Previous day | Jump to hour: 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | Next day

Seconds: Show Hide | Joins: Show Hide | View raw
Font: Serif Sans-Serif Monospace | Size: Small Medium Large

Click in the nick column to highlight everything a person has said.
The Logo icon identifies that the person is a core developer (has commit access).

#rockbox log for 2011-11-06

00:03:58 Join webguest [0] (
00:06:14 Quit webguest (Client Quit)
00:06:56bluebrotheramazingly long visit
00:11:41TheSevenjhMikeS: I just don't get any clicking sound at all while USB is connected
00:12:04TheSevenhow did you manage to get one?
00:12:39jhMikeSTheSeven: Just turn on keyclick. Connect USB storage. Press some keys.
00:13:41TheSeventhat's what I did. no sound at all.
00:14:32jhMikeSI'd guess that means there is some interference between PCM and USB storage on that target afterall.
00:15:03TheSevenwell, yeah, but I'd guess at a higher level
00:16:15jhMikeSmost likely, unless some later action by the driver is culprit, one that isn't hit unless storage is fully running
00:17:20*TheSeven just triggered a usb lockup without playing any audio
00:26:02JdGordon|TheSeven: are you nuking oyur .rockbox dir every reinstall?
00:30:18TheSevenJdGordon|: not on every reinstall, but often enough
00:31:28 Join ageis [0] (
00:33:48 Join Keripo [0] (
00:34:34TheSevenemcore install/uninstall/update testing, FTL testing, ...
00:35:18JdGordon|ok, fair enough
00:35:24TheSeven=> lots of reformatting going on
00:35:42JdGordon|you might want to just change the default value in settings_list.c for the show files setting if that annoys you
00:35:45TheSevendoes that build click on your ipod while USB is connected?
00:36:52jhMikeSTheSeven: I can't test that. The only iPod is have is a 3g with a pp5002
00:37:29TheSevenah, so "Just FYI: Keyclick while USB is connect seems to be fine on the targets I own. I suppose it should be expected to work." refers to non-nano2g devices, which explains a lot
00:37:45JdGordon|jhMikeS: thats a hard disk target right? any chance you could test a patch for me?
00:38:55jhMikeSTheSeven: that's what I meant. I did mention in that task that I didn't have the apropriate hardware :)
00:39:08jhMikeSJdGordon|: sure, specifically on that one?
00:39:13 Join Reimu_Hakurei [0] (~znc@2607:f358:1:fed5:22:0:8c2c:6b65)
00:39:59JdGordon|jhMikeS: yes, I'm trying to see if i managed to speed up .cfg loading
00:40:09JdGordon|do you want me to build for oyu?
00:40:32jhMikeSwould probably be faster if you did
00:42:10jhMikeSit seems to be badly in need of a charge too :\
00:43:25TheSevenjhMikeS: making sure usb_core_control_request will never get called also helps
00:43:56TheSevenso the USB interrupt handling code doesn't seem to be directly involved here
00:44:11TheSevenpossibly a high rate of interrupts during lots of USB traffic?
00:44:40 Quit Reimu_Hakurei (Quit: Baibai.)
00:45:38jhMikeSwhy should that be an issue though? a measly 173 ints/sec from PCM?
00:47:32TheSevencommenting out the call to usb_core_transfer_complete in the IN EP IRQ handler doesn't help!
00:47:37jhMikeSthe fact it doesn't start at all once storage is active should be telling
00:48:11TheSevenso receiving a control request (but not passing it to our code) doesn't cause the crash, but sending the response does!
00:48:26JdGordon|jhMikeS:"> and on both please note the number that gets splashed at boot, and also try manually loading a cfg (preferably a full one saved from the setting menu) and not the number again
00:48:39*TheSeven double checks
00:49:00jhMikeSJdGordon|: oh, now it's TWO tests :)
00:49:29TheSeventhe behavior is consistent so far...
00:49:33JdGordon|4 technially :) 2 builds
00:49:58TheSevenbtw, one more thing that I dislike: keyclick repeat
00:50:14jhMikeSTheSeven: well, turn it off! :)
00:50:26JdGordon|keyclick repeat needs to be fixed to only beep the first repeat
00:50:41TheSevenyou have the option to either just get a single click when scrolling around in a list (doesn't make any sense to me), or getting a really nasty sound when trying to e.g. power off the ipod
00:50:47TheSevenor enter charging mode
00:50:57jhMikeSI disagree if that involved the wheel or something where hearing every step is important
00:51:36TheSevenyeah, i want to hear every step of the wheel, but not every typematic-like event of a button that's being held down
00:51:48TheSevenand that's apparently not possible currently
00:51:51jhMikeSJdGordon|: should I test the unmodified build too or just the provided ones?
00:52:04JdGordon|just the provided 2
00:52:15JdGordon|the second one with svn with the splash added to get a baseline number
00:52:19JdGordon|TheSeven: sure it is
00:52:22 Quit ageis (Quit: XChat 2.8.8 Linux 3.0.0-1-686-pae [i686/2.99GHz])
00:52:37*jhMikeS doesn't even know if newer builds still run on that target at all
00:53:51JdGordon|TheSeven: keyclick happens in action.c so just check the button value
00:54:08 Join Reimu_Hakurei [0] (
00:54:40TheSevenJdGordon|: I meant with the current code, from a user's perspective
00:55:03TheSeventhe options we're currently providing just don't make sense for clickwheel-based targets
00:55:14 Quit lebellium (Quit: ChatZilla 0.9.87 [Firefox 8.0/20111102223350])
00:55:16JdGordon|100% agree
00:55:39JdGordon|i dont see why we need a setting for it though
00:56:18TheSeveni'd be perfectly fine with a sane default behavior, but we should at least provide a setting for a sane behavior, and not just the two useless options :)
00:56:56JdGordon|It shouldnt be hard to fix the repeat behaviour to be not moronic :)
00:57:00jhMikeSarg, wtf. hope I don't have to reboot. it seem to be having trouble connecting on the firewire
00:58:02jhMikeSguess I'd better
00:59:59TheSevenmessing with the usb send code made it end up with an ATA error 1 while connecting to USB
01:00:24TheSevenJdGordon|: I'd actually consider this a release blocker
01:00:51jhMikeSI guess it was a problem of having too much crap open :\
01:01:10JdGordon|TheSeven: its not new since 3.9 so good luck! :)
01:01:36TheSevenhm, I never realized it was *that* broken
01:02:26TheSevenpossibly because I usually have the hw keyclick patch applied which seems to fix the most annoying parts of it
01:02:59JdGordon|hwclick is just as bad
01:04:54CIA-22New commit by theseven (r30905): usb-s3c6400x.c: