might_ work with suspend, will check and submit it later Aug 25 13:02:06 or b) use gpsd directly from /dev/smd27 Aug 25 13:02:08 xrmx: yes Aug 25 13:02:15 mickey|patio, I want to do a) Aug 25 13:02:21 ok Aug 25 13:02:25 then you have to do some python :) Aug 25 13:02:31 I know python a bit Aug 25 13:02:37 but I failed at doing so Aug 25 13:02:41 i will add the activator to the android-rpc now Aug 25 13:02:42 do you have any pointers Aug 25 13:02:48 xrmx: it allows to save ~30 mA on my PDA (90 mA -> 60 mA in idle) Aug 25 13:02:53 didn't look at the code for long, and it's not even my code Aug 25 13:02:58 i'd have to look at it first Aug 25 13:03:01 mickey|patio,basically here are the required pointers: Aug 25 13:03:07 there is a NMEA class Aug 25 13:03:12 anarsoul, sound interesting :) Aug 25 13:03:14 if I use it it register the GPS resource Aug 25 13:03:19 and I can't reigster mine Aug 25 13:03:27 I want to get rid of this registration Aug 25 13:03:38 in the msm plugin which inherit nmea Aug 25 13:04:19 btw, what about s3cmci/dma-related patches? :) are there any volunteers to test them? Aug 25 13:04:54 anarsoul: may be just decreasing HZ will be enought to archive same effect? Aug 25 13:05:18 gena2x: maybe Aug 25 13:07:14 if it's for example 30ma decreasing hz 4 times should make it 8ma, which sounds acceptable. i tried to builds with hz100, but it wont boot, i have no chances to debug it as no dboard :( Aug 25 13:08:07 mickey|patio, 3) Aug 25 13:08:09 use gpsd Aug 25 13:08:16 and provide compatibility with old api? Aug 25 13:08:35 gena2x: I'm not sure that decreasing HZ would help, NOHZ _really_ decreases irq count per second Aug 25 13:08:37 I want openBmap.... Aug 25 13:08:43 I want to play with some gps apps Aug 25 13:09:05 what about gps providers Aug 25 13:09:10 like the openBmap provider? Aug 25 13:09:23 openbmap is not a provider Aug 25 13:09:29 ah? Aug 25 13:09:30 it's a consumer, isn' it? Aug 25 13:09:35 anarsoul: but HZ100 should _really_ decrease twice amount of irqs generated in idle and ++ less task switch :) Aug 25 13:09:35 both Aug 25 13:09:40 there is in oe a consumer Aug 25 13:09:42 but out of oe Aug 25 13:09:44 there is a git Aug 25 13:09:49 with a location provider Aug 25 13:09:54 that uses openbmap data Aug 25 13:10:10 yeah, frameworkd has an old copy of that data Aug 25 13:10:24 the easiest way is to fix frameworkd to use the new GPS resource Aug 25 13:10:27 on dream Aug 25 13:10:29 anarsoul: less task switch is benefit in less cache clean, while load :) Aug 25 13:11:06 indeed Aug 25 13:11:09 so option 3 Aug 25 13:11:25 but I fear that I will hit my head against walls again Aug 25 13:11:45 wait a sec Aug 25 13:11:53 new gps resource means what option? Aug 25 13:11:58 1,2 or 3? Aug 25 13:12:36 adding a new class to frameworkd Aug 25 13:12:50 1) write an ogspd device plugin for external NMEA resources Aug 25 13:12:54 2)use gpsd directly from /dev/smd27 Aug 25 13:13:10 3) use gpsd directly from /dev/smd27 + make an ogpsd compatibility layer Aug 25 13:13:37 nah, 3) makes no sense. no one is using the obsolete API except the adapter Aug 25 13:14:16 anarsoul: also, i think HZ60 is more than enought for hardware, and if this also decrease /4 power consumption, (i thought of it as a more performance due to less task switch), this sounds even more sweet. Aug 25 13:15:30 what about that: Aug 25 13:15:31 dbus.exceptions.DBusException: org.freedesktop.DBus.Error.ServiceUnknown: The name org.freesmartphone.ogpsd was not provided by any .service files Aug 25 13:15:49 self._dbusobj = dbus.SystemBus().get_object('org.freesmartphone.ogpsd', '/org/freedesktop/Gypsy') Aug 25 13:15:56 that's from openBmap Aug 25 13:16:03 the gps consumer version Aug 25 13:16:12 should I fix that app? Aug 25 13:16:20 gena2x: I suspect that Ben chose 200HZ for some reason :) you can look through plat-samsung/time.c to check what's wrong when HZ is 100 Aug 25 13:16:45 what do you mean "fix"? it's looking for frameworkd Aug 25 13:17:01 if you provide the ogspd device plugin, it will find Aug 25 13:17:41 gena2x: and, btw, with nohz power consumption is only 30% lower (90 mA -> 60 mA) Aug 25 13:17:41 one sec...I'll reread Aug 25 13:17:46 anarsoul: my guess, this value just were somewhere in samsumg evaluation board sdk, so nobody actually choose it. Aug 25 13:18:16 gena2x: sure, our glamo eats much more than cpu, but 30mA still matter. Aug 25 13:18:42 mickey|patio, ahh ok so you were talking about option 1? Aug 25 13:18:49 how would that work Aug 25 13:18:58 I mean, I want to use NMEA class Aug 25 13:19:02 anarsoul: my guess is based on older view of defconfigs of different arm boards Aug 25 13:19:05 and you didn't look at the code Aug 25 13:19:09 for a long time Aug 25 13:19:13 who should I talk to? Aug 25 13:19:30 the only ones who did are gone Aug 25 13:19:38 ouch Aug 25 13:19:40 i could look, but not now Aug 25 13:19:52 let me add the activator now Aug 25 13:19:57 ok Aug 25 13:20:09 thanks a lot!!! Aug 25 13:20:54 gena2x: it seems that TIMER_USEC_SHIFT from plat-samsung/time.c should be changed too when you change HZ Aug 25 13:21:09 anarsoul: interesting Aug 25 13:21:20 however not sure Aug 25 13:21:24 anarsoul: i grep cat * |grep CONFIG_HZ in arm defconfigs Aug 25 13:21:29 what about writing something new in vala? Aug 25 13:21:39 I could use/copy the android lib's code Aug 25 13:22:03 seem values are - 100, 128 and 200 Aug 25 13:22:10 100 - 111 cases Aug 25 13:22:15 128 - 25 cases Aug 25 13:22:21 and 200 - 9 cases Aug 25 13:23:07 ah makes no sense Aug 25 13:23:23 hmmm I'm doing too much things at once Aug 25 13:23:25 let me look Aug 25 13:23:43 gena2x: AFAIK there's generic_clock on pxa and omap Aug 25 13:23:47 so we with our 'older hardware' have higest values of HZ :) Aug 25 13:24:11 almost all new hardware has generic_clock support :) Aug 25 13:24:56 configs with HZ_200: gta02_defconfig gta02_drm_defconfig mini2440_defconfig s3c2410_defconfig s5p6440_defconfig s5p6442_defconfig s5pc110_defconfig s5pv210_defconfig tct_hammer_defconfig Aug 25 13:25:09 gena2x: btw, timer interrupt doesn't mean that context will be switched Aug 25 13:25:23 so... Aug 25 13:25:28 gpsd is a tcp daemon Aug 25 13:25:28 but this is granularity of sheduler Aug 25 13:25:29 right Aug 25 13:25:34 yes? Aug 25 13:25:35 fso has a dbus api Aug 25 13:25:44 for gps Aug 25 13:25:44 right Aug 25 13:25:47 some apps use that gps api Aug 25 13:25:53 like openbmap Aug 25 13:26:00 are there a lots of app who does? Aug 25 13:26:05 gena2x: yep, it seems so Aug 25 13:26:08 so, if we have 2 tasks with same priotity, it will shitch it each HZ. Aug 25 13:26:10 IIRC no one except the adapter uses gypsy Aug 25 13:26:19 ok Aug 25 13:26:21 gypsy is a zombie Aug 25 13:26:22 adapter means? Aug 25 13:26:31 fso-gpsd Aug 25 13:26:34 ah ok Aug 25 13:26:52 we want fsotdld to speak two APIs Aug 25 13:27:01 one is gpsd for all those legacy apps Aug 25 13:27:08 one is a new location API Aug 25 13:27:23 taking into account what we learned when working with gypsy Aug 25 13:27:36 and geoclue et. al. Aug 25 13:28:19 anarsoul: ok, thanks, i'll give it other try with plat-samsung/time.c changed. a bit later, as now have some real life things to do. Aug 25 13:28:51 good luck :) Aug 25 13:29:25 anarsoul: but this anyway directly related to NOHZ. it NOHZ reduce power consumtion on X, chnaging HZ will reduce it X/2 or X/4. Aug 25 13:29:31 anarsoul: thanks :) Aug 25 13:29:53 openBmap does Aug 25 13:29:53 should I modify openbmap to use the new api? Aug 25 13:29:53 so everyone uses the tcp part? Aug 25 13:29:53 right Aug 25 13:29:53 no dbus for no-one exept fso-gpsd Aug 25 13:29:54 ? Aug 25 13:29:55 if an app uses dbus Aug 25 13:30:02 freesmartphone.org: 03mickey 07utilities * rfba54b698e23 10/android/rpc/ (7 files in 2 dirs): android-rpc: add fso-msm-gps-activator Aug 25 13:30:28 are you 100% sure that openbmap uses gypsy? Aug 25 13:30:41 hm, nm Aug 25 13:30:42 they do Aug 25 13:31:15 indeed Aug 25 13:31:17 yes, to ease the transition, it might be good to add gpsd as an alternative source to openbmap Aug 25 13:31:24 at least until i have our new location API in place Aug 25 13:31:27 ok I'll do that then Aug 25 13:31:27 which can take some months Aug 25 13:31:31 ok Aug 25 13:32:41 thanks a lot Aug 25 13:33:10 I'll be back very soon Aug 25 13:34:07 oki Aug 25 13:36:21 freesmartphone.org: 03mickey 07utilities * r740b9fbab4b2 10/android/rpc/ (6 files in 2 dirs): android-rpc: install dbus service configuration and shadow resource configuration for fso-msm-gps-activator Aug 25 13:38:21 freesmartphone.org: 03mickey 07utilities * r869629dc9d7d 10/android/rpc/AUTHORS: android-rpc: authors++ Aug 25 13:40:55 anarsoul: btw, why your patches rely on watchdog times? i've read that s3c has 5 timers, why not use other one? Aug 25 13:42:35 anarsoul: i mean 5+watchdog Aug 25 13:47:04 freesmartphone.org: 03mickey 07utilities * rf76b19ec3003 10/android/rpc/fso-msm-gps-activator/Makefile.am: android-rpc: cosmetics Aug 25 13:58:27 gena2x: it has 4 + watchdog Aug 25 13:59:20 and it is not my patch, andrzej zaborowski is original author Aug 25 13:59:30 I just ported it to recent kernel Aug 25 14:00:05 oops, it has 5 timers, you're right Aug 25 14:00:36 btw, here's comment on that: /* GTA02 uses all PWM outputs (0-3) for LEDs */ Aug 25 14:18:53 hi ,is there a list of program that doesn't use gpsd tcp interface Aug 25 14:19:04 but uses fso's gps dbus interface Aug 25 14:19:12 does settings do that? Aug 25 14:21:53 anarsoul: omg. all for leds! what a trashing of resources!! Aug 25 14:23:27 anarsoul: can't all leds share 1 pwn? Aug 25 14:23:35 *pwm Aug 25 14:24:34 hmmm the settings uses dbus too Aug 25 14:24:36 too bad Aug 25 14:24:39 I bet a lots of apps do Aug 25 14:24:45 hi JaMa|GoNe Aug 25 14:25:57 gena2x: I'm not familiar with gta02 hardware :) I'm using another s3c2442-based PDA Aug 25 14:26:45 WAIT Aug 25 14:26:51 anarsoul: stay here :) Aug 25 14:27:11 anarsoul: can you test memory r/w speed of your device? Aug 25 14:27:17 anarsoul: please :) Aug 25 14:27:34 gena2x: what tool should I use? Aug 25 14:27:56 anarsoul: i have mine, you may also use lmbench. Aug 25 14:28:29 what is system on your pda? Aug 25 14:28:44 glibc or uclibc? Aug 25 14:29:27 anarsoul: wait a bit, i'll put my small test to web Aug 25 14:29:35 both, I'm using uclibc-based initramfs with busybox for testing kernel stuff and glibc-based angstrom Aug 25 14:30:06 gena2x: I can test it only in few hours, I have no usb cable with me Aug 25 14:30:20 ok, no hurry :) Aug 25 14:32:54 gena2x: I doubt that it much differs from gta02, memory clock is same (100mhz iirc) :) Aug 25 14:33:31 gta is 200mhz. but i want to test raw memory r/w speed. Aug 25 14:33:57 ah. Aug 25 14:33:59 HZ Aug 25 14:34:01 not HZ Aug 25 14:34:04 memory clock. Aug 25 14:34:05 ok Aug 25 14:34:19 just want to check it. Aug 25 14:34:24 ok :) Aug 25 14:34:26 this will not take much time Aug 25 14:36:01 ok, just publish sources of your test somewhere Aug 25 14:36:13 sure, wait a bit. Aug 25 14:43:21 gena2x, is your patch touchscreen_ignoreunexpectedintr29.patch needed on om-gta02 2.6.34? Aug 25 14:43:40 xrmx: sure. Aug 25 14:44:12 xrmx: you may notice this if your ts will hang. Aug 25 14:44:29 xrmx: but it is not 29.patch, but 34patch Aug 25 14:44:59 xrmx: look to track, .29 and .32 share 1 ts driver, .34 has other driver. Aug 25 14:45:20 gena2x, anyway the patchs is still not in git.openmoko.org Aug 25 14:45:20 xrmx: patches are in ticket. Aug 25 14:45:48 xrmx: i have no control over git.openmoko.org, and think i can't commit to it. Aug 25 14:46:44 xrmx: all my patches miss git.openmoko.org. two times i asked larsc about inclusion of ar6000.patch, but bot times got silence as answer. Aug 25 14:47:31 xrmx: but about ts things are more precise - larsc beleive i should send it upstream Aug 25 14:47:51 gena2x, what about sending it upstream? :) Aug 25 14:48:01 what's wrong with touchscreen driver? Aug 25 14:48:17 it works fine for me :) (2.6.35-2.6.36) Aug 25 14:48:19 xrmx: i won't as i think upstream driver is not appropriate for openmoko Aug 25 14:48:38 anarsoul: on gta02? Aug 25 14:48:47 on rx1950, but it uses same driver Aug 25 14:48:51 s3c2410_ts Aug 25 14:50:10 anarsoul: it's ok - you have no buzz, this is main thing, and second thing is ignoreunexpectedintr29. i didn't investigate reason, but sometimes system seem goes to some unusual state, and ts driver recieves confusing interrupts. Aug 25 14:50:47 anarsoul: this confusing interrupts lead to counter in ts driver underflow and complete hang of ts driver. Aug 25 14:51:26 anarsoul: i think this is really due to flaw in driver, but it happens only in some extreme conditions. Aug 25 14:51:29 I'm pretty sure that 23c239be223d7427da51656d6f196bde965d9796 from linus' tree should fix it Aug 25 14:52:00 let me see... Aug 25 14:52:53 http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=23c239be223d7427da51656d6f196bde965d9796 Aug 25 14:53:16 being out of tree is so painful Aug 25 14:53:35 anarsoul: hehe Aug 25 14:54:10 anarsoul: no, not related Aug 25 14:55:25 gena2x: without this patch touchscreen behavior was really weird, driver started events spam on single touch :) Aug 25 14:55:53 anarsoul, do you have a public git tree with your patches? Aug 25 14:56:17 xrmx: nope, but almost all of them are somewhere in maillist archives Aug 25 14:57:15 hmmm Aug 25 14:57:20 but there's only few s3c24xx-generic: now it's s3c24xx dma fix and generic_time (resumed to work on it today) Aug 25 14:57:22 why not using plain gypsy? Aug 25 14:57:28 anarsoul, i was trying to avoid collecting patches from mailing lists :) Aug 25 14:57:39 anarsoul: what?? ah, this is some old patch Aug 25 14:57:50 anarsoul: it is already included to .34. Aug 25 14:57:57 oh, I see Aug 25 14:58:11 anarsoul: so, no relation. Aug 25 14:58:25 anarsoul: this driver has other multiple issues imo. Aug 25 14:59:09 xrmx: so, in short, as soon as you'll get first ts hang you'll be sure that my patch is needed. Aug 25 14:59:38 btw, why you don't report this issue upstream? :) Aug 25 14:59:57 gena2x, i trust you, i'm trying to build a deb from 2.6.34 kernel and was looking inside debian/patches Aug 25 15:01:16 xrmx: this require significant effort, like producing better logs. also, i really don't know that is real cause of interrupts. only guesses. Aug 25 15:01:46 xrmx: also, i think that filtering driver should be used for gta02. Aug 25 15:02:00 xrmx: so... all is not so clean. Aug 25 15:02:30 gena2x, this was a response to anarsoul upstream question i suppose Aug 25 15:02:33 xrmx: also, i have hacky patch to significantly reduce ts buzz without filtering. Aug 25 15:03:24 with this too, usage of upstream ts is possible, but still far from solved question... Aug 25 15:05:53 hm, is your hacky patch available somewhere? Aug 25 15:06:26 noone so far were brave enough to be interested in it, so no. Aug 25 15:06:43 publish it :) Aug 25 15:07:06 it's too hacky to publish. Aug 25 15:07:14 :) Aug 25 15:07:31 but how it works? Aug 25 15:07:50 now there's only averaging to reduce jitter Aug 25 15:08:36 reduce amount of ts measurements... to 1 :). reduce rate of event to 50hz. slow down glamo pixclock while doing adc measurements. no magic :) Aug 25 15:09:13 and it has much less buzz that default. Aug 25 15:09:36 for example, i cat get reliably get right click from tslib Aug 25 15:09:39 *can Aug 25 15:09:56 still buzzs, but much less. Aug 25 15:10:51 actually, I have almost no buzz on rx1950, but I'm using average value of 32 samples Aug 25 15:11:15 heh, averaging will not help on gta Aug 25 15:11:35 because default buzz level is too high Aug 25 15:11:37 some hardware fault? Aug 25 15:11:48 i tried 16, it didn't help Aug 25 15:12:03 as it is related to glamo clock, do guesses :) Aug 25 15:12:23 but s3c ts hardware itself is fault imo. Aug 25 15:12:37 for finger-friendly input as a minimun. Aug 25 15:12:49 it works fine for h1940 and rx1950 :) Aug 25 15:12:54 it has very tricky race. Aug 25 15:13:08 may be this also depends on touchscreen Aug 25 15:13:20 but i found other problem in it Aug 25 15:14:38 if you press with 'not enought' pressure, it will still adc your data, but do not notice that fact. as a result, you'll get value from almost completely other place Aug 25 15:15:16 this happens only if you press with small pressure and only with finger. Aug 25 15:16:15 so my, guess is that actually ts reports that finger if up/down fast, but s3c do not notice this. Aug 25 15:17:01 and continue measurement, which lead to other kind of data. resulting pointer is always located in 'diagonal' to finger position. Aug 25 15:17:16 but it also has other problem! :) Aug 25 15:17:34 which may cause of any kind of races imo Aug 25 15:18:06 it has 2 registers, each hold x and y position _and_ pen up/down state bit. Aug 25 15:18:52 _but_, seems that pen up/down state is not from measurement time, but instead _current_ state. Aug 25 15:19:34 so, this lead to tricky things. Aug 25 15:19:54 well, I doubt I can reproduce your bugs on hardware I own :) Aug 25 15:19:55 as driver has no way to get pen state while measurement. Aug 25 15:20:12 it works fine. Aug 25 15:20:22 until some corner thing happens. Aug 25 15:20:36 but hm. Aug 25 15:20:46 can you press with small pressure? Aug 25 15:20:49 your ts? Aug 25 15:20:58 yep Aug 