**** BEGIN LOGGING AT Sun Jul 11 02:59:57 2010 Jul 11 06:52:32 mrmoku|away: yeah.. sounds like you have the exact same failure as me Jul 11 06:52:43 (it should break during build, because the source code is currently broken) Jul 11 07:36:58 gena2x: ping Jul 11 08:55:54 * Weiss installs Debian, having calmed down a bit from last night Jul 11 10:13:59 gena2x: no need to convince me to remove debug :) I was using 2.6.32 without debug before you published you analysis.. so my point was to switch to 2.6.32 sooner instead of improving 2.6.29-rc3... (but it took many months :/) Jul 11 10:23:57 <[Rui]> JaMa|GoNe: I haven't had any of those hangups I noticed with wifi if I used iliwi instead of a script that launched wpa_supplicant and udhcpc Jul 11 10:24:28 <[Rui]> JaMa|GoNe: it was based upon PaulFertser's script and one thing it did is relaunch udhcpc if the connection was cut Jul 11 10:24:55 <[Rui]> I think that for some reasone the connection was being frequently cut and perhaps I was exhausting the memory with udhcpc instances Jul 11 10:25:10 <[Rui]> I can't prove it, but I don't see much other explanations Jul 11 10:26:03 <[Rui]> and I notice that I frequently have to stop and launch iliwi again in order to stop and start the wifi interface because the connection was hung (got IP, wifi connection, but no traffic) Jul 11 10:26:06 badcloud: pong Jul 11 10:32:15 JaMa|GoNe: 1) nobody did attribute (except lars maybe) speed change with nodebug, everyone were talking about how useful is debug info and how much it is needed. 2) yes, talk were about .29, which were changed to .32 after 9 months, and only after accident, and still, speed change were attributed to gcc and udev :) But really, problem is gone, as everyone now using nodebug kernel. And problem really were with Doc who told something like 'do he ask any Jul 11 10:32:16 questions?', and after his phrase, i were unable to grab any attention to problem. Jul 11 10:35:25 <[Rui]> rin* Jul 11 10:35:27 <[Rui]> *grin* Jul 11 10:36:45 gena2x: well I've said back then, that defconfig used in 2.6.32 doesn't have debug and that it's faster... (even with old gcc and with udev) Jul 11 10:37:30 gena2x: and for last 9 months or more, I was trying to convince other shr devs that new kernel is worth the troubles ... Jul 11 10:38:47 JaMa|GoNe: ok, really, no problem now. i were just (again) were frustrated that doc were attempt to destroy my idea, which i hope may be very good (did you took a look to context?). Jul 11 10:38:56 <[Rui]> JaMa|GoNe: so far I'm happy (safe for that WS bug). gena2x no, I haven't compiled that ThibG's patch Jul 11 10:40:33 gena2x: I'm still very gratefull for you debug analysis, and also your ts patch fixed .32 show stopper for us, to thanks again Jul 11 10:40:34 sometimes WS happens here (but i haven't noticed any pattern like mrmoku did :P), but it's acceptable Jul 11 10:41:10 JaMa|GoNe: ah, may be include ThibG patch to fix WSed also Jul 11 10:41:12 JaMa|GoNe: ah, may be include ThibG patch to fix WSes also Jul 11 10:41:16 and thanks for reminding, i wanted to add rule to my rules.yaml to make neo suspending after holding power button for 3 sec :D Jul 11 10:41:56 if you're takling about engines initialization patch from ThibG then it's included already Jul 11 10:42:24 JaMa|GoNe: let me check.... Jul 11 10:44:49 mrmoku|away: can we push that fsogsmd patch to cornucopia repo? iirc dos1 talked with mickey|bbl about it, only debian would need to apply oposite patch, but maybe ThibG will upgrade debian kernel to .32 or .34 sooner then new fsogsmd debian package Jul 11 10:45:26 JaMa|GoNe: which fsogsmd patch? Jul 11 10:45:27 JaMa|GoNe: you mean "0001-glamo-core-initialize-engine-states-as-disabled.patch" Jul 11 10:45:36 JaMa|GoNe: no, i mean other patch. Jul 11 10:46:10 dos1: sysfs nodes (only in OE) Jul 11 10:46:49 JaMa|GoNe: i talked about fsodeviced and UsbHost, which i already pushed Jul 11 10:46:50 gena2x: ah, then I didn't notice it on ML Jul 11 10:47:13 JaMa|GoNe: we discussed it with Rui a bit here. Jul 11 10:47:43 dos1: mrmoku|away updated fsogsmd patch in OE for newer cornucopia rev.. so I expected it to be pushed upstream and removed in OE.. Jul 11 10:47:48 JaMa|GoNe: it fixes all kinds of WSes in blank (may be in suspend too, i have no suspend in