**** BEGIN LOGGING AT Tue Jan 05 02:59:56 2010 Jan 05 03:01:36 max_posedon: so ubi is using 2k blocks, is that what gta02 nand supports? Jan 05 03:01:49 also, is oob custom? Jan 05 03:02:15 tmzt, I think I use 512b on sheevaplug Jan 05 03:02:21 but its doesn't work on neo Jan 05 03:02:32 so, its hardware limitation I think to use 2j Jan 05 03:02:33 2k Jan 05 03:03:30 and if ubifs works, next improvment step will be use squashfs) Jan 05 03:03:48 yes Jan 05 03:03:59 2k is what we have to use on msm Jan 05 03:04:13 I think somebody with vogue can try it and see if it boots faster Jan 05 03:04:30 the total debug settings and patches for 2k? Jan 05 03:04:45 was the underlying mtd driver set up for it already? Jan 05 03:04:50 (2k oob) Jan 05 03:07:31 max_posedon: It just toggles few lines in kernel - reread all written data, and if it fails.... return EIO :) Mb we can disable this one also in distibutions? Jan 05 03:07:56 gena2x, don't ask me, ask others Jan 05 03:08:14 but I'm not sure that its so bad, this is how bad blocks detecting works Jan 05 03:08:28 also, I'm not sure that ubi works on top of mtd layer Jan 05 03:08:37 not directly on nand driver Jan 05 03:08:42 reread written data? Jan 05 03:08:53 mtd is nand, mtdblock is the block layer Jan 05 03:09:05 ubi is unsorted block interface or something Jan 05 03:09:20 max_posedon: just in case you working in that area. Jan 05 03:09:36 max_posedon: bad block detection hmm... Jan 05 03:09:48 gena2x, I working only on ubifs freankly speaking) (at least right now) Jan 05 03:10:03 5am, I didn't sleep, so, I'll forgot all what you say) Jan 05 03:10:24 max_posedon: same here, you know. Jan 05 03:10:51 if it doesn't use mtd at all where do you make it work with 2k blocks? Jan 05 03:10:56 that's what I wonder Jan 05 03:11:03 tmzt: yes. Jan 05 03:11:10 gena2x: reread. Jan 05 03:11:11 I mean use mtd's oobsize and other settings Jan 05 03:11:34 tmzt, I just use ubi.mtd=6,2048 in kernel params Jan 05 03:11:48 I'm not sure that it means, I don't understand all background Jan 05 03:12:12 This adds an extra check when data is written to the flash. The Jan 05 03:12:15 NAND flash device internally checks only bits transitioning Jan 05 03:12:17 from 1 to 0. There is a rare possibility that even though the Jan 05 03:12:20 device thinks the write was successful, a bit could have been Jan 05 03:12:22 flipped accidentally due to device wear or something else. Jan 05 03:13:01 6 is the mtd device then? Jan 05 03:13:09 tmzt: i think ubi works on top of MTD layer. And 6,2048 means 6th mtd partition, 2048 block size. Jan 05 03:13:39 right Jan 05 03:13:40 ok Jan 05 03:13:42 PaulFertser: mb disable this? Jan 05 03:14:00 but the question is more if the mtd paramters are used Jan 05 03:14:09 gena2x: mb Jan 05 03:14:11 :) Jan 05 03:14:13 PaulFertser: 'extra check' :) Jan 05 03:14:17 and why does it need ,2048 shouldn't it get the caps from mtd? Jan 05 03:14:33 reads are much faster then writes Jan 05 03:14:47 tmzt: ubiformat /dev/mtd6 -s 2048 -O 2048 ; ubimkvol /dev/ubi0 -N rootfs -m Jan 05 03:14:54 it's crazy because google didn't fully write mtd for msm_nand, they put important parts in yaffs2 directly Jan 05 03:14:57 max_posedon: 2 Mb/s? Jan 05 03:14:58 ubifs write data only by big data Jan 05 03:15:13 I don't think that reread can kill it Jan 05 03:15:18 gena2x, I don't know Jan 05 03:15:19 max_posedon: do you think this is so fast? Jan 05 03:15:25 max_posedon: dd Jan 05 03:15:29 so switching fs means extracting that stuff Jan 05 