25 15:21:10 will you pointer be always at center of your finger? Aug 25 15:21:22 open some drawable (like xpaint) Aug 25 15:21:25 and check. Aug 25 15:21:28 or ts events. Aug 25 15:21:41 nope, it will jitter Aug 25 15:21:52 jitter is other thing Aug 25 15:21:53 but not too much Aug 25 15:22:08 but will it be in complete other position? Aug 25 15:22:16 nope Aug 25 15:22:32 may be this is because of your averaging of 32. Aug 25 15:22:42 try to see with averaging 1 Aug 25 15:23:06 with averaging 1 driver is pretty unusable :) Aug 25 15:23:30 it starts to work reliable from 8 Aug 25 15:23:31 main sign is that it tends to stay in place, diagonal to right location. Aug 25 15:23:50 so, may be you just hiding issue. Aug 25 15:24:31 of course, i may be wrong also. if someone tricky (or wrong memory accesing) resets my register on the way. Aug 25 15:24:44 upstream driver all based on timeouts. Aug 25 15:24:48 like this: Aug 25 15:25:15 Aug 25 15:25:40 adc_callback: Aug 25 15:26:14 timer_happens: Aug 25 15:26:51 yep, I know how it works :) Aug 25 15:27:01 notice , which is separate from Aug 25 15:27:52 gena2x: yep, it doesn't cancel adc scheduled conversions, but I see no race condition Aug 25 15:28:00 heh Aug 25 15:28:17 let me tell... Aug 25 15:28:55 ok. first thing as you see Aug 25 15:29:12 it introduces delay of HZ/2 on average measurement Aug 25 15:30:51 but let me describe race, it is tricky.... Aug 25 15:31:04 irq comes -> s3c_adc_start -> select(1) -> disable ts irq -> ts_conversion(n times) -> select(0) -> mod_timer -> emit event if pen is down -> enable ts irq Aug 25 15:31:58 nono Aug 25 15:32:04 no 'disable ts irq' Aug 25 15:32:13 :) Aug 25 15:32:18 <[Rui]> freerunner's touch screen driver supports right click emulation? (long press?) Aug 25 15:32:26 look through s3c24xx_ts_select function in s3c2410_ts.c Aug 25 15:32:33 [Rui]: tslib supports Aug 25 15:32:37 writel(S3C2410_ADCTSC_PULL_UP_DISABLE | AUTOPST, Aug 25 15:32:38 ts.io + S3C2410_ADCTSC); Aug 25 15:32:46 it disables ts irqs Aug 25 15:32:59 wait, wait... Aug 25 15:33:09 <[Rui]> gena2x: it's just because of that L|R button that's occupying some space (I managed to take it out, revealing another problem) Aug 25 15:33:27 <[Rui]> gena2x: it's not enabled by default in current images, right? Aug 25 15:33:56 [Rui]: wait a bit, i'll finish with ts, and tell all i now about RMB Aug 25 15:34:20 <[Rui]> gena2x: minutes or days? :) Aug 25 15:34:30 hehe Aug 25 15:34:37 minutes... i hope Aug 25 15:34:52 anarsoul: partially right Aug 25 15:35:14 <[Rui]> gena2x: well, you could be doing some work which would take a few days, that's "a bit" as well if it's big work :) Aug 25 15:35:50 anarsoul: but right sequence: irq comes -> s3c_adc_start -> select(1) -> disable ts irq -> ts_conversion(n times) -> select(0) -> enable irq -> mod_timer -> emit event if pen is down Aug 25 15:36:00 stop Aug 25 15:36:08 wait a bit... Aug 25 15:36:20 yep Aug 25 15:36:24 thing is Aug 25 15:36:59 that ADCUPDN reg value is invalid when TS mode is not WAIT4INT Aug 25 15:37:21 anarsoul: exactly. Aug 25 15:37:49 that's why timer's required :) but where's race? Aug 25 15:37:52 but not in this words Aug 25 15:37:53 <[Rui]> how does one add gadgets to the top bar? and to the popup bar? Aug 25 15:38:17 anarsoul: you can still do adc, while timer is happening Aug 25 15:38:26 mm Aug 25 15:38:32 no, you can't Aug 25 15:38:38 why? Aug 25 15:38:50 select callback is called when adc conversions are finished Aug 25 15:39:08 I mean select callback with 0 as arg, i.e. deselect Aug 25 15:39:13 select callback is called on _each_ adc. Aug 25 15:39:32 but timer armed in Aug 25 15:39:48 so it not depend on select at all Aug 25 15:40:49 well, with 0 as arg its called only when conversion is finished Aug 25 15:41:00 s3c_adc_irq in plat-samsung/adc.c Aug 25 15:41:05 yes. Aug 25 15:41:11 you didn't got. Aug 25 15:41:18 we have 2 ints Aug 25 15:41:22 1-conversion Aug 25 15:41:28 2-pen up/down Aug 25 15:41:40 in pen_down we arm timer to jiffies+1 Aug 25 15:41:43 and start adc. Aug 25 15:41:57 so timer may happen sooner than adc will be finished Aug 25 15:42:59 and if we'll strictly interpret manual as ' ADCUPDN reg value is invalid when TS mode is not WAIT4INT' Aug 25 15:43:10 we'll get invalid value. Aug 25 15:43:30 maybe something was changed in 2.6.35, because timer is armed only _after_ all conversions finished, in s3c24xx_ts_select Aug 25 15:43:36 there's only one mod_timer in driver Aug 25 15:43:42 and it's in s3c24xx_ts_select Aug 25 15:44:07 let me see code (i am watching .34) Aug 25 15:44:08 and TS irq can't come when ADC is doing conversion Aug 25 15:45:18 yes Aug 25 15:45:28 TS can't. Aug 25 15:45:43 but i am not about it Aug 25 15:46:00 btw it can came after adc, but before timer Aug 25 15:46:18 but let me check .35 code first Aug 25 15:46:27 and it will be UP irq, so what? :) Aug 25 15:46:40 may be immediately follow down irq Aug 25 15:46:49 wait a bit Aug 25 15:46:56 may be i am wrong in my last sentence Aug 25 15:47:25 nope, it can't be DOWN, as DOWN irq is masked until timer handler is called Aug 25 15:48:34 ah Aug 25 15:48:59 i recalled one more thing. Aug 25 15:49:01 anyone knows why a just flashed image (latest qi, latest uboot and latest image file available for unstable) would not allow GPS to function? with the GPS enabled, I only get GPSD,O=? GPSD,Y=? when running gpspipe -r 127.0.0.1 2947 Aug 25 15:51:42 gena2x: I'm leaving for now, can continue discussion in ~1 hour :) Aug 25 15:52:03 don't forget to publish your memory test Aug 25 15:52:03 anarsoul: it will be very interesting for me to finish it. Aug 25 15:52:10 i already did it Aug 25 15:52:35 [Rui]: ok, Aug 25 15:52:38 tslib Aug 25 15:53:00 yes, rmb works with tslib, but require hack Aug 25 15:53:08 in tslib or in kernel Aug 25 15:53:13 both one-liners Aug 25 15:53:21 rebelThor, shr-u? Aug 25 15:53:57 rebelThor: go to street and wait 5 mins Aug 25 15:54:18 rebelThor: so far this worked for everyone who had problems with gps in shr Aug 25 15:54:33 [Rui]: still interested? Aug 25 15:54:55 <[Rui]> gena2x: yeah, sure :) Aug 25 15:55:09 [Rui]: so, which way you prefer? Aug 25 15:55:13 GNUtoo|laptop: yeah, SHR-U, http://build.shr-project.org/shr-unstable/images/om-gta02/full-om-gta02.jffs2 just fetched 20 minutes ago Aug 25 15:55:14 * [Rui] was coding support to viewing related statuses in elmdentica (ie, those one replied to). Aug 25 15:55:37 gena2x: shouldn't I get something like $GPGGA,235946.99,,,,,00,00,5.0,,M,0.0001999,M,0.0020199,*57 even though I'm inside the room? Aug 25 15:55:42 i mean still interested after word 'hack' :) Aug 25 15:55:50 <[Rui]> gena2x: a way to have right mouse click without having a button wasting top bar space :) Aug 25 15:55:52 rebelThor: go to street :) Aug 25 15:56:24 <[Rui]> gena2x: I remember having r-m-click emulation when I had a smartq7 (before it broke), so it should be possible Aug 25 15:56:37 [Rui]: ok, you'll need tslib Aug 25 15:56:46 (do shr have one) Aug 25 15:56:51 ? Aug 25 15:56:54 <[Rui]> gena2x: I think shr uses it Aug 25 15:56:55 <[Rui]> no? Aug 25 15:56:59 i think no Aug 25 15:57:03 rebelThor, freerunner I bet Aug 25 15:57:08 <[Rui]> we have an /etc/ts.conf Aug 25 15:57:16 GNUtoo|laptop: yep, FR Aug 25 15:57:23 rebelThor, I don't know....but I can tell you how GPS works so you can debug it Aug 25 15:57:28 <[Rui]> it's installed Aug 25 15:57:42 basically fsotdlt should have gps provider disabled Aug 25 15:57:51 so frameworkd register its ogpsd Aug 25 15:58:00 that's plain gypsy Aug 25 15:58:02 rebelThor: really, try in open space. i know 5 people who return with 'thanks' after this suggestion. i am anong this people. Aug 25 15:58:05 I mean gypsy interface Aug 25 15:58:09 then.... Aug 25 15:58:16 fso-gpsd uses that interface Aug 25 15:58:21 and translate to gpsd format Aug 25 15:58:26 which is a tcp daemon Aug 25 15:58:34 gypsy is dbus Aug 25 15:58:54 [Rui]: check xorg.conf Aug 25 15:58:55 and gpsd is tcp Aug 25 15:59:04 <[Rui]> gena2x: it's using evdev :| Aug 25 15:59:16 [Rui]: it should be Driver "tslib" Aug 25 15:59:16 Option "Device" "/dev/input/event1" Aug 25 15:59:16 Option "EmulateRightButton" "1" Aug 25 15:59:26 for tslib && button emulation Aug 25 15:59:30 [Rui], you had a smartQ? Aug 25 15:59:54 [Rui], with xf86-input-tslib right click emulation is possible Aug 25 15:59:55 <[Rui]> GNUtoo|laptop: more like in I had. I let it fall down and the touchscreen film borked completely (otherwise looks fine). Aug 25 16:00:24 did you have time to hack on it? Aug 25 16:00:28 what sources are lacking? Aug 25 16:00:35 hm... Aug 25 16:01:00 you need to recompile one of kernel ts driver or tslib. Aug 