my .34) Jul 11 10:48:07 gena2x: ah.. I can read log tomorrow.. today I'm here only for few mins Jul 11 10:49:13 JaMa|GoNe: sure. just may be useful for users to have this patch before ThibG will prepare something polished. Jul 11 10:51:36 JaMa|GoNe, "only debian would need to apply oposite patch"? Jul 11 10:53:25 so, Debian boots, but I can't seem to get a getty on the serial console Jul 11 10:57:06 is there something I'm missing? Jul 11 10:57:30 Weiss, maybe you have to add ttySAC2 to /etc/inittab? Jul 11 10:58:47 I did, but I get a message about "S respawning too fast" (and no useful error message that I can get to) Jul 11 10:59:45 fortunately I can still edit files on Debian from SHR (in NAND) to try to get the network up.. Jul 11 11:01:29 now the FS is corrupted so that dropbear can't start :/ Jul 11 11:01:39 Weiss: %) Jul 11 11:02:04 in debian, on sd? Jul 11 11:02:05 knew I should've taken an image of the SD card straight after installation Jul 11 11:02:09 yep Jul 11 11:02:37 it's damn hard to archieve corruptes fs :) Jul 11 11:03:10 find: cannot delete `./dropbear.pid': Input/output error Jul 11 11:03:19 [ 14.150000] EXT2-fs error (device mmcblk0p1): ext2_lookup: deleted inode referenced: 94661 Jul 11 11:03:23 [ 14.165000] EXT2-fs error (device mmcblk0p1): ext2_lookup: deleted inode referenced: 94661 Jul 11 11:03:26 [ 14.180000] EXT2-fs error (device mmcblk0p1): ext2_lookup: deleted inode referenced: 94661 Jul 11 11:03:29 find: `./network': Input/output error Jul 11 11:03:30 fantastic. Jul 11 11:03:31 etc Jul 11 11:03:51 you able to archieve fs error everywhere. Jul 11 11:03:56 reinstalling.. Jul 11 11:04:04 yep, and this is the FR without the NAND problems Jul 11 11:04:58 i used ext2 for weeks, doing really weird expirements, sudden reboots and so, and never had something like your errors. Jul 11 11:05:17 * Weiss got errors on his fourth unclean reboot Jul 11 11:05:34 (can't do a clean shutdown when neither SSH nor serial console works...) Jul 11 11:05:56 you need to add sync after boot :) Jul 11 11:06:36 Weiss: ah, add xterm and X11. this may also help to bring usb0 for a first time. Jul 11 11:06:52 really? why? Jul 11 11:06:59 I noticed nothing in /etc/network/interfaces Jul 11 11:07:12 (I'm trying to do a minimal installation + GCC, then I'm going to bring X up from scratch) Jul 11 11:07:29 (actually, what's the X version in Debian? I need the newest..) Jul 11 11:07:32 ah, your situation is more special, of course. Jul 11 11:07:37 Weiss: use ext3 instead of ext2 Jul 11 11:07:57 actually ext3 is too slow. Jul 11 11:08:12 but it doesn't fail like that :P Jul 11 11:08:32 yes, yeah. Jul 11 11:08:38 * Weiss will settle for anything which lets him hack without unspeakable agony right now :( Jul 11 11:08:55 right now i am using xfs :) Jul 11 11:09:03 hehe, cool Jul 11 11:09:14 but can't recommend it :) Jul 11 11:09:35 512 mb card is very fast to backup. Jul 11 11:10:05 eh. i have few time today till may be evening... Jul 11 11:10:18 yeah... ideally, I'd make a 512Mb "Minimal SD image for X hackers" Jul 11 11:32:24 Weiss: FYI from openmoko-debian: Last sid upgrade broke X and USB ethernet :/ Jul 11 11:32:25 :) poor weiss... Jul 11 11:32:25 (EE) XKB: Couldn't compile keymap Jul 11 11:32:25 and usb ethernet? Jul 11 11:32:25 kernel upgraded? Jul 11 11:32:27 no Jul 11 11:32:29 no idea Jul 11 11:32:31 It's just working randomly Jul 11 11:34:25 gena2x: that's ok.. Sid's X is too old for me anyway, so I'll be installing from scratch I think Jul 11 11:34:44 I had ethernet working just before the FS died.. Jul 11 11:36:22 and kernel will be my own, of course.. Jul 11 11:36:47 (with g_ether compiled in rather than messing around with modules Jul 11 11:45:58 gena2x: I'm not sure the ethernet is more broken than it was before anyway. Jul 11 11:48:52 Don't take my frowning and frustration too seriously. Although I can get X working anymore now. Jul 11 11:48:54 can't* Jul 11 11:51:28 thrashold: it's ok, i am just very interested in Weiss's review of my timings idea and waiting for his setup, so informed him about possible troubles in debian installation. hope this didn't hurt anyone :) Jul 11 11:52:27 raw debian installed, backup up SD... Jul 11 11:52:52 backing* Jul 11 11:55:18 What is this timing idea? Jul 11 11:58:10 thrashold: glamo throughtput increase x1.8 times. Jul 11 11:58:21 This is something I'd want Jul 11 11:58:26 When I get my X working, that is :P Jul 11 12:01:12 thrashold: it not ready to use right now. Jul 11 12:01:39 I don't have anything working right now, breaking it more won't make a difference :P Jul 11 12:04:32 so, now ethernet and SSH work, but I don't know the default root password.. Jul 11 12:05:35 docs say "changeme", but that doesn't work.. Jul 11 12:06:14 chroot from other distro and change it :) Jul 11 12:07:02 just did :) Jul 11 12:08:09 root@(none):~# Jul 11 12:08:10 (wohoo) Jul 11 12:08:24 and, only took 143Mb Jul 11 12:08:55 Also, having a default root password on any distro/installer is a bad idea, IMHO. Jul 11 12:10:56 ls: cannot access /dev/ttySAC2: No such file or directory Jul 11 12:11:03 probably explains my getty problems Jul 11 12:15:22 NB: root's PATH should usually contain /usr/local/sbin, /usr/sbin and /sbin. Jul 11 12:15:29 (from apt-get) Jul 11 12:16:01 hi. I can not flash the kernel anymore, neither through SHR, not uboot. the dfu name of the fr is "rootfs" and it says the "kernel" space does not exist. any tips ? Jul 11 12:16:33 tsaitgaist: could you paste the command you're trying? Jul 11 12:18:25 weiss : http://pastebin.com/zLMU5ptf Jul 11 12:22:48 tsaitgaist: download dfu-util binary from openmoko site Jul 11 12:22:55 tsaitgaist: known bug. Jul 11 12:23:09 ok. thx Jul 11 13:32:06 ~seen GNUtoo Jul 11 13:32:09 gnutoo <~GNUtoo@host12-84-dynamic.51-79-r.retail.telecomitalia.it> was last seen on IRC in channel #openmoko-cdevel, 3d 14h 53m 36s ago, saying: 'there is a free tool to interact with their bootloader...'. Jul 11 13:54:34 ~seen PaulFertser Jul 11 13:54:35 paulfertser is currently on #openmoko (4d 9h 27m 8s) #openmoko-cdevel (4d 9h 27m 8s) #openjtag (4d 9h 27m 8s). Has said a total of 165 messages. Is idling for 1d 7h 56m 10s, last said: 'pabs3: you can have arbitrary number of fields in opimd, each field has a type (arbitrary amount of types), and each field can have any number of entries.'. Jul 11 13:56:00 PaulFertser: "arbitrary number of fields"? Jul 11 13:56:39 PaulFertser: only types are predefinied, you can define new fields via dbus and if you use non-definied field in your entry it just falls back to "generic" type Jul 11 14:02:07 * Weiss notices a potentially interesting thing about WS: it's never happened to him while showing just the console Jul 11 14:02:40 Weiss: you mean blank/unblank? Jul 11 14:02:48 yeah Jul 11 14:03:09 Weiss: with -fb, it easy thing to get in console. Jul 11 14:03:18 also with KMS Jul 11 14:03:38 Weiss: also you can put fr to fridge. Jul 11 14:04:09 hehe Jul 11 14:04:22 alternatively, I could change the way the mode set is performed after unblanking Jul 11 14:04:24 Weiss: this way before ThibG fix, i were able to get 100% sometimes recoverable, sometimes not recoverable, sometimes WSOD. Jul 11 14:05:05 Weiss: after fix, i forgot it blanked in fridge. seem it lay where for 30 minutes. Jul 11 14:05:06 yeah Jul 11 14:05:25 Weiss: and unblanked without any problems. Jul 11 14:13:30 it's <0C in fridge. Jul 11 14:13:42 i can't test suspend/resume :( Jul 11 14:14:16 but while thing is very interesting as seem this voltage stabilization time depends on temperature. Jul 11 14:16:26 I haven't seen wsod with andy-tracking a3587e4ed77974ad / Qi 9ef7754b8243457c for almost 6 months Jul 11 14:18:38 lindi-: yeah, I broke things with KMS. it's not usually "oD", but the modesetting is slightly screwy on unblank Jul 11 14:38:50 JaMa|GoNe: (patch to cornucopia) if mickey|sofa is ok with it... though probably PaulFertser will complain :P Jul 11 14:39:27 how to change uSD clock after boot? echo 10000000 > /sys/module/glamo_mci/parameters/sd_max_clk has no effect to real clock Jul 11 14:40:27 clock is computed based on a minimum value, and the maximum you provide with sd_max_clk Jul 11 14:41:29 changing sd_max_clk in runtime has no effect Jul 11 14:41:54 (it is only used in probe) Jul 11 14:42:42 Weiss: I have it with X stopped too (WS) Jul 11 14:43:23 is there any way how to change it in runtime? need to find sd_max_clk value for my uSD card Jul 11 14:43:36 max value which works Jul 11 14:44:12 I fear there isn't (atm, that is) Jul 11 14:48:04 mrmoku|eiscafe: ah ok, thanks