03:15:32 max_posedon: dd told us. Jan 05 03:15:41 and moving it to mtd where it belongs Jan 05 03:15:45 fsck, my new kernel doesn't boot, unknown block Jan 05 03:15:59 rootdelay? Jan 05 03:16:37 tmzt: i think ,2048 is unnecessary Jan 05 03:17:00 PaulFertser, I think its really necesserry, Jan 05 03:17:07 max_posedon:no, really 6mb/s Jan 05 03:17:09 at least I had problems with mount Jan 05 03:17:27 gena2x, you have ubifs on neo with .32 kernel? Jan 05 03:17:35 e.g. where are you doing such tests) Jan 05 03:17:47 kernel can do upto 30Mb cache remember it Jan 05 03:17:48 gena2x: not yet. Just dd if=/dev/mtd6 bs=256k of=/dev/null Jan 05 03:18:00 ah, its about seq read Jan 05 03:18:56 Now i wonder very much why distributions on flash are working in same speed to one on SD Jan 05 03:19:13 SD is 2.6M/b, flash is 6MB/s Jan 05 03:19:20 Should be twice faster Jan 05 03:19:24 strange Jan 05 03:19:29 jffs2 is sync Jan 05 03:19:40 on writes i guess? Jan 05 03:19:48 it means when you write 10b, you write 2k Jan 05 03:20:00 and if you do sequence writes of 10b files Jan 05 03:20:01 that's true of nand Jan 05 03:20:10 max_posedon: but reads is 99% of bootup? Jan 05 03:20:24 gena2x, I think yes Jan 05 03:20:53 max_posedon: hm... mb we have sync syslogd? Jan 05 03:21:06 syslog log to tmpfs Jan 05 03:21:28 max_posedon: really? /var/log on tmpfs? Jan 05 03:21:54 max_posedon: I don't believe Jan 05 03:22:03 max_posedon: in shr? Jan 05 03:22:56 max_posedon: tmzt: the second parameter is "Optional \"vid_hdr_offs\" parameter specifies UBI VID header position and data starting position to be used by UBI." Jan 05 03:26:17 hmm Jan 05 03:27:10 rootdelay=10 helps Jan 05 03:27:20 so, I have fully working default shr Jan 05 03:27:31 fast enough booting on ubifs Jan 05 03:28:02 PaulFertser, http://pastebin.ca/1738321 Jan 05 03:28:11 my patch to shr! kernel Jan 05 03:28:17 tmzt: according to the comment in the source it should align to the first page boundary by default, no idea why it doesn't work without explicit specification, probably some bug. Jan 05 03:28:50 PaulFertser, are you ready commit it? or you want me to do few more plays/games Jan 05 03:29:13 max_posedon: i want to sleep again right atm, let's commit it tomorrow. Jan 05 03:29:28 ok, I'll remind all when I woke up Jan 05 03:29:41 PaulFertser: so, why not disable what verify? Jan 05 03:29:45 JaMa, http://pastebin.ca/1738321 Jan 05 03:29:55 but lets now do some speed tests) Jan 05 03:30:19 max_posedon: it's also of limited use anyway because not all SHR users are able to do your boot-from-SD/format manually/etc trick instead of flashing. Jan 05 03:30:45 gena2x: i will if it's not present in defconfigs of other widely used devices. Jan 05 03:30:54 * PaulFertser goes to sleep Jan 05 03:30:59 PaulFertser, may be I'll find time to test ubinize Jan 05 03:31:05 may be somebody else do it Jan 05 03:31:07 PaulFertser: :) Jan 05 03:31:13 max_posedon: results of speed tests are of more interest imho Jan 05 03:31:19 PaulFertser: i thought - how to check it :)) Jan 05 03:31:25 but people will have working ubi in kernel its nice Jan 05 03:31:36 PaulFertser: and what is source of all options of that sort :) Jan 05 03:32:08 gena2x: heh :) Jan 05 03:32:40 max_posedon: pleas, publish result of your speed measurement... Jan 05 03:32:55 installing coreutils(dd) now Jan 05 03:33:09 max_posedon: I wanted to do similar quest with ubifs in my own. Jan 05 03:33:49 max_posedon: So, very interesting to read. Jan 05 03:34:55 ok, what tests you