25 16:01:16 if where is not tslib on shr, you'll have to recompile it anyway. Aug 25 16:01:23 get tslib sources Aug 25 16:01:24 ah you can choose? Aug 25 16:01:33 I thought it needed a little modification Aug 25 16:01:37 GNUtoo|laptop: yeah :) Aug 25 16:01:55 like BTN_TOUCH -> TOUCH_SOMETHINGELSE Aug 25 16:02:11 because else it doesn't register a click Aug 25 16:02:16 but you can still move arround Aug 25 16:03:56 [Rui] here i hack i am using: http://www.bsdmn.com/openmoko/tslib/tslib_hack.diff Aug 25 16:03:59 GNUtoo|laptop: hey I am sorry I didn#t reply yesterday regarding htcleo GPS Aug 25 16:04:05 s/i hack/is hack/ Aug 25 16:04:05 gena2x meant: [Rui] here is hack i am using: http://www.bsdmn.com/openmoko/tslib/tslib_hack.diff Aug 25 16:04:18 GNUtoo|laptop: I was falling asleep and I was on ubuntu on the hd2 with on screen keyboard :> Aug 25 16:04:22 ok Aug 25 16:04:27 np Aug 25 16:04:36 so you may want to respond now Aug 25 16:04:56 GNUtoo|laptop: i can't exactly recall details of issue, i applied this patch half year ago :) Aug 25 16:05:12 ok Aug 25 16:05:15 I recall Aug 25 16:05:25 I think it was Aug 25 16:05:26 ABS_PRESSURE Aug 25 16:05:33 for tslib Aug 25 16:05:41 and BTN_TOUCH for evdev Aug 25 16:05:45 not sure tough Aug 25 16:05:49 [Rui] so, basically after that your tslib should work. if not, kick me to redo diff. Aug 25 16:06:44 i just added too many debugging prints, and may be missed something while removing. i'll recheck it if it'll not work. Aug 25 16:15:58 <[Rui]> back Aug 25 16:16:02 <[Rui]> I have to report that using tslib worked better than the L|R button! Aug 25 16:17:50 <[Rui]> but it seems to enter into a bit of conflict with Elementary's popup menus Aug 25 16:18:14 hm... how to talk with Rui??? Aug 25 16:18:27 :) Aug 25 16:32:02 gena2x: I'm back Aug 25 16:36:55 anarsoul: ok, i'm here too. Aug 25 16:37:55 anarsoul: so, timer arming.... Aug 25 16:39:43 mrmoku|a`, FINALLY! (thanks) Aug 25 16:40:28 JaMa|GoNe, pushed the changes. Aug 25 16:40:33 JaMa|GoNe, to the gadgets. Aug 25 16:42:56 xrmx: just retested generic_time patch, unfortunatelly my PDA can't resume from suspend with it Aug 25 16:43:39 hi, Aug 25 16:43:41 about GPS: Aug 25 16:43:53 anarsoul, that would have been too good to be true :) Aug 25 16:43:54 did someone test openBmap and cellhunter? Aug 25 16:43:58 anarsoul: yes, you right Aug 25 16:44:05 cellhunger seem to work with gpsd Aug 25 16:44:09 openbmap no Aug 25 16:44:22 but openbmap has not only an app in oe that collect data Aug 25 16:44:27 but it has an app for using it too Aug 25 16:44:38 but I bet it uses gypsy too Aug 25 16:44:46 I tried gypsy daemon Aug 25 16:44:47 xrmx: maybe I'll spend sometime to fix it, but I don't promise that it will be soon :) Aug 25 16:44:49 it didn't work for me Aug 25 16:44:57 by gypsy I mean plain daemon Aug 25 16:44:58 anarsoul: timer armed at deselect. this is my jitter patch (i am using jiffie + 3) Aug 25 16:45:28 anarsoul: sorry for confusion Aug 25 16:46:05 gena2x: it's ok :) what about your memory test and testing my dma-related patches? :) Aug 25 16:46:14 anarsoul: wait a bit Aug 25 16:46:24 anarsoul: lets finish up with ts first! Aug 25 16:46:24 anarsoul, if you fix it push it upstream or write a message to openmoko-kernel :) Aug 25 16:46:28 anarsoul: don't mind? Aug 25 16:46:40 gena2x: ok Aug 25 16:46:54 xrmx: sure Aug 25 16:46:55 hey guys, anyone with knowledge on where to look at: bluetoothd is not run on bluetooth power up, and bluez can't find default adapter. maybe something's borked in kernel? Aug 25 16:47:21 anarsoul: ok. now i wathing to virgin .35 driver to avoid any further confusion Aug 25 16:48:11 anarsoul: next thing. Aug 25 16:49:31 mickey|zzZZzz: I am reading FSO doc api, fsoraw and org.freesmartphone.Usage it is the same thing right ? Aug 25 16:50:41 mickey|zzZZzz: I read your discution with GNUtoo|laptop og this afternoon, did those change around fso dealing with the GPS could ad something todo with the dropdown of accurency of the GPS in the Freerunner ? Aug 25 16:51:44 mickey|zzZZzz: what does fsotdlt stand for ? Aug 25 16:52:56 anarsoul: ok. lets imagine that up interrupt happens Aug 25 16:53:04 ok Aug 25 16:53:08 anarsoul: but get_down return down state. Aug 25 16:53:11 mickey|zzZZzz: and at last , I am planning to dev an app using the GPS throught FSO, what do I have to conerned by ? Aug 25 16:53:36 anarsoul: now, all this happens while touch_timer_fire is active Aug 25 16:53:37 gena2x: but that's pretty impossible Aug 25 16:54:03 anarsoul: unfortunately, my patch fix exatly this situation. Aug 25 16:54:13 it can happen only if there's a big irq latency Aug 25 16:54:31 this is called in other words 'race'. Aug 25 16:54:58 and no Aug 25 16:55:19 this may happen without any latency Aug 25 16:55:34 gena2x: so, just adapt your patch for 2.6.36 and send it to linux-input maillist Aug 25 16:56:28 ok. Aug 25 16:56:38 you also noticed other big problem by the way Aug 25 16:56:55 you told interrupts disabled while adc? Aug 25 16:57:02 masked Aug 25 16:57:19 just look through s3c2442 datasheet Aug 25 16:57:27 ok. so, how while adc it will know if current pen state is down or up? Aug 25 16:57:37 there's conversion mode and wait for interrupt mode Aug 25 16:57:47 gena2x: it should check after conversion is finished Aug 25 16:58:18 but it is possible that pen will be up during conversions and become down just after them. Aug 25 16:58:30 and btw, not so clean with interrupts. Aug 25 16:58:41 in fact i GOT interrupts while adc. Aug 25 16:58:50 gena2x: adc is fast enough Aug 25 16:58:58 32 of adc? Aug 25 16:59:12 with delay? Aug 25 16:59:28 i have logs of such interrupts - this is reproducible. Aug 25 16:59:40 gena2x: one conversion is 5 us (max) Aug 25 16:59:50 and i GOT such unsync things (have logs too) Aug 25 17:00:03 160 uS for 32 conversions Aug 25 17:00:14 and that's worst case Aug 25 17:00:20 did you took in account delay register? Aug 25 17:00:31 or only conversion time register? Aug 25 17:02:12 gena2x: s3c2442 datasheet, page 16-3 Aug 25 17:02:36 ok, it is in from of me. Aug 25 17:02:43 hehe Aug 25 17:02:57 i has 2 registers Aug 25 17:03:21 p 16-7 ADCDLY Aug 25 17:04:14 and p ADCON (PRSCVL) Aug 25 17:04:33 PRSCVL is conversion time as i understand Aug 25 17:05:08 while DELAY is delay before each adc Aug 25 17:05:18 may be not each but serie Aug 25 17:05:19 gena2x: I got your point; just send your patch with appropriate description to linux-input Aug 25 17:05:20 :) Aug 25 17:05:52 eh. Aug 25 17:06:04 do i understand correctly or i am not right? :) Aug 25 17:07:38 there's no much info about that reg in datasheet, and I'm not sure what's its purpose Aug 25 17:08:14 this is why this always reqiere more time... Aug 25 17:08:36 ok. anyway. i am not planning to submit it right now :) Aug 25 17:08:46 let's got to memory timing. Aug 25 17:08:54 not timing but speed atm. Aug 25 17:09:47 about ts: as you see, i need stonger understaning and better arguments to submit Aug 25 17:09:55 :) Aug 25 17:10:21 http://www.bsdmn.com/openmoko/sdramtest/ Aug 25 17:10:31 here is makefile and bunck of files Aug 25 17:10:45 i ask you to compile with 'make' for device Aug 25 17:10:49 and run 3 tests Aug 25 17:11:07 out of compiled stuff Aug 25 17:12:20 ./memtest_memcpy Aug 25 17:12:57 ./memtest_read Aug 25 17:13:01 ./memtest_write Aug 25 17:13:09 for fr i have: Aug 25 17:13:25 ./memtest_memcpy - 57Mb/s Aug 25 17:13:45 ./memtest_write - 185Mb/s Aug 25 17:14:01 ./memtest_read - 83Mb/s Aug 25 17:14:11 and this 'read' puzzles me much Aug 25 17:14:47 anarsoul: so, please, can you run this test on something with similar cpu :) Aug 25 17:17:17 ohhh, can you publish it in tar.gz? (yeah, I know about wget -r) Aug 25 17:17:30 sure Aug 25 17:19:39 anarsoul: ahh, this were wrong did anyway. right link: http://www.bsdmn.com/openmoko/memtest/memtest.tgz Aug 25 17:19:50 s/did/link/ Aug 25 17:19:50 gena2x meant: anarsoul: ahh, this were wrong link anyway. right link: http://www.bsdmn.com/openmoko/memtest/memtest.tgz Aug 25 17:24:38 arghhh, std=c++0x Aug 25 17:25:47 hi JaMa|GoNe Aug 25 17:26:01 is it safe to update on dream if I compiled just now? Aug 25 17:27:17 anarsoul: it ware very quick-written. you g++ have no support? :) Aug 25 17:27:24 s/ware/were/ Aug 25 17:27:24 gena2x meant: anarsoul: it were very quick-written. you g++ have no support? :) Aug 25 17:28:25 for uclibc I have no c++ compiler at all, now trying to compile it with another toolchain :) Aug 25 17:31:50 gena2x: memcpyspeed.cc lacks #include (it uses printf for some reason), and makefile was not cross-compile friendly Aug 25 17:32:12 this is really quick-written, sorry. Aug 25 17:32:35 