Jul 11 14:48:14 Debian installed, NFS set up, starting to install X stuff.. Jul 11 14:49:20 (but anyway, glamo mmc clock shouldn't be card-depend, should it?) Jul 11 14:51:43 sd_max_clk=2500000 is max which works with my Sandisk 16 GB on 2.6.32 Jul 11 14:52:13 what value was on 2.6.29? it worked out of the box Jul 11 14:52:47 2500000 really? That's pretty low... Jul 11 14:53:12 Did you use some other parameters, back with 2.6.29? Jul 11 14:53:34 apt-get install chromium-browser: http://lindi.iki.fi/lindi/openmoko/chromium-browser/chromium-browser-5.0.375.99~r51029-3-openmoko1.png Jul 11 14:53:37 JaMa|GoNe: do you have an idea why local builds do not work anymore? Jul 11 14:53:57 lindi-: how fast is it? Jul 11 14:53:57 sd_drive, I mean Jul 11 14:54:03 dos1: got a benchmark? Jul 11 14:54:08 or s/fast/slow/ ;) Jul 11 14:54:10 lindi-, how fast does it consume all your RAM? Jul 11 14:54:29 lindi-: how do you feel? :P Jul 11 14:55:04 ThibG: no, I used only default params on 2.6.29 Jul 11 14:55:12 martix, ok Jul 11 14:56:02 I've been using chromium as my desktop browser, but discarded it as it consummed too much memory (twice as much as my previous one!) Jul 11 14:56:16 and I have 1GB memory... Jul 11 14:58:11 martix, default was 16666667 in 2.6.29 Jul 11 14:59:09 oh, remembering something! Jul 11 15:01:44 JaMa|GoNe, do you have this one in om? http://git.openmoko.org/?p=kernel.git;a=commit;h=5202f1be17877bf9d032ad0a747d21104e16965a Jul 11 15:02:31 This one is another potential bugfix too http://git.openmoko.org/?p=kernel.git;a=blobdiff;f=drivers/mfd/glamo-core.c;h=d7bcce3ecd7b4f38e9800842d7831afc89e7337b;hp=af50b82981a42d669e36335af417c46165d3c138;hb=b02fa027be49b9d9bc9642b2aa08dc2f3c7e32ec;hpb=8335be6be5033ca548bd0231b3d0c5db392e5bec Jul 11 15:04:40 * Weiss +1's the patch which un-breaks the hostbus registers Jul 11 15:04:59 likely to fix a lot of weirdness.. Jul 11 15:06:19 hdparm -t /dev/mmcblk0p5 -- 941.70 kB/sec :-/ Jul 11 15:06:44 * Weiss just created a Debian SD image which boots and has working SSH, sudo, bash, NFS, autotools and GCC. anyone interested? Jul 11 15:06:51 (I'll probably blog it anyway) Jul 11 15:07:01 (..but no X, Illume, FSO etc etc) Jul 11 15:07:14 cool :-) Jul 11 15:07:41 hopefully soon it'll be the same, but plus bleeding edge X.org Jul 11 15:08:12 Weiss, huh, was it so difficult? Jul 11 15:08:14 Weiss: blog it :) Jul 11 15:08:24 it can be useful Jul 11 15:08:46 ThibG: moderately fiddly (about 5 on a scale of zero to OpenEmbedded) Jul 11 15:09:21 as you may know, I'm working on Debian Installer support Jul 11 15:09:28 I expect building X to be about 10% as annoying as doing the same with OE, if perhaps more time-consuming the first time Jul 11 15:09:51 ThibG: yep, I read :) Jul 11 15:09:55 ok :) Jul 11 15:10:39 so far it works quite good (can install the whole thing without problem), however, bootloader configuration isn't done atm Jul 11 15:11:23 and u-boot is the only option for now (although PaulFertser is working on Qi missing features ;)) Jul 11 15:11:41 ThibG: problems were: FS corruption the first time, root password not set (as in, not possible to log in at all), no /etc/network/interfaces by default (I copied the one from SHR-u, plus my modifications (different subnet needed) without problems), seemed unhappy with rootfs ro (in Qi parameter list) Jul 11 15:12:11 getting NFS up was a bit more of a pain, but that was more to do with Fedora than Debian Jul 11 15:13:00 FS corruption: odd ; root password not set: hm, okay, would not happen in d-i :) Jul 11 15:13:25 yeah.. relatively minor problems, and not too soul-destroying to fix Jul 11 15:14:12 (compared to OE's habit of forcing you to fix crazy unrelated problems to get your one small package to compile, and being separated from what's really going on by crazy scriptage) Jul 11 15:14:29 ThibG: which Qi missing features? Jul 11 15:14:52 multi-file u-boot image support Jul 11 15:15:51 ThibG: kernel + initramfs? Jul 11 15:15:55 yes Jul 11 15:15:59 ok Jul 11 15:26:25 ThibG: related, is this normal? "All rc.d operations denied by policy" whenever apt-get tries to start or stop a service Jul 11 15:26:41 according to what I read, this is normal for a chrooted installation, but not for a "real" one Jul 11 15:27:02 this isn't normal if you