want? Jan 05 03:35:21 I guess is up to you :) Jan 05 03:35:39 I planned bonnie++, but really unsure it is best testsuite. Jan 05 03:35:39 so, plz wait) Jan 05 03:37:07 gena2x, shr haven't bonnie++,) Jan 05 03:37:25 max_posedon: i guess shr have to testsuite at all Jan 05 03:37:44 max_posedon: what's why i am using debian. Jan 05 03:38:10 max_posedon: everything is included to it :) Jan 05 03:38:25 thats why I'm working on gentoo) Jan 05 03:39:00 oh, update-alternatives.. Jan 05 03:39:01 hm.. on freerunner? Jan 05 03:39:24 With native compilation? Jan 05 03:39:26 gena2x, yep, not all done, but will be some day Jan 05 03:39:41 I prefer native on sheevaplug, or cross Jan 05 03:40:00 Ohhh. now i got idea with sheevaplug......... Jan 05 03:40:44 sheevaplug is awesome) and its really good for testing any embedded stuff Jan 05 03:40:54 I already test on it btrfs and ubifs) Jan 05 03:41:01 where did you get it? Jan 05 03:41:12 bought online Jan 05 03:41:30 max_posedon: in russian shop? Jan 05 03:41:43 nope, in some USA's Jan 05 03:42:08 Such things not so easy to buy here... Jan 05 03:42:12 with 3days delivery, but final price was 2 times bigger Jan 05 03:42:17 here, where? Jan 05 03:42:24 max_posedon: in spb. Jan 05 03:42:37 just by it by mail) Jan 05 03:43:06 max_posedon: with price x2... Jan 05 03:43:27 I was payed x2 because of belarussian laws Jan 05 03:43:48 max_posedon: we have very similar taxes i guess. Jan 05 03:44:03 nope Jan 05 03:44:20 max_posedon: ??? Jan 05 03:44:23 http://max-posedon.livejournal.com/18230.html more info here (in russian) Jan 05 03:44:41 our minimal taxfree mail price 10 eur Jan 05 03:44:52 your about 500 eur Jan 05 03:45:01 so, you will pay only for delivery Jan 05 03:45:23 I payed for (device+deliver)*1.3(30% taxes) Jan 05 03:45:46 + some crazy payments just for entering to airport mail office Jan 05 03:48:15 how does btrfs work on arm? Jan 05 03:48:18 http://pastebin.ca/1738336 Jan 05 03:48:25 tmzt, works nice) Jan 05 03:48:42 1st test at link, if somebody have jffs2 and can compare will be nice Jan 05 03:49:37 http://pastebin.ca/1738338 Jan 05 03:49:46 max_posedon: Funny :) selling tickets to custom :) Jan 05 03:53:12 so, I don't know which tests I should do to compare with jffs2 Jan 05 03:53:29 Just do bonnie++ Jan 05 03:53:45 I have results for sd card I think Jan 05 03:54:16 yes, with ext2 Jan 05 03:54:22 I haven't bonnie++ on shr Jan 05 03:54:36 just compile it... Jan 05 03:55:26 I disagree, not today Jan 05 03:55:31 ) Jan 05 03:56:11 ok, ok. finally, I can go throught that quest too for xp :) Jan 05 03:57:20 I just see more interesting test Jan 05 03:57:26 lest compare result) Jan 05 03:57:29 which one? Jan 05 03:57:55 max_posedon: bonnie have simple makefile Jan 05 03:58:05 max_posedon: so it's easy to compile Jan 05 03:58:15 ah its wrong, I need reboot Jan 05 03:58:29 I just want test time between framework started, and gsm connected Jan 05 03:59:03 I'll test time since poweron to gsm connected (sim pin disabled) Jan 05 03:59:31 max_posedon: not so bad. 