need to remove c++0x, remove int64_t Aug 25 17:32:38 to long long Aug 25 17:32:46 and of course add cstdio Aug 25 17:32:52 int64_t is ok, it's in c99 Aug 25 17:33:00 but not in c++ :) Aug 25 17:34:01 that's why i just added c++0x :) Aug 25 17:36:19 it's divided to different translation units to avoid optimizations for sure Aug 25 17:38:53 gena2x: http://pastebin.ca/1925223 Aug 25 17:40:03 anarsoul: aha. thanks. it looks almost similar to gta02 values. just a bit slower. Aug 25 17:40:32 why Memcpy sanity: fail? Aug 25 17:40:38 it's ok Aug 25 17:40:42 as this is not memcpy Aug 25 17:41:14 but simple read or write. Aug 25 17:41:34 ok. thanks. so now i know that such thing is not fr-specific. Aug 25 17:41:58 same picture - read is 2 times slower than write. Aug 25 17:42:42 there were discussion about that on lakml Aug 25 17:43:23 really??? Aug 25 17:43:30 very interesting! Aug 25 17:43:43 can you point me or tell some keywords? Aug 25 17:44:37 hi dcordes Aug 25 17:44:45 about gps here's how it works: Aug 25 17:44:45 yo Aug 25 17:44:56 gena2x: trying to find it :) Aug 25 17:45:09 first mickel wrote the first fso implementation: Aug 25 17:45:11 frameworkd Aug 25 17:45:19 then he wrote cornucopia in vala Aug 25 17:45:31 frameworkd was written in python Aug 25 17:45:36 well nice Aug 25 17:45:40 I know that Aug 25 17:45:43 so...now: Aug 25 17:45:58 we don't have a complete cornucopia gps Aug 25 17:46:01 why are you making things so userspace speicifc. a gps.c that works on any distro would be nice Aug 25 17:46:06 so it's disabled Aug 25 17:46:09 in fsotdlt.conf Aug 25 17:46:11 let the user turn on gps manually with a shortcut and your done. Aug 25 17:46:16 only old ogpsd from frameworkd Aug 25 17:46:19 no fso*.conf Aug 25 17:46:28 no that doesn't work Aug 25 17:46:32 how to stop it then? Aug 25 17:46:34 why what is the problem about that Aug 25 17:46:38 how to integrate with existing apps? Aug 25 17:46:42 with a stop-gps shortcut Aug 25 17:46:50 hmmm no Aug 25 17:46:54 any gps app let you speciy gps device Aug 25 17:46:56 done Aug 25 17:47:01 not true Aug 25 17:47:04 navit Aug 25 17:47:10 you have to edit xml files Aug 25 17:47:16 so here's the state: Aug 25 17:47:16 and? Aug 25 17:47:16 hm, that's not discuss, just single question Aug 25 17:47:17 http://www.spinics.net/lists/arm-kernel/msg71551.html Aug 25 17:47:24 what's the problem about editing xml file Aug 25 17:47:26 let me explain the state fisrt Aug 25 17:47:30 and how is that any related Aug 25 17:47:32 there are 2 GPS api: Aug 25 17:47:35 dbus: gypsy Aug 25 17:47:40 tcp: gpsd Aug 25 17:47:43 and no response Aug 25 17:47:46 some apps expect dbus Aug 25 17:47:49 some expect gpsd Aug 25 17:48:05 well some openmoko apps might expect dbus Aug 25 17:48:11 indeed Aug 25 17:48:11 don't use them Aug 25 17:48:15 you are going the android way ! Aug 25 17:48:38 stay modular Aug 25 17:48:41 that's all I can say Aug 25 17:48:48 hmmm Aug 25 17:48:52 but the apps are so great Aug 25 17:48:57 there is openBmap Aug 25 17:49:11 openBmap can map cell towers, and use them to get a fix Aug 25 17:49:21 the use them is not in oe Aug 25 17:49:27 s/fix/position Aug 25 17:50:15 don't get me wrong I am not saying that fso is bad Aug 25 17:50:16 <\marco> hi all Aug 25 17:50:24 I just wouldn't have it handle _everything_ Aug 25 17:50:49 <\marco> I've just do an upgrade and, after the efreet problem, I've this: Aug 25 17:51:00 mabye it is just the android trauma sorry :D Aug 25 17:51:01 <\marco> Configuring util-linux-ng-readprofile. Aug 25 17:51:02 <\marco> update-alternatives: Error: cannot register alternative readprofile to /sbin/readprofile since it is already registered to readprofile.util-linux-ng 100 Aug 25 17:51:45 dcordes, gps state in fso isn't great actually Aug 25 17:51:48 GNUtoo|laptop: how is GPS activated on freerunner phones in kernel level? Aug 25 17:51:53 but the issue is not fso Aug 25 17:51:57 it's the 2 apis Aug 25 17:52:02 and the api change of gpsd Aug 25 17:52:09 dcordes, sys node Aug 25 17:52:13 I'll check Aug 25 17:52:26 GNUtoo|laptop: and how is it handled in userspace ? Aug 25 17:52:34 GNUtoo|laptop: same gps.c dance like on dream ? Aug 25 17:52:50 anarsoul: that man has something complete strange Aug 25 17:52:58 I swear if you put a button to send the init and one to send gps terminate magic everybody will be happy Aug 25 17:53:12 and you don't have to worry about adding that code in 25512314³ different programs Aug 25 17:53:22 not at all Aug 25 17:53:25 look here: Aug 25 17:53:40 anarsoul: 130Mb/s vs 1Gb/s look waaay to much Aug 25 17:53:45 /usr/lib/python2.6/site-packages/framework/subsystems/ogpsd Aug 25 17:53:47 look there Aug 25 17:53:52 om.py Aug 25 17:53:56 got a git link Aug 25 17:54:05 DEVICE_POWER_PATH_NEW = "/sys/bus/platform/devices/gta02-pm-gps.0/power_on" Aug 25 17:54:08 yes Aug 25 17:54:13 file:///usr/lib/python2.6/site-packages/framework/subsystems/ogpsd Aug 25 17:54:16 (joke) Aug 25 17:54:19 look in your device Aug 25 17:54:32 ok so you got my idea ? Aug 25 17:54:41 just want to save you work Aug 25 17:54:48 <\marco> forgot to mention: I'm using shr-u. Do you have any idea? Aug 25 17:54:55 dcordes, ? Aug 25 17:55:02 so I should use gpsd? Aug 25 17:55:02 GNUtoo|laptop: the button thing Aug 25 17:55:03 anarsoul: but i had complete different values even on completely old arms or i testes n800. Aug 25 17:55:09 dcordes, that's already done Aug 25 17:55:15 GNUtoo|laptop: leave that up to the program. Aug 25 17:55:20 anarsoul: hm. may be worth posting to lkml too? :) Aug 25 17:55:23 GNUtoo|laptop: if it want to use /dev/smd27 or gpsd. Aug 25 17:55:34 GNUtoo|laptop: you just care how to turn it on or offf Aug 25 17:55:47 basically it works like that Aug 25 17:56:08 I made gps.c dbus aware Aug 25 17:56:10 so .... Aug 25 17:56:15 it registers to fso Aug 25 17:56:16 gena2x: post, if you're so curious :) don't forget to add Ben Dooks to cc Aug 25 17:56:24 then fso can power it down or up Aug 25 17:56:32 then...an application like navit does: Aug 25 17:56:39 fsoraw -R GPS navit Aug 25 17:56:41 anarsoul: this question eated me for several weeks. Aug 25 17:56:42 but not in lkml, but in lakml (linux-arm-kernel) Aug 25 17:56:55 fsoraw goal is to power gps as long as navit is on Aug 25 17:57:17 GNUtoo|laptop: ok cut the dbus, fsoraw, .., .., .. and use my proposed method Aug 25 17:57:17 or as long as navit+tangogps+other programs are on Aug 25 17:57:19 anarsoul: and it still unsolved :) Aug 25 17:57:26 uhhh no Aug 25 17:57:29 because: Aug 25 17:57:32 *it's already done Aug 25 17:57:38 *mickeyl wants that Aug 25 17:57:44 gena2x: maybe it's some hw bizzarity? Aug 25 17:58:07 also there is already such button Aug 25 17:58:08 or gcc bug Aug 25 17:58:09 GNUtoo|laptop: *it is zero portable Aug 25 17:58:10 in shr-settings Aug 25 17:58:14 GNUtoo|laptop: *only makes sense when you use fso Aug 25 17:58:18 dcordes, I gave you code Aug 25 17:58:22 indeed Aug 25 17:58:24 anarsoul: i checked assembler Aug 25 17:58:25 and? Aug 25 17:58:26 GNUtoo|laptop: *soon as fso API changes once again you have the same bullshit again Aug 25 17:58:29 I use only fso Aug 25 17:58:31 no Aug 25 17:58:35 just my opinion Aug 25 17:58:39 anarsoul: i redid read in assembler too. Aug 25 17:58:44 the API I was talking about is the NMEA parsing Aug 25 17:58:48 not the activator Aug 25 17:58:48 GNUtoo|laptop: you are talking like android guy Aug 25 17:58:55 GNUtoo|laptop: 'I only use android' Aug 25 17:58:57 anarsoul: so, not assembler Aug 25 17:59:26 anarsoul: would be much more interesting to see it is possible to fix this :) Aug 25 17:59:45 dcordes, fso can be ported to debian etc... Aug 25 17:59:50 I really doubt that it's fixable :) Aug 25 18:00:01 GNUtoo|laptop: but as I said it is only my thought on it Aug 25 18:00:05 ok Aug 25 18:00:16 but it seem that you like things like universal packages Aug 25 18:00:24 exactly Aug 25 18:00:28 I prefer each distro packaging its stuff Aug 25 18:00:34 like in openembedded Aug 25 18:00:39 one distro is for routers Aug 25 18:01:01 why not keeping fso for what it's supposed to do ? Aug 25 18:01:06 they don't care about Xorg support in mplayer for instance Aug 25 18:01:07 handling the phone portion Aug 25 18:01:10 gena2x: and what about testing my patches? :) Aug 25 18:01:12 pim etc Aug 25 18:01:16 it's not ophono either Aug 25 18:01:27 basically it's autotools Aug 25 18:01:29 I don't see the neede to let fso do *PM *GPS *... Aug 25 18:01:33 it can be packaged for debian Aug 25 18:01:38 ah? Aug 25 18:01:39 anarsoul: seem i can't select simple questions. but so far i fixed everything i thought wrong. from touchscreen buzz (which were impossible) to fr speed (which were impossible too). Aug 25 18:01:43 so where should it be? Aug 25 18:01:46 anarsoul: you mean s3cmci? Aug 25 18:01:48 * lindi- is struggling with an mplayer issue at the moment: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=594093 Aug 25 18:01:50 manually enable gps and forget about it? Aug 25 18:01:55 GNUtoo|laptop: yes Aug 25 18:02:04 GNUtoo|laptop: that's how I would do gps Aug 25 18:02:06 forget about it means forget to shutdown it Aug 25 18:02:11 gena2x: yep, patches 0021 and 0022 from http://downloads.tuxfamily.org/linuxrx1950/patches/bjdooks-next-samsung/25_aug_2010/ Aug 25 18:02:15 GNUtoo|laptop: yes I get that and I see the downside Aug 25 18:02:17 anarsoul: i have only gta02 :( if i can test something - i'll do, just tell me... Aug 25 18:02:32 dcordes, on dream we can do both Aug 25 18:02:35 on freerunner too Aug 25 18:02:40 bypassing fso is possible Aug 25 18:02:45 then should it be the default Aug 25 18:02:47 http://htc-linux.org/wiki/index.php?title=Main_Page Aug 25 18:02:48 definitely no Aug 25 18:02:50 gena2x: apply patch, enable DMA for s3cmci in dma config and test WiFi speed/cpu load during Wifi transfer Aug 25 18:02:52 not on SHR Aug 25 18:02:53 Focus of the HTC Linux project is to make Linux in all of its flavors available to as many devices as possible. Aug 25 18:03:01 see that's my focus Aug 25 18:03:02 yes Aug 25 18:03:04 ok Aug 25 18:03:07 it's a nice focus Aug 25 18:03:12 *in kernel config Aug 25 18:03:13 I agree with that Aug 25 18:03:16 anarsoul: ah, WIFI!? Aug 25 18:03:19 but we don't agree on the how Aug 25 18:03:33 if someone makes a gps lib for android Aug 25 18:03:36 gena2x: yep, it's connected to s3c2442 mci Aug 25 18:03:43 anarsoul: ok. but this will take some time, as i need to bring back online my wifi. Aug 25 18:03:50 you think like: they should do that, we can't use it Aug 25 18:04:01 I think: I'll take it and modify it for SHR Aug 25 18:04:01 GNUtoo|laptop: when I patch phh / your gps.c to make it work for htcleo I primarily want to make it work in any distro Aug 25 18:04:01 anarsoul: i thought it is about gta01 mmc transfer only. Aug 25 18:04:09 ensure that CONFIG_MMC_S3C_DMA is set in kernel config Aug 25 18:04:11 dcordes, you can do that Aug 25 18:04:17 anarsoul: ok, just wait. Aug 25 18:04:20 I don't see why what I do is an issue Aug 25 18:04:26 just do that for you Aug 25 18:04:30 I can even help Aug 25 18:04:30 not CONFIG_MMC_S3C_PIO Aug 25 18:04:35 but I won't do that for me Aug 25 18:04:36 <\marco> damn ... I've tried to reinstall busybox.. Aug 25 18:04:41 I don't see the issue Aug 25 18:04:50 GNUtoo|laptop: then in a second step, if somebody wants to use it in an environemnt which allows for complicated mechanisms (that of course can make life easier once finished) like fso in shr or whatever meego has Aug 25 18:04:54 and btw TV makes thinking hard for me Aug 25 18:04:59 GNUtoo|laptop: then you can make a distro specfici thing Aug 25 18:05:09 <\marco> but it can't re-download itself since there's not a wget -_- Aug 25 18:05:16 dcordes, indeed that's SHR Aug 25 18:05:20 I use SHR Aug 25 18:05:24 so I port to SHR Aug 25 18:05:25 <\marco> ... -_- Aug 25 18:05:32 anarsoul: ok. Aug 25 18:05:33 I'm open to people wanting to port debian Aug 25 18:05:35 or else Aug 25 18:05:41 I can help Aug 25 18:05:45 GNUtoo|laptop: if I had the dream I would sure want to use GPS in debian Aug 25 18:05:47 like help on irc Aug 25 18:05:52 say how I did this/that Aug 25 18:05:56 but I won't do it Aug 25 18:06:05 indeed that would be easy: Aug 25 18:06:17 git clone fso-uttilities Aug 25 18:06:27 cd uttilities/android Aug 25 18:06:30 sh autogen.sh Aug 25 18:06:35 ./configure Aug 25 18:06:36 make Aug 25 18:06:45 and ship it Aug 25 18:06:52 and run: Aug 25 18:06:52 gps Aug 25 18:07:00 and you've data on /dev/smd27 Aug 25 18:07:06 <\marco> umpf.. -_- have to reflash.. good -_ Aug 25 18:07:06 it's possible Aug 25 18:07:10 we made it possible Aug 25 18:07:22 if someone wants to do that we won't stop him and will help him Aug 25 18:07:32 but I won't do it myself Aug 25 18:08:18 on hd2: some things can be made distro-independant Aug 25 18:08:23 but not everything Aug 25 18:09:15 GNUtoo|laptop: yes true Aug 25 18:09:35 gena2x: please let me know about results in private (I'll be afk for some time) Aug 25 18:09:46 dcordes, I helped you and will continue to help but I don't want to undo SHR integration of the htcdream Aug 25 18:09:59 anarsoul: ok, it will take a while. btw, can i apply it to .34? Aug 25 18:10:09 you still have the choice to use "gps" Aug 25 18:10:52 GNUtoo|laptop: of course. and I am not trying to tell you to drop that work or change the way you integrate it Aug 25 18:11:03 gena2x: yep, I think it should work Aug 25 18:11:13 GNUtoo|laptop: after all it's not a bad thing when you have thame to be as user friendly as possible Aug 25 18:11:15 ok, i'll inform. Aug 25 18:11:22 indeed Aug 25 18:11:25 if the 2 are possible Aug 25 18:11:31 and that user friendly is the default Aug 25 18:11:34 it's great no? Aug 25 18:11:45 GNUtoo|laptop: so un your roadmap to total dream user satisfaction there are not many things left right ? Aug 25 18:11:57 *GPS Aug 25 18:12:15 *OPKG GUI Aug 25 18:13:22 dcordes, keyboard,GUI,GPS,alsa-buffer-underrun-free,and wifi PSM Aug 25 18:13:38 I can even view big buck bunny fullscreen Aug 25 18:13:46 under some special conditions Aug 25 18:14:54 GNUtoo|laptop: ok it sounds like a perfect beta condition to me :) Aug 25 18:15:12 and we have official images Aug 25 18:15:22 which is great and will help for wifi Aug 25 18:15:32 because kvalo will help if he'll have time Aug 25 18:15:36 he already booted SHR Aug 25 18:15:48 how do official images relate to wifi ? Aug 25 18:15:49 he'll help only for wifi btw Aug 25 18:15:58 because kvalo doesn't have a lot of time Aug 25 18:16:04 aha ok Aug 25 18:16:04 and he didn't do any work on dream Aug 25 18:16:13 because he didn't have a GNU/Linux image Aug 25 18:16:24 and bob copeland used android for doing his wifi work Aug 25 18:16:31 because of the same reason Aug 25 18:16:42 do you remember me trying to convince leviathan and you to do 'official release' ? Aug 25 18:16:42 and even better Aug 25 18:16:46 you can test without installing Aug 25 18:16:56 no I don't remember...sorry Aug 25 18:17:15 I needed something booting for the release Aug 25 18:17:34 so I didn't do it when you had a lot of manual tweaks just to get a boot to a GUI Aug 25 18:17:38 GNUtoo|laptop: the 'ffocial images' are a perfect base Aug 25 18:17:42 indeed Aug 25 18:17:59 GNUtoo|laptop: what you need to add is the possibility to install it from a windows computer to your htc dream Aug 25 18:18:28 I'll do that at the end, there are 2 more important things: Aug 25 18:18:34 *finish my urgent todo list Aug 25 18:18:39 GNUtoo|laptop: and that is what we were dicussing about Aug 25 18:18:39 *make a promotion video Aug 25 18:18:40 GNUtoo|laptop: do it now. Aug 25 18:18:48 the promotion video is important Aug 25 18:18:59 yep Aug 25 18:19:03 I've no idea on how to install from windows Aug 25 18:19:05 but.... Aug 25 18:19:06 iirc I also proposed that Aug 25 18:19:16 I already started for the vide Aug 25 18:19:21 I tried: Aug 25 18:19:23 mplayer //fails Aug 25 18:19:28 gst-launch //fail Aug 25 18:19:36 cheese //success Aug 25 18:19:40 I suggested to: *make it installable from windows computer *create noob friendly step by step documentation *shoot small video Aug 25 18:20:29 hmmm: Aug 25 18:20:30 *create documentation Aug 25 18:20:30 *shoot video Aug 25 18:20:30 *make leviathan do the noob firendly installer Aug 25 18:20:42 GNUtoo|laptop: I can help you with the installer as well Aug 25 18:20:52 GNUtoo|laptop: if you point me to basic documentation and image Aug 25 18:20:56 GNUtoo|laptop: I will create initrd for you Aug 25 18:21:04 ohh so nice!!!! Aug 25 18:21:12 I'll help you on gps first tough Aug 25 18:21:38 how do you write the kernel commandline like initrd= ? Aug 25 18:21:47 no wait Aug 25 18:21:55 we must pass that to the bootloader Aug 25 18:22:02 basically I have no clue how the bootloader works Aug 25 18:22:07 in any android device Aug 25 18:24:34 GNUtoo|laptop: what we do on wince devices: 1) to the bootlaoder we pass: zImage, initrd, cmdline 2) kernel will mount initrd exectue init 3) init mounts the sd card (can be unpartitioned card or first vfat parittion), init loop mounts rootfs.ext2 from sdcard, init switches to rootfs Aug 25 18:24:39 tada Aug 25 18:24:53 so you basically give people the kernel, initrd, rootfs.ext2 Aug 25 18:25:10 make them put it on SD , configure the bootloader and your set :) Aug 25 18:25:26 the best thing about