aren't in a chroot Jul 11 15:27:39 IIRC, it's a script installed by (c)debootstrap that prevent service launching from the chroot Jul 11 15:27:53 and this script is removed at the end of the bootstrap procedure Jul 11 15:31:00 hmm.. can I just remove it now? Jul 11 15:33:40 doesn't seem to be such a big deal Jul 11 15:33:48 * pespin flashing newer kernel+rootfs from build.shr-project.org Jul 11 15:44:41 pespin: dunno if that's a good idea :/ Jul 11 15:45:54 mrmoku, well, I had an old version, I was not using it atm :) Jul 11 15:46:08 mrmoku, why it's not a good idea? some borked things inside? Jul 11 15:46:26 pespin: last image from 7th has broken busybox without networking Jul 11 15:46:57 Weiss: I found why local builds do not work anymore... dunno yet how to fix though Jul 11 15:47:15 (and IIUC you dropped the idea of using OE for that anyway :P) Jul 11 15:47:35 mrmoku: yep to both (saw your message last night, this morning) Jul 11 15:47:59 if my Debian/native build strategy works, it'll be loads more comfortable for the kind of development I'm doing Jul 11 15:49:13 mrmoku, any plans on building newer images? :P Jul 11 15:53:27 btw guys, I'm just curious to know if the demonstration we did here in catalonia yesterday (1,5 milion people) arrived to your media ;) Jul 11 15:53:44 any news about DMA for the glamo? I wasn't home for a few days ... Jul 11 15:54:35 Flyser: not DMA, but gena2x found more tweaks Jul 11 15:55:25 like? :-) I am really curious for updates :) Jul 11 15:59:18 Flyser: i can do 640x480 memcpy to glamo at 32fps... Jul 11 15:59:28 wow Jul 11 16:00:04 Flyser: by default you can get 16fps. Jul 11 16:00:08 will that also work in the drm driver? Jul 11 16:00:27 Flyser: currently, this is fb-only. Jul 11 16:01:03 Flyser: and only possibility. but it works. Jul 11 16:01:45 Flyser: also i coded DMA without tweaks. Jul 11 16:02:23 Flyser: which has same tranfser speed to memcpy, but cpu load is 0%. Jul 11 16:02:39 mrmoku, in shr.wizard, when I press Finish button, nothing happens ;) Jul 11 16:03:31 Flyser: DMA i coded is separate demo of possibility. Jul 11 16:04:31 Flyser: not used in MMC of any of glamo drivers, just shows that dma is ok and without fps drop. Jul 11 16:04:45 Flyser: DRM/non-DRM shouldn't make a difference: the speed boost should apply either way Jul 11 16:04:46 how does it scale? can you do memcpy with 480x360 at 56fps or with 640x360 at 42fps? That would be interesting for video playback ... Jul 11 16:05:10 pespin: got a weak password? Either use something passwd considers strong or leave it empty Jul 11 16:05:13 nice Jul 11 16:05:19 pespin: someday dos1 is going to fix that Jul 11 16:05:21 Flyser: it's raw transfer speed. divide of you wish. Jul 11 16:05:31 pespin: (new image) already building Jul 11 16:05:52 Flyser: I'm going to try to do pixmap migration using DMA Jul 11 16:05:59 (that'll be DRM-only, though) Jul 11 16:06:12 Weiss: may be you try timings thing first? Jul 11 16:06:35 Weiss: with different timing, dma is not needed. Jul 11 16:06:51 Weiss: as it is actually slower. Jul 11 16:06:58 nice :-) what about xv using dma? that should make it possible to play videos smoothly on the freerunner Jul 11 16:07:25 Flyser: you can already play smooth 640x480x12 mpeg2 on framebuffer. Jul 11 16:07:39 who wants to watch a 12fps video? Jul 11 16:07:57 Flyser: 12fps is watchable. Jul 11 16:08:07 Flyser: most time is spend in decoding. Jul 11 16:08:07 ugh Jul 11 16:08:20 Flyser: not in glamo as before. Jul 11 16:08:35 Flyser: i think with default timings, it is 8fps. Jul 11 16:08:50 Flyser: which is really unwatchable. Jul 11 16:08:52 gena2x: of course, I meant in the medium-term Jul 11 16:08:56 pespin: about media - when i'm home alone i don't watch tv at all ;) Jul 11 16:09:09 8fps? I remember of less than 5 Jul 11 16:09:13 but it was last year Jul 11 16:09:18 but if you would use dma, then mplayer could use 100% cpu to Jul 11 16:09:27 ThibG: 5 were with debugging :) Jul 11 16:09:29 dos1, ok, np, thanks anyway :D Jul 11 16:09:46 gena2x: but, freeing up the CPU is very important: X spends a huge amount of time migrating pixmaps, so anything that can be done to make it less resource-intensive gives a big overall performance boost Jul 11 16:09:55 pespin: oh, the first info on one of the biggest