'use case test' Jan 05 04:00:28 I have loglevel=8, but anyway) Jan 05 04:01:21 75sec to X Jan 05 04:01:53 105 sec to enlightment Jan 05 04:02:27 hm. how to you wish to calculate jffs2 timings? Jan 05 04:02:36 nohow) Jan 05 04:03:04 180 sec to gsm connected Jan 05 04:03:18 not awesome, not so bad Jan 05 04:03:19 ) Jan 05 04:03:30 max_posedon: that's why i measure ext2 before going with my experiment ;) Jan 05 04:04:00 max_posedon: In fact i wanted a bit different thing - best fs for sd. Jan 05 04:04:24 gena2x, I think it will be btrfs without compression for now Jan 05 04:04:29 for really big sds Jan 05 04:04:36 (like 8G+) Jan 05 04:04:51 max_posedon: btrfs? on flash? why? Jan 05 04:05:03 not fastest, but optimized for ssd and for reliablity Jan 05 04:05:04 max_posedon: i thought about some kind of logfs Jan 05 04:06:12 gena2x, have you your frameworkd.log near you? (nice if you have sim pin disabled) Jan 05 04:06:51 90sec between framework starts and gsm connected Jan 05 04:06:59 max_posedon: which framework, fso2 (vala stuff)? Jan 05 04:07:01 I really like thats results Jan 05 04:07:12 max_posedon: I have one from debian only Jan 05 04:07:30 tmzt, shr-unstable I don't know Jan 05 04:07:38 some fso2 framework mix) Jan 05 04:08:22 how fast does it usually start up? Jan 05 04:08:28 max_posedon: think btrfs is too heavy. guess all that feature not needed in fr. Jan 05 04:08:48 tmzt, longer, but... I don't know Jan 05 04:09:21 may be I just want ubifs to be really faster) Jan 05 04:09:30 max_posedon: btw it's usual optimiation problem - you do something, think it became better, but measurement with real clock 50% of time shows it is now :) Jan 05 04:09:56 max_posedon: I fall in that pit several times Jan 05 04:10:43 max_posedon: that's why unbiased bonnie is needed :) Jan 05 04:11:01 ok) Jan 05 04:11:11 but now I must sleep Jan 05 04:11:15 at least 4h Jan 05 04:11:19 me2 Jan 05 05:40:49 hi, failed to apply patch to local tar.bz2 file using bitbake file, any clue ? **** ENDING LOGGING AT Tue Jan 05 05:57:21 2010 **** BEGIN LOGGING AT Tue Jan 05 05:58:14 2010 Jan 05 06:53:24 morning Jan 05 06:53:47 satish: can you paste the .bb, the .patch and the link to the source somewhere? Jan 05 06:53:52 satish: probably path problems Jan 05 06:53:57 satish: also, the error log would be interesting Jan 05 08:49:40 max_posedon, PaulFertser, UBI-fans: http://git.openembedded.org/cgit.cgi/openembedded/commit/?id=590b54897fcca0e8568cdc2acd50cf74002b50fb Jan 05 08:50:32 thanks!) Jan 05 08:50:45 JaMa, its really works here, I'm happy Jan 05 08:54:52 anyone know why /sys/devices/system/timer/timer0/dyn_tick doesn't exist on SHR? trying to squeeze out some more battery life by experimenting with the stuff in http://markmail.org/message/nuoduc23qlty2slk Jan 05 08:56:36 max_posedon: btw do you have any idea if there is some "filesystem created timestamp" in ubifs header? Jan 05 08:58:38 max_posedon: just because while creating ubifs which is later ubinized and ubifs after that with OE, ubifs are a bit different in first few bytes.. Jan 05 08:58:38 at least I saw some header during mount Jan 05 09:00:01 max_posedon: I'm going to remove standalone ubifs from IMAGE_FSTYPES for om-gta02 because its generated because of "ubi" which is ubinized Jan 05 09:00:28 I'm ok with it Jan 05 09:00:40 but I'm not sure, that images works at all Jan 05 09:01:06 I'm not sure that ubinize and kernel have same opinions about ubifs Jan 05 09:01:17 max_posedon: here is that header difference http://www.mail-archive.com/shr-user@lists.shr-project.org/msg02410.html Jan 05 09:02:14 problem that I face similar problems also on sheevaplug with .32 Jan 05 09:04:32 max_posedon: you can also check there that error you said yesterday was almost imposible, this was from mkvol.. but the same I got with ubiformat while reseting that err. count Jan 05 09:05:29 but with "kernel BUG at mm/slab.c:3002!" its not sure where the real error was created.. Jan 05 09:07:48 [459531.915000] Internal