it is they will be able to update things easily Aug 25 18:25:48 'configure the bootloader' part is the only thing you need to help me with Aug 25 18:25:50 I know Aug 25 18:25:56 it's easy Aug 25 18:26:00 if you have such a mechanism working, you can make oe put everything in a .zip Aug 25 18:26:03 done Aug 25 18:26:07 people will love it ! Aug 25 18:26:10 fastboot -c or mkbootimg --cmdline Aug 25 18:26:15 ok something like that Aug 25 18:26:25 awesome Aug 25 18:26:37 so you send kernel and cmdline in a package ? Aug 25 18:26:46 in the kenrel package Aug 25 18:26:48 what about the initrd ? how can the android bootloader get it ? Aug 25 18:26:53 s/package/recipe Aug 25 18:26:53 ok Aug 25 18:27:00 ok Aug 25 18:27:01 just fastboot it Aug 25 18:27:06 I'll explain: Aug 25 18:27:09 yes please Aug 25 18:28:02 http://pastebin.com/QwAqQgXE Aug 25 18:28:10 you do a ramdisk Aug 25 18:28:13 cpio.gz Aug 25 18:28:17 and you do mkbootimg Aug 25 18:28:23 on the ramdisk and the kernel Aug 25 18:28:24 and it's ok Aug 25 18:28:46 so kernel initramfs cmdline Aug 25 18:28:51 all is in one file ? Aug 25 18:29:43 yes Aug 25 18:29:47 perfect Aug 25 18:29:50 it's easy Aug 25 18:29:54 now to the flashing Aug 25 18:30:05 fastboot boot ../your_image.fastboot Aug 25 18:30:09 but... Aug 25 18:30:13 1) I can't test Aug 25 18:30:13 we can't expect users to get a shell in the bootloader and run fastboot ?? Aug 25 18:30:26 nand is somewhat broken Aug 25 18:30:29 maybe I can test on something else than system Aug 25 18:30:31 like cache Aug 25 18:30:38 wait a second Aug 25 18:30:42 2)we have only yaffs2 Aug 25 18:30:50 which result in a very small rootfs Aug 25 18:30:55 is there no mechanism to hold down a button to boot image.fastboot from SD ? Aug 25 18:31:06 I don't know Aug 25 18:31:09 lol Aug 25 18:31:12 but fastboot is available on windows Aug 25 18:31:16 aha Aug 25 18:31:18 I think it's possible Aug 25 18:31:23 but I don't know how to do it Aug 25 18:31:29 anyway.... Aug 25 18:31:32 so the usual htc dream windows geek has the ability to fastboot Aug 25 18:31:34 you really want to install to sd Aug 25 18:31:37 so we are all set ! Aug 25 18:31:38 else you have 60M Aug 25 18:31:42 or something like that Aug 25 18:31:46 because it's partitioned Aug 25 18:31:54 yes he has fastboot Aug 25 18:32:19 ok so where do you see all the problems Aug 25 18:32:20 the problem is how to install to sd? Aug 25 18:32:23 you can't test ? why ? Aug 25 18:32:31 broken "system" nand Aug 25 18:32:38 but I can test on "cache" Aug 25 18:32:42 or something else Aug 25 18:32:53 you don't need nand to fastboot right ? Aug 25 18:33:00 no Aug 25 18:33:03 ok Aug 25 18:33:05 you can boot on whatever Aug 25 18:33:07 nand,sd Aug 25 18:33:19 you only need to execute our new custom kernel+initramfs+cmdline.fastboot Aug 25 18:33:19 basically the cool stuff would be: Aug 25 18:33:22 install on sd Aug 25 18:33:24 then.... Aug 25 18:33:31 fastboot the kenrel Aug 25 18:33:37 and you can try without installing anything Aug 25 18:33:38 ok first problem solved. you can boot. Aug 25 18:33:40 and test Aug 25 18:33:42 indeed Aug 25 18:33:46 hehe Aug 25 18:33:49 you finally got my point :) Aug 25 18:33:56 sorry I keep complicating things Aug 25 18:34:00 see the 1) 2) 3) above Aug 25 18:34:02 but how....to install to sd? Aug 25 18:34:03 it cuts the need of any flashing Aug 25 18:34:06 that is what I want to do. Aug 25 18:34:12 because : Aug 25 18:34:22 1)boot from cpio.gz Aug 25 18:34:33 2)download the rootfs Aug 25 18:34:37 downlioad it ? Aug 25 18:34:48 3) format/unpack/install the microsd? Aug 25 18:34:53 aha :) Aug 25 18:34:54 download or push Aug 25 18:34:55 ok no problem Aug 25 18:35:00 the issue is: Aug 25 18:35:01 I have to leave in a few for sports Aug 25 18:35:05 all that will be done on ram Aug 25 18:35:12 I will handle that no problem Aug 25 18:35:13 ram is about 100M Aug 25 18:35:16 so.... Aug 25 18:35:20 no excratction Aug 25 18:35:31 on the 100M you'll have to have: Aug 25 18:35:32 we prepare the rootfs filesystem file before hand Aug 25 18:35:34 *the shr image Aug 25 18:35:40 *the sdio.gz Aug 25 18:35:44 we will create like 400M.ext2 Aug 25 18:35:44 and the ram Aug 25 18:35:48 stop Aug 25 18:36:04 an shr-image is about: Aug 25 18:36:10 if we can fastboot custom kernel+initramfs+commandline Aug 25 18:36:15 I do the rest :) Aug 25 18:36:22 ok ? Aug 25 18:36:30 I prepare everything Aug 25 18:36:45 73M for the full Aug 25 18:36:48 the neede commandline plus initrd plus rootfs file Aug 25 18:36:51 ok thanks a lot Aug 25 18:37:00 can I eat? Aug 25 18:37:08 and you need a Aug 25 18:37:15 hm Aug 25 18:37:24 can you put the link for rootfs we will use for testing ? Aug 25 18:37:31 best will be a somehwat known working one Aug 25 18:37:41 http://shr.bearstech.com/shr-unstable/images/htcdream/ Aug 25 18:37:51 http://shr.bearstech.com/shr-unstable/images/htcdream/shr-full-eglibc-ipk--20100823-htcdream.rootfs.tar.gz Aug 25 18:37:56 ok you tested that ? Aug 25 18:38:00 no Aug 25 18:38:04 anyone know what the last shr-u release was that actually installs on the freerunner? Aug 25 18:38:06 you got one you tested ? Aug 25 18:38:06 I'm not sure Aug 25 18:38:10 no Aug 25 18:38:13 I'm on 3g Aug 25 18:38:17 ok no hurry Aug 25 18:38:21 I'll look if I tested something Aug 25 18:38:29 just give me known working link.tar.gz Aug 25 18:38:35 so I can prepare will look it up in scrollback then Aug 25 18:38:46 i'm trying out the 17 august image currently ... Aug 25 18:38:47 ok Aug 25 18:38:50 I'll do but later Aug 25 18:38:57 80M is too big for 3g Aug 25 18:39:06 I prefer sepnding GB for downloading source Aug 25 18:39:41 I'll eat, be back after Aug 25 18:42:10 GNUtoo|laptop: bon appetit I'm also in bbl mode now Aug 25 18:44:18 ok thanks a lot Aug 25 19:58:20 hello all, bit of a panic here. I removed libphone-ui-shr as I've built a local copy. when I try to install my ipk I get an error: Aug 25 19:58:22 Package libefreet1 wants to install file /usr/lib/efreet/efreet_desktop_cache_create Aug 25 19:58:23 But that file is already provided by package * efreet Aug 25 19:58:23 * opkg_install_cmd: Cannot install package libphone-ui-shr. Aug 25 19:59:14 --force-depends don't have any effect. That probably only works for removal. Can anybody give me a quick fix? Aug 25 20:03:15 Arigead: well, not probably correct, but i would do --force-overwrite Aug 25 20:03:54 vanous: thanks a million I've only a weak signal and no man pages for opkg. Aug 25 20:04:07 Arigead: np Aug 25 21:20:31 hi folks Aug 25 21:22:01 my SHR-T on my Freerunner won't boot any more, the filesystem is filled up so badly that it can't extract /etc/dev.tar anymore Aug 25 21:29:12 hi, Aug 25 21:29:15 hi mickey|zzZZzz Aug 25 21:29:19 hmmm Aug 25 21:29:23 too bad Aug 25 21:29:23 I had an idea Aug 25 21:29:27 use gypsy Aug 25 21:29:29 the real one Aug 25 21:30:50 it's in C Aug 25 21:30:56 it might be faster no? Aug 25 21:36:10 a vala/python implementation is too slow? Aug 25 21:36:48 you can code vala quite c-ish Aug 25 21:37:09 GNUtoo|laptop: any idea what's the name of the gypsy package in debian? Aug 25 21:37:29 I don't know Aug 25 21:37:32 I'll look Aug 25 21:37:35 afaik there's no package Aug 25 21:37:38 ah ok Aug 25 21:37:45 I tried plain gypsy Aug 25 21:37:47 I made it start Aug 25 21:37:49 but then... Aug 25 21:37:52 how do I use it? Aug 25 21:37:56 gypsy is quite dead for about 2(?) years Aug 25 21:38:03 not true: Aug 25 21:38:15 http://gypsy.freedesktop.org/releases/ Aug 25 21:38:24 09-Jun-2010 08:29 Aug 25 21:38:32 maybe time to update it in oe Aug 25 21:38:34 GNUtoo|laptop: yeah that's why I was wondering where it is in distros Aug 25 21:38:47 ok Aug 25 21:39:05 oh. missed some releases Aug 25 21:39:21 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=494175 is the ITP Aug 25 21:39:51 "Thu, 10 Jun 2010" : "I think I will upload the package during the week-end if you have no Aug 25 21:39:54 comment. Aug 25 21:39:57 " Aug 25 21:40:07 ok so it's not quite there yet Aug 25 21:41:25 hmm, finally: dpkg-deb: building package `mplayer' in `../mplayer_1.0~rc4~try1.dsfg1-1_armel.deb'. <3 Aug 25 21:42:30 GNUtoo|laptop: hi Aug 25 21:42:42 dcordes, hi Aug 25 21:42:54 GNUtoo|laptop: did you find a minute to test a recent build ? Aug 25 21:43:04 dcordes, I'm still on 3g Aug 25 21:43:31 that means that if your build is very small in size....it's fine Aug 25 21:43:33 else...hmmm Aug 25 21:44:02 well I am talking about the shr rootfs we will show the masses Aug 25 21:51:43 GNUtoo|laptop: will you be home later ? Aug 25 21:52:17 GNUtoo|laptop: it is not a big deal soon as I have the rootfs you want me to use Aug 25 21:52:26 no but I can go to the