polish internet portals in category "World" :D Jul 11 16:09:55 decode the video and the copying to the glamo wouldn't take cpu time Jul 11 16:09:56 "Hiszpania: protestowało ponad milion Katalończyków" Jul 11 16:09:56 gena2x, oh, yes, debugging Jul 11 16:10:42 ("Spain: over milion catalonians demostrated") Jul 11 16:10:48 oh, great ^^ Jul 11 16:11:00 Weiss, Flyser : not so easy. as with dma even if memory bus is not completely blocked and you can still access memory, but memory access is still slow. Jul 11 16:11:58 Weiss, Flyser, ThibG: i did lmbench testing under constant dma transfer. most things are much slower. Jul 11 16:12:17 but work :) Jul 11 16:12:48 but its better then blocking the cpu completely Jul 11 16:12:50 gena2x: btw, whats wrong on internal Glamo MPEG4 decoder? sw decoding is faster? Jul 11 16:12:51 ? Jul 11 16:13:21 martix: nothing wrong. but i am speaking not about video decoding in fact. Jul 11 16:13:43 martix: what he does is to increase speed of transfering data to glamo Jul 11 16:13:48 gena2x: ok, its benchmark :-) Jul 11 16:13:50 martix: but about all glamo operations in fact. Jul 11 16:14:11 martix: good test for that is playing video - with acceleration it doesn't make sense ;) Jul 11 16:14:31 martix: and also we have no working glamo driver with mpeg4 acceleration. Jul 11 16:15:11 martix: for recent kernels. Jul 11 16:15:41 btw, mpeg4 speed may be also increased this way :) Jul 11 16:18:23 mrmoku, hmm usb networking is working for me Jul 11 16:23:45 gena2x: hmm.. how much slower? Jul 11 16:24:18 Weiss: 20% or so. Jul 11 16:25:26 Weiss: only with changed timings. with default - ~2% slower. Jul 11 16:25:42 Weiss: you can test easily yourself :) Jul 11 16:26:06 interesting.. Jul 11 16:26:44 if I understand the previous experiments (~2007), the transfer used to be 50-100% slower with CPU jobs virtually locked out Jul 11 16:26:52 pespin: really? which image? Jul 11 16:27:16 Weiss: not really. i have idea why - dma transfer is 4-word only, while cpu cache can do more. Jul 11 16:27:46 Weiss: so, then you not using page mode - you see almost no change. Jul 11 16:28:07 mrmoku, shr-full-eglibc-ipk--20100707-om-gta02.rootfs.jffs2 uImage-2.6.32.13-oe3.2+gitr6+a9254be10ac2294ea20165a87c09ea6afcf66d94-r0-om-gta02.bin Jul 11 16:28:07 Weiss: sdram burst is 4words too. Jul 11 16:28:18 mrmoku, already upgraded from the repo and all went fine ;) Jul 11 16:28:27 Weiss: in both cpu and dma. Jul 11 16:28:45 gena2x: no change from what? Jul 11 16:28:51 mrmoku, only thing I did before was chaning the subnet to 2 instead of 0 (I use that subnet for my fr at home ;)) Jul 11 16:29:32 Weiss: ok. 4 cases : dma/memcpy; default/page mode. Jul 11 16:29:59 pespin: hmm... interesting... I tried the lite one and it did not work Jul 11 16:30:34 Weiss: in default, dma speed == memcpy speed. dma is 4word burst, memcpy is done by cpu of course. Jul 11 16:30:37 pespin: will retry Jul 11 16:30:59 mrmoku, ok :) Jul 11 16:31:29 Weiss: in page mode, dma speed == ~0.8 memcpy speed. may be 0.9 i can't recally exactly. Jul 11 16:31:49 Weiss: but in all modes, something is readed from sdram and written to glamo. Jul 11 16:32:08 Weiss: sdram may be read only by 4 words burst is far as i understand. Jul 11 16:32:38 Weiss: in case of cpu data is read to dcache. Jul 11 16:32:50 Weiss: in case of dma to some dma buffer. Jul 11 16:33:28 Weiss: in case of default, data is written word-by-word (divided to 16bit). Jul 11 16:33:52 Weiss: in case of page mode data is written by page. Jul 11 16:34:18 Weiss: so if page size if 16, it is 8 words in 16bit words. Jul 11 16:34:31 Weiss: but dma can only do 4 words. Jul 11 16:34:44 Weiss: this is how i understand now. Jul 11 16:40:36 Weiss: emmm. i am clean? Jul 11 16:41:06 Weiss: or all this obvoius? Jul 11 16:46:32 JaMa, it looks like default locale in catalan lang in OE is set to ca-it, but it should be ca-es. Any idea on where to change that? Jul 11 16:51:28 PaulFertser: ping. Jul 11 16:57:22 JaMa: what about building SHR toolchain? seems that everything needed we can just copy from task-openmoko-toolchain-* and change glibc to eglibc Jul 11 16:57:51 or probably better, ${LIBC} (or something like that ;]) Jul 11 17:04:25 Weiss: don't you know is it possible somehow to get glamo docs? Jul 11 17:05:08 