error: Oops: 817 [#1] PREEMPT <- Jan 05 09:07:58 I think disabling BGT fixes such errors Jan 05 09:08:24 and yes, exactly same backtrace was on sheevaplug) Jan 05 09:08:40 but isnt BGT for "more cleaver logic to lower wearing"? Jan 05 09:09:01 bgt just background thread as I understand, I'm not sure now Jan 05 09:09:30 http://www.anandtech.com/storage/showdoc.aspx?i=3631 Jan 05 09:09:31 but this is what I touched, bgt and just debug info Jan 05 09:10:52 max_posedon: from this article I got idea that there is lots of logic needed to check nand usage ie in page 4 about cleaning lady iirc Jan 05 09:11:36 PaulFertser: to pick up last night's discussion ... one problem with exposing internal NAND would be that you could read what their secret algorithms are doing. now, any self-respecting reverse engineer would laugh at this, but then their lawyers may now. (and sometimes, the lawyers have good arguments, such as the fine distinction between something being accessible and something being made accessible.) Jan 05 09:13:53 JaMa, btw, can you add bonie++ in feeds?) Jan 05 09:15:23 JaMa, this is big problem for SSD, but not for ubifs Jan 05 09:15:44 SSD/mmc/etc, have problems with info about their content Jan 05 09:16:19 and isn't it handled in ubifs in that BGT ? Jan 05 09:16:33 nope Jan 05 09:16:51 BGT just option of implementation Jan 05 09:16:56 (as I understand) Jan 05 09:17:56 anyway, ubifs doing write/read in optimized way, just because it knows their backend Jan 05 09:18:58 hmm from Kconfig "The thread may be also be enabled/disabled via UBI sysfs." Jan 05 09:19:23 of course we can continue working on config, but *now* we have defenetly working kernel Jan 05 09:19:43 so, we can enable/disable on-by-one to test where is all fails Jan 05 09:20:45 but I think thereis one more interesting step, pooting rootfs to squashfs Jan 05 09:21:34 As I wrote before on ML.. it works for me only partially.. it haven't survived few subsequent mounts for me :/ Jan 05 09:21:52 but maybe other users like you will have more luck :) Jan 05 09:23:02 you can try our current kernel + with creating ubi manually Jan 05 09:23:17 may be it was smth wrong not with luck, but with kernel) Jan 05 09:23:31 max_posedon: I tried many times in last few moths :) trust me .. Jan 05 09:23:56 max_posedon: I'm testing it from time to time with every kernel bump I do in OE repo.. Jan 05 09:24:10 but you never know, does kernel correct or not, but now can be sure) Jan 05 09:24:22 max_posedon: flashin ubinized as well as created on neo.. Jan 05 09:24:43 I just doesn't understand how jffs2 can work, and ubifs doesn't Jan 05 09:25:03 of course if jffs2 doesn't work for you, its another story) Jan 05 09:25:29 and worse is that even if it works and I can mount it.. then it fail after reboot (same kernel, same nand etc.) Jan 05 09:25:46 thats why I always just try and then forgot for at least a week :) Jan 05 09:26:07 ok, but for me, I did few reboots now Jan 05 09:26:18 works reliable Jan 05 09:26:56 now with your success test and someone (maybe ist also you) success story on SHR ML.. it make me to push updated ubinize params to OE (which I had locally for some time and wasn't sure if it makes things better or worse :) Jan 05 09:27:12 max_posedon: still on that 2.6.29-rc3 without drm right? Jan 05 09:27:21 max_posedon: can you try with 2.6.32? Jan 05 09:27:26 default shr-U, I think it means yes Jan 05 09:27:34 JaMa, yes, but at evening Jan 05 09:27:55 max_posedon: btw bonnie++ added to