wifi cafee Aug 25 21:54:08 GNUtoo|laptop: ok well let's do it like so Aug 25 21:54:35 GNUtoo|laptop: if you find some time to get bandwidth you download test and chose the image you want for 'beta release Aug 25 21:54:58 GNUtoo|laptop: then you tell me where I can find it and I will prepare the files Aug 25 21:55:02 ok Aug 25 21:55:10 dcordes, what I'll do will be that: Aug 25 21:55:15 I'll dowload official image Aug 25 21:55:17 and test it Aug 25 21:55:49 GNUtoo|laptop: ok. you might as well consider to create a custom image to include some fancy extra things Aug 25 21:56:31 GNUtoo|laptop: e.g. when I will put an shr image for leo I think I will add some common networking analysy things, scummvm, .. :) Aug 25 21:58:49 ah ok Aug 25 21:59:07 not sure if it's a good idea Aug 25 21:59:08 but... Aug 25 21:59:16 a video may temp people to install things Aug 25 21:59:19 like battle for wesnoth Aug 25 21:59:20 pingus Aug 25 21:59:31 supertux-qvga Aug 25 21:59:45 mokomaze? Aug 25 21:59:50 doesn't work Aug 25 21:59:55 it require a higher resolution Aug 25 21:59:59 GNUtoo|laptop: if you have a chance, please test whether the new version of android-rpc builds fso-msm-gps-activator ok Aug 25 22:00:03 and sensor doesn't work Aug 25 22:00:07 mickey|zzZZzz, it works Aug 25 22:00:09 mickey|zzZZzz, hi Aug 25 22:00:14 mickey|zzZZzz, I've tried plain gypsy Aug 25 22:00:17 ah, good to know. then we can start to improve from her Aug 25 22:00:18 e Aug 25 22:00:19 it loads Aug 25 22:00:28 but I don't know how to use it Aug 25 22:01:01 http://pastebin.com/Mq2X5ExE Aug 25 22:01:06 how do I use that? Aug 25 22:01:25 org.freedesktop.Gypsy.Server.Create /dev/smd27 ? Aug 25 22:01:27 i don't know Aug 25 22:01:29 ah Aug 25 22:01:34 right Aug 25 22:01:39 you didn't implement gypsy Aug 25 22:01:39 like i said, i didn't do ogspd Aug 25 22:01:47 daniel and jan did Aug 25 22:01:50 not even fso-gpsd then Aug 25 22:01:54 ok Aug 25 22:01:54 but both are gone with their own share of private problems Aug 25 22:02:00 hmmm Aug 25 22:02:07 i will do fsotdld Aug 25 22:02:17 it already has the basics of NMEA Aug 25 22:02:18 mickey|zzZZzz, what do you think about modifying plain gypsy? Aug 25 22:02:36 by plain gypsy I mean the official implementation at: Aug 25 22:02:41 gypsy is dead, really Aug 25 22:02:52 better teach gpsd about FSO GPS resources Aug 25 22:02:56 http://gypsy.freedesktop.org/releases/ Aug 25 22:02:58 and then openbmap to use gpsd Aug 25 22:03:01 dead? Aug 25 22:03:06 look at the last release Aug 25 22:03:11 ok Aug 25 22:03:12 noone uses it except nokia Aug 25 22:03:17 it has never reached adoption Aug 25 22:03:21 right nokia Aug 25 22:03:27 and if you look at the DBus API, you know why... Aug 25 22:03:27 that's what I wanted to say Aug 25 22:03:33 lol ok Aug 25 22:03:39 I tried to make it work Aug 25 22:03:45 but when I create /dev/smd27 Aug 25 22:04:06 and then when I try to access it the daemon dies Aug 25 22:04:08 ok Aug 25 22:04:09 so gpsd Aug 25 22:04:13 I'll try to fix the apps Aug 25 22:04:14 then Aug 25 22:04:22 it's a complete braindead dbus API, like with bitfields and stuff. they basically moved gps tcp protocol to dbus, not taking into account any good DBus practices. It's a shame that I didn't veto against using it. Aug 25 22:04:35 ok Aug 25 22:04:37 ouch Aug 25 22:04:58 now we're stuck with supporting an obsolete protocol and an adapter that moves this protocol to gpsd protocol :) a real sucky situation Aug 25 22:05:03 that's why i want to start from scratch Aug 25 22:05:07 with a new location API Aug 25 22:05:19 and gpsd support Aug 25 22:05:27 and gypsy support too? Aug 25 22:05:30 no Aug 25 22:05:34 ok Aug 25 22:05:35 gypsy has no adoption. Aug 25 22:05:38 no one wants it Aug 25 22:05:39 :) Aug 25 22:05:42 openBmap Aug 25 22:05:47 cell locator Aug 25 22:05:49 etc... Aug 25 22:05:54 yes, but just because we used it Aug 25 22:05:56 even shr-settings Aug 25 22:05:57 ok Aug 25 22:06:01 if it wasn't for us, they would not have used it Aug 25 22:06:04 but if we got help for moving them.... Aug 25 22:06:05 so we are to blame :D Aug 25 22:06:28 because the idea of mapping the cell towers + getting back your location with a client Aug 25 22:06:40 that would be great Aug 25 22:06:43 well Aug 25 22:06:45 basically : Aug 25 22:06:48 FSO has it since like... milestone 5 Aug 25 22:06:49 :) Aug 25 22:06:52 *less accuracy: Aug 25 22:06:54 but... Aug 25 22:06:58 *less power consumation Aug 25 22:07:00 *fix in no time Aug 25 22:08:07 * GNUtoo|laptop hesitate to help mickey|zzZZzz on fsotldt Aug 25 22:09:17 GNUtoo|laptop: ok as you have a keyboard I think you will not ask to much typing in opkg install wesnoth e.g. Aug 25 22:09:35 wesnoth reminds of the fosdem. there were those guys working on it Aug 25 22:09:38 indeed, can be done via ssh too Aug 25 22:09:43 indeed....lol Aug 25 22:09:45 in the hacking room Aug 25 22:09:50 yeess Aug 25 22:09:55 I showed them wesnoth Aug 25 22:09:59 it was so slow Aug 25 22:10:02 now it's a lot faster Aug 25 22:10:07 it was also because of: Aug 25 22:10:09 *debug kernel Aug 25 22:10:11 does it require keyboard ? Aug 25 22:10:13 *supertux Aug 25 22:10:19 no keyboard needed Aug 25 22:10:30 cool then I bet it will rock on leo Aug 25 22:10:33 it worked on om-gta02, and maybe still works Aug 25 22:10:39 indeed Aug 25 22:10:56 it require a minimum of: Aug 25 22:11:09 400x300 resolution Aug 25 22:11:09 so we are sticking with an official build.. that's best actually. then you can just refer to the autobuild link in some release readme Aug 25 22:12:42 I need to contact openBmap people btw Aug 25 22:12:46 if I succeed Aug 25 22:12:52 because I would have modified the client Aug 25 22:13:09 and so they want to be notified, for uploads Aug 25 22:13:18 because they want to handle clients issues Aug 25 22:14:10 GNUtoo|laptop: should we agree for a time tomorrow to handle the release ? Aug 25 22:14:30 not sure Aug 25 22:14:35 release of what exactly? Aug 25 22:14:43 well to prepare it Aug 25 22:14:48 ahh ok Aug 25 22:14:56 the windows-ready release Aug 25 22:14:59 yep Aug 25 22:15:09 hmmm, here's the issue: Aug 25 22:15:18 on the day I'm supposed to do some maths Aug 25 22:15:25 after I've some free time....but.... Aug 25 22:15:30 the wifi cafee closes early Aug 25 22:15:38 it's really a pain Aug 25 22:15:46 I can't download stuff Aug 25 22:15:51 I can't test official images Aug 25 22:16:01 I can't opkg update;opkg upgrade the freerunner Aug 25 22:16:02 are there no wep or open APs in sight :> Aug 25 22:16:07 there is one Aug 25 22:16:17 only mails + irc passes trough Aug 25 22:16:20 strange isn't it Aug 25 22:16:29 so I've put a tcp vpn on the mail port Aug 25 22:16:30 can't you tunnel through irc port ? Aug 25 22:16:37 but that server is down Aug 25 22:16:40 ah ok Aug 25 22:16:45 it will be up soon tough Aug 25 22:16:49 but the issue is that one: Aug 25 22:16:57 the server has normal ADSL Aug 25 22:16:59 you know what. will prepare a little tarball for you with everything needed Aug 25 22:17:07 then you can handle it on your own Aug 25 22:17:16 in ADSL there is Asyncronous Aug 25 22:17:19 if I remember well Aug 25 22:17:33 asyncronous means about 50k upload Aug 25 22:17:40 with tcp and not udp tunnel Aug 25 22:17:43 you can imagine the speed Aug 25 22:18:07 GNUtoo|laptop, asymmetric Aug 25 22:18:09 and you have frequent disconnections etc.... Aug 25 22:18:18 xrmx, yes indeed, sorry Aug 25 22:20:20 is 'openmoko shr' naming correct ? Aug 25 22:20:22 or good to use ? Aug 25 22:21:16 or should I rather go an openembedded based phone distro named SHR Aug 25 22:21:34 corporate identity lol Aug 25 22:22:08 Stable Hybrid Release (of what? openmoko?) Aug 25 22:23:05 :> Aug 25 22:23:19 please let the openmoko term die for anything but hardware and kernel Aug 25 22:23:37 mickey|zzZZzz: that's my thought Aug 25 22:23:45 SHR has nothing to do with openmoko, except that it also runs there Aug 25 22:23:57 mickey|zzZZzz: as to most people who have heard of it, but are not really involved it is just that. dead / a myth Aug 25 22:24:00 just use SHR as brand, no need to communicate what is stands for Aug 25 22:24:10 ok you're right Aug 25 22:26:54 GNUtoo|laptop: about how big is the offical dream shr extracted ? Aug 25 22:27:18 it's a tar.gz Aug 25 22:27:21 so it can be smaller Aug 25 22:27:38 73M Aug 25 22:37:18 GNUtoo|laptop: I am writing a guide in wiki Aug 25 22:37:43 nice Aug 25 22:37:46 SHR wiki? Aug 25 22:38:22 because I wrote an installation howto in SHR wiki Aug 25 22:38:23 I will put it on htc-linux but you can of course copy it Aug 25 22:38:31 this is general Aug 25 23:22:17 ok **** ENDING LOGGING AT Thu Aug 26 02:59:57 2010