Weiss: sign NDA, is it possible from Russia? is it possible via mail? Jul 11 17:06:01 Weiss: is it possible at all? Jul 11 17:09:39 ok, guys, /me is going to take things into his own hands Jul 11 17:09:49 let's do some C development :D Jul 11 17:10:57 it's little too host, but at least i won't watch third episode of The Big Bang Theory today - i won't to have something to watch later too ;D Jul 11 17:11:03 s/won't to/want to/ Jul 11 17:11:05 dos1 meant: it's little too host, but at least i won't watch third episode of The Big Bang Theory today - i want to have something to watch later too ;D Jul 11 17:11:20 s/host/hot/ Jul 11 17:11:21 dos1 meant: it's little too hot, but at least i won't watch third episode of The Big Bang Theory today - i won't to have something to watch later too ;D Jul 11 17:11:53 yeah, looking at those mistakes it's really too hot ;x Jul 11 17:12:10 gena2x: should be.. I did mine via scan+email Jul 11 17:12:24 dos1: apt is just not so smart :) Jul 11 17:12:46 gena2x: not sure who to talk to at the moment, thouhg.. someone else asked on OM-Community recently Jul 11 17:13:08 it's 2^5 C here ATM Jul 11 17:13:15 wolfspraul: ^ (NDA?) Jul 11 17:13:48 and i think i would prefer 2^4 :P Jul 11 17:16:00 if it had been as much colder in February as February was than today, it would have rained dry ice Jul 11 17:17:21 Weiss: damn. i forgot about my multimeter in fridge. Jul 11 17:18:05 hehe Jul 11 17:18:15 gena2x, do you actually have something to eat in your fridge? Jul 11 17:18:28 ThibG: :) Jul 11 17:19:25 ThibG: no, only fr and multimeter, as actually i am trying to work on fr now :) Jul 11 17:21:19 * dos1 had to put chocolate into fridge in order to eat it today... Jul 11 17:42:52 gena2x: how's the qi 533 b.loader coming? ;) Jul 11 17:43:30 badcloud_: i'm not doing qi, dos1 doing qi's. Jul 11 17:43:49 badcloud_: or PaulFertser also did some versions. Jul 11 17:44:13 gena2x: i'll ask you later about CAMDIVN Jul 11 17:44:29 dos1: you may ask now :) Jul 11 17:44:37 gena2x: now i'm doing something different :D Jul 11 17:44:44 dos1: k Jul 11 18:02:07 handy timesaving tip: set up NFS so that the FR's user maps to your normal dev user on the server, then you can "git clone" on the "real" machine, and compile on the FR Jul 11 18:03:21 Weiss: heh. setup nis ? :) Jul 11 18:03:54 Weiss: why not just use qemu then? ;) Jul 11 18:03:56 I just did the anonymous user mappy things Jul 11 18:04:12 works fine :) Jul 11 18:04:16 I build on FR only when I need to move around and don't have stable connection Jul 11 18:04:26 Weiss: also it sounds better to compile at host too, and run on fr transparently Jul 11 18:04:38 gena2x: sure, but then you're back to cross-compiler doom Jul 11 18:04:56 Weiss: apt-cross is not doom, but easy Jul 11 18:05:04 if you run Debian on your host Jul 11 18:05:13 Weiss: ahh. sure :) Jul 11 18:05:42 I don't trust any of the packages involved to build properly in a cross environment - being fed up with screwing with that kind of thing is my motivation right now Jul 11 18:05:50 I don't mind waiting for slow builds Jul 11 18:06:26 (and automake/autoconf on FR is sllloooowww...) Jul 11 18:07:10 dash helps a bit Jul 11 18:07:10 slow builds is ok. but sometimemes you getting _really_ slow builds :) Jul 11 18:08:18 hehe Jul 11 18:08:26 time is cheap, sanity is not Jul 11 18:08:29 gena2x: thanks Jul 11 18:08:37 do you know Paul's server? Jul 11 18:08:44 or ratherm FR-time is cheap, Weiss-sanity is not Jul 11 18:09:07 good point Jul 11 18:09:30 and my sanity was seriously beginning to be compromised.. Jul 11 18:10:26 or anyone else Jul 11 18:10:32 badcloud_: you'd have to wait for Paul. i recommend you to try dos1 qi's. link in community ml. Jul 11 18:10:54 gena2x: thanks. I've actually been testing for him at present Jul 11 18:11:16 running at 450mhz :) Jul 11 18:11:35 badcloud_: 465, 1.55V versions are really best choise from any point of view if works flawlessly. Jul 11 18:12:08 Weiss: hm. you may try distcc with 2 freerunners. Jul 11 18:12:15 gena2x: nice idea :) Jul 11 18:12:31 distcc with cross-compilers is easier :) Jul 11 18:12:33 Weiss: also are you using overclocking? :) Jul 11 18:12:41 hm...... Jul 11 18:12:48 lindi- is really right. Jul 11 18:12:52 very right. Jul 11 18:13:06 and not only