feeds.. I'll start build on buildhost.. Jan 05 09:28:03 and now back to daywork.. Jan 05 09:28:20 max_posedon: thanks for testing and results.. I really apreciate that Jan 05 09:28:52 max_posedon: first good results and good info after long time (before that only questions :)) Jan 05 09:32:52 is there any known issue with 2.6.31 and the SD card? It doesn't look like it can see my card using this kernel Jan 05 09:33:41 rebelThor: I was booting from SD card with 2.6.31 built from OE Jan 05 09:34:14 JaMa: I don't have the folders inside /media, maybe it's a different issue then Jan 05 09:34:19 perhaps the automounter or something Jan 05 09:34:31 I'm not using automounter.. Jan 05 09:34:54 shouldn't I have some mmc* devices in /dev? they're not there Jan 05 09:34:55 rebelThor, I think you should check, which fs supported Jan 05 09:35:00 and which modules you have Jan 05 09:35:20 but I had same problems with .31 some_gta_defconfig Jan 05 09:35:41 and maybe its not enabled in .31 config.. its building much less modules/features than what old defconfig-packaging did Jan 05 09:36:45 i just got one from the mrmoku folder , maybe it doesn't contain everythign it should.. it's way faster though Jan 05 09:36:55 rebelThor: or /dev/block if using mdev Jan 05 09:37:27 faster because less is built in? Jan 05 09:38:11 tmzt: i have some mtdblock[1-6] in /dev/block, you think these are my sd partitions? :) Jan 05 09:38:52 no Jan 05 09:38:58 mtd is flash not sd Jan 05 09:39:21 right.. then it's not there :) maybe I can find a 2.6.32 image I can flash :P Jan 05 09:40:04 rebelThor: yes thats OE built.. the same version I have.. Jan 05 09:40:31 JaMa|Wrk: so you're saying I should be able to see the sd? that's odd Jan 05 09:40:37 tmzt: faster because it has Weiss's DRM patches.. Jan 05 09:40:53 rebelThor: I'm saying that I was able to boot from sd with that :) Jan 05 09:41:20 rebelThor: and yes.. I could see sd after that too (as mmcblk0p1 etc) Jan 05 09:41:44 good Jan 05 09:41:59 add rootdelay=1 at least Jan 05 09:42:32 i'm getting some "cannot load /lib/modules/2.6.31 ..." errors when booting via uboot Jan 05 09:42:42 what am I missing? Jan 05 09:44:35 rebelThor: have you downloaded and unpacked modules tarball? Jan 05 09:44:45 nope, I didn't know I have to :D Jan 05 09:45:22 then you're missing all modules :) Jan 05 09:46:42 i even got a nice WSOD :)) just installed the modules, hopefully it'll work now, as it's too damn fast to give up on Jan 05 09:47:38 rebelThor: you got a nice WSOD every time phone tryies to suspend.. Jan 05 09:47:53 rebelThor: thats unwelcome sideeffect of DRM patches making it way faster.. Jan 05 09:48:15 interesting.. Jan 05 09:48:29 hmm, suspended, though it doesn't even want to come back from suspend at all Jan 05 09:48:50 rebelThor: you can reboot it over ssh in most cases.. Jan 05 09:49:06 DRM typically slows things down a little, apart from the possibility of the waitqueue stuff (maybe makes a big difference- not sure: my FR is nice and fast with FIFO+waitq, but I haven't tried FIFO-only for long enough to judge) Jan 05 09:49:52 it feels faster to me.. I haven't tried some objective benchmark Jan 05 09:50:53 it's kind of bad that suspend doesn't work at all atm :( Jan 05 09:51:58 rebelThor: yes :/ that's why I'm using 2.6.32 only with neo connected to pc (its most of the time) and I'm switching back to 2.6.29-rc3 when going out with phone.. Jan 05 09:52:04 rebelThor: are you using 2.6.29+drm or 2.6.3{1,2}? Jan 05 09:52:21 Weiss: 2.6.31 atm, didn't