easy, but also very sane. Jul 11 18:13:42 gena2x: the 465 have all provided poor results w/qtmoko v24 Jul 11 18:14:06 best case stuck at cli login or qpe startup, worst case kernel [anic Jul 11 18:14:08 *panic Jul 11 18:14:19 badcloud_: at which voltage? Jul 11 18:14:28 hold on Jul 11 18:14:46 1.5, 1.55, 1.6, 1.65, 1.7 Jul 11 18:15:04 gena2x: i've build 450 and 465 with all those voltages Jul 11 18:15:30 dos1: nice idea. Jul 11 18:15:47 dos1: i had to write script and build 100 u-boots too :) Jul 11 18:16:56 gena2x: http://pastebin.com/n1BJdX00 Jul 11 18:17:33 using a Nokia BL-6C battery btw Jul 11 18:18:34 badcloud_: all you can try other than this is 1.9 volt on 1.8V bus. but this not really good idea for long run. Jul 11 18:18:53 gena2x: or move proper variables to environment ;-) Jul 11 18:19:28 gena2x: I'm using 450-112-1.5-1.8 at present Jul 11 18:19:44 ...seems ok. should I be aware of anything Jul 11 18:19:47 ? Jul 11 18:20:29 martix: it's impossible to change clk2/clk3 while running from sdram. so at least 2 version :) Jul 11 18:21:08 gonna watch some people kick around an inflatable ball :) Jul 11 18:21:09 badcloud_: i thinks you running safiest possible version. Jul 11 18:21:14 gena2x: noice Jul 11 18:21:17 gotta go Jul 11 18:22:45 gena2x: how safe it 465/118 1.55/1.8 then? ;) Jul 11 18:22:58 s/it/is/ Jul 11 18:23:00 dos1 meant: gena2x: how safe is 465/118 1.55/1.8 then? ;) Jul 11 18:26:15 gena2x: u-boot must be loaded in SDRAM before it loads u-boot_env ? Jul 11 18:28:44 martix: i am not really 100% sure. i think it setup sdram before copying something to it. as it has only 1 sdram setup afair. Jul 11 18:29:25 s/as/and/ Jul 11 18:29:26 gena2x meant: martix: i am not really 100% sure. i think it setup sdram before copying something to it. and it has only 1 sdram setup afair. Jul 11 21:17:52 sigh, why is it I keep running into uber-strange issues with OE? Jul 11 21:19:46 zub: OE is a beast... sometimes :P Jul 11 21:19:50 > arm-oe-linux-gnueabi/bin/ld:.libs/libgdk_pixbuf-2.0.ver:2: syntax error in VERSION script Jul 11 21:19:57 heh Jul 11 21:20:06 I think I figured out WHAT is happening, but I seldomly figure out why Jul 11 21:29:07 | arm-oe-linux-gnueabi-libtool: link: /usr/bin/nm .libs/gdk-pixbuf.o .libs/gdk-pixbuf-animation.o ... Jul 11 21:29:13 | /usr/bin/nm: .libs/gdk-pixbuf.o: File format is ambiguous Jul 11 21:29:14 ... Jul 11 21:29:30 this produces broken linker version script Jul 11 21:30:11 why does libtool invoke /usr/bin/nm instead of arm-oe-linux-gnueabi-nm? Jul 11 21:34:11 zub: something in your env? Jul 11 21:34:32 I have problems when having stuff like CPATH set to something... maybe same accounts for LD_LIBRARY_PATH? Jul 11 21:35:17 let me see, but I thought bitbake/OE pretty much kills all env vars with some exceptions Jul 11 21:36:16 anyway I "solved" it for now: ln -s .../tmp/cross/bin/arm-oe-linux-gnueabi-nm /usr/bin/nm Jul 11 21:36:21 how "nice" :) Jul 11 21:36:29 :P Jul 11 21:39:46 hm, env looks ok Jul 11 21:50:17 looking into sysroots/x86_64-linux/usr/bin/arm-oe-linux-gnueabi-libtool I see: NM="arm-oe-linux-gnueabi-nm" Jul 11 21:50:39 why, oh why... Jul 11 21:50:48 did it invoke /usr/bin/nm Jul 11 21:56:29 Sucks. Jul 11 21:57:21 My 3 favorites (had to compromise because of premature knock-outs) lost :P Jul 11 21:57:47 Lucky for me, I don't gamble :P Jul 11 21:57:53 Otherwise I would have lost all of my money :P Jul 11 21:57:57 night. Jul 11 21:58:04 night Jul 11 21:58:49 gena@work:~$ sudo dfu-util -a u-boot -D /tmp/qtmoko-debian-v24.jffs2 Jul 11 21:58:49 dfu-util - (C) 2007-2008 by OpenMoko Inc. Jul 11 21:58:50 ... Jul 11 21:58:50 File seems a bit too small... Jul 11 21:59:04 ^^^ nice error Jul 11 22:02:01 precise timing. he changes his nick back at 8:00am and I'll think he's a cron script :) Jul 11 22:03:18 hmm... may be we can use it to syncronize clock. Jul 11 22:29:21 <[Rui]> hi again, but g'night ye all :) Jul 11 22:29:28 * [Rui] lurks into bed Jul 11 22:34:27 another funny issue: | ../../../src/glsl/pp/libglslpp.a: could not read symbols: Archive has no index; run ranlib to add one Jul 11 22:35:16 if I just run arm-oe-linux-gnueabi-ranlib on the file and run compile again, it passes Jul 11 22:35:26 but why do I have to run it by hand? **** ENDING LOGGING AT Mon Jul 12 02:59:57 2010