find any image for 2.6.32 to test Jan 05 09:52:53 rebelThor: I'm adding it to shr buildhost atm Jan 05 09:53:15 Weiss: is 2.6.29+drm as fast as 2.6.31? :) i'm willing to give it a try Jan 05 09:53:25 rebelThor: but you can build it from OE .. its there from the day larsc announced it for experimental freaks :) Jan 05 09:53:47 if it fails on *suspend* (not resume) with 2.6.3x, try enabling MMC_UNSAFE_RESUME_or_whatever_it_was Jan 05 09:54:04 2.6.29 should be as fast as .31 as far as graphics is concerned.. Jan 05 09:54:05 JaMa|Wrk: i gave up on building stuff myself due to time constraints and other factors :) I just dont have the time to wait for it Jan 05 09:54:32 Weiss: .31 just feels faster in everything, booting up, launching apps, etc Jan 05 09:55:26 cool :D Jan 05 10:00:05 weird, FAT: codepage cp437 not found when trying to mount a vfat on 2.6.31 Jan 05 10:00:38 missing cp437 in config i guess.. Jan 05 10:01:11 kernel config, right Jan 05 10:01:13 ? Jan 05 10:01:36 yes Jan 05 10:02:03 well then, back to .29 for me :) Jan 05 10:05:00 wpwrak: (secret algorithms) eww :/ Jan 05 10:05:17 PaulFertser: i can make Qi turn on the LED during resume even before RAM is working, right? Jan 05 10:06:58 lindi-: yes, you can :) Jan 05 10:10:17 PaulFertser: how about going to suspend with watchdog running? Jan 05 10:10:29 PaulFertser: and having Qi wake up every 40 seconds to ACK the watchdog? Jan 05 10:10:39 mrmoku|away: ping Jan 05 10:11:45 lindi-: how do you distinguish between user-set RTC alarm and an RTC alarm set by Qi to wake up itself? Jan 05 10:15:10 lindi-: and do i understand it right that you want to increase reliability by using watchdog but you want to save power by quickly suspending again, without going through full kernel suspend/resume? Jan 05 10:15:47 PaulFertser: yes Jan 05 10:16:45 PaulFertser: but having to add RTC and watchdog drivers to steppingstone code does not sound fun Jan 05 10:16:56 lindi-: exactly Jan 05 10:17:25 PaulFertser: any other ideas? Jan 05 10:21:30 lindi-: To me it seems that core FR is already stable enough. Probably the kernel problems will be solved by upgrade to new mainline release. If i had an energy i'd spent it on finally upstreaming all the stuff and switching to recent kernels. Also i'd migrate to the fso2 as much as i could and extended fso.el to use opimd. Jan 05 10:25:21 PaulFertser: i think a silent crash every two months is still too much. it's a different thing if it crashes when the user is using it and knows he need to reboot it Jan 05 10:25:31 s/need/needs/ Jan 05 10:26:04 PaulFertser: i still don't know if it crashes during suspend, resume or somewhere in the middle :P Jan 05 10:28:33 lindi-: what is the maximum time hardware watchdog can be configured for? Jan 05 10:29:26 did someone build Qi with latest toolchain used in SHR? Jan 05 10:29:28 PaulFertser: 42 seconds i think Jan 05 10:29:53 http://pastebin.ca/1738603 Jan 05 10:29:55 PaulFertser: bit too short for doing full suspend/resume Jan 05 10:30:30 JaMa|Wrk: i think it's the same problem that was reported recently on ML (about debian armel native compilation). Jan 05 10:30:31 PaulFertser: maybe i need an external watchdog that will start to beep if it is not ACKed :) Jan 05 10:30:47 JaMa|Wrk: i'll check/apply his fix today. Jan 05 10:31:15 PaulFertser: can serial console wake the system up? Jan 05 10:32:02 lindi-: no idea, i need to check the manual to answer that. Jan 05 10:35:07 PaulFertser: however it needs extra hardware so it is not very practical Jan 05 10:37:10 VDVsx: hi Jan 05 10:37:31 VDVsx: are you the remoko author? Jan 05 10:38:08 PaulFertser: argh, i don't like trac at all, it only says "2 weeks ago" and not exact date of a bug comment Jan 05 10:38:22 lindi-: i do not like trac too :| Jan 05 10:38:37 lindi-: there's some way to get a date from it, it's encoded in the hyperlink or something. Jan 05 10:38:43 PaulFertser, yes, but I've the project in standbye, don't have time to improve it atm :( Jan 05 10:38:57 VDVsx: can you at least update us on the current status? Jan 05 10:39:09 It's not obvious from the commit log. Jan 05 10:39:12 lindi-: (console) unfortunately not, because the clock is stopped. you could run rx to an eint-capable pin but you'd still lose the first character(s) Jan 05 10:39:57 PaulFertser, my last commits were to support bluez4 thus SHR, nothing new there Jan 05 10:40:21 VDVsx: yes, but does the latest version work with bluez4? Jan 05 10:40:35 PaulFertser, yes Jan 05 10:40:54 last svn version Jan 05 10:42:46 VDVsx: ok, thanks a lot! Jan 05 10:44:06 PaulFertser: great! let me know if that patch works I'll apply it to OE.. Jan 05 10:46:21 mickey|zzZZzz: ping, seems like fsotdld is missing in sane-srcrevs-fso.inc, which revision should be used? ${FSO_CORNUCOPIA_SRCREV} there is too old because there is no code in fsotdld and latest revision fails to compile here :/ Jan 05 10:46:43 mickey|zzZZzz: with this http://pastebin.ca/1738619 Jan 05 10:47:36 JaMa|Wrk: let me do the right thing and fix it in the repo please ;) Jan 05 10:50:20 PaulFertser: even better.. but still ping me to bump revision :) Jan 05 10:50:40 JaMa|Wrk: please bump the remoko version to r148 Jan 05 10:50:45 Counts? Jan 05 10:50:46 ;) Jan 05 10:50:57 yes.. i guess when its from you :) Jan 05 10:51:12 but I already did that when you said it first in #openmoko :) Jan 05 10:51:16 just not pushed Jan 05 10:51:21 because of lunch.. Jan 05 10:51:29 JaMa|Wrk: enjoy your meal ;) Jan 05 11:05:42 JaMa|Wrk: try building vala 0.7.9.3 and then fsotdld Jan 05 11:10:37 hmm is yesterdays unstable image recommended? Jan 05 11:14:31 \o/ finally found the resizing bug in my notes application :D Jan 05 11:19:29 * PaulFertser wished to know why the hell Andy used ulong/ushort/uint and u_int32_t etc when he should have used C99-compatibles uint32_t etc... Jan 05 11:19:35 wpwrak: do you know? Jan 05 11:20:12 PaulFertser: i often ask myself that question, with a lot of code from all kind of people ... Jan 05 11:21:43 PaulFertser: another favourite of mine are those libraries that blissfully rename all basic types Jan 05 11:21:59 wpwrak: especially ulong and friends is what looks strange, they shouldn't be used even if there's no C99 support, it just doesn't make sense. If you need unsigned long, write unsigned long. If you want some particular size, name it appropriately... Jan 05 11:22:23 wpwrak: heh, he also used u32 :) Jan 05 11:22:29 PaulFertser: and let's not forget the well-meant foo_t types that mimick the ANSI C type naming style but completely overlook the fact that these names are reserved by that very same standard Jan 05 11:22:45 PaulFertser: (u32) there's at least in good company :) Jan 05 11:22:51 :D Jan 05 11:22:54 s/there's/there he is/ Jan 05 11:25:42 mickey|zzZZzz: it was after upgrading vala-native Jan 05 11:27:19 perhaps it wasn't rebuilt Jan 05 11:27:29 fsotdld compiles fine with vala >= 0.7.9.1 here Jan 05 11:27:44 hm. Jan 05 11:28:09 mickey|zzZZzz: which revision of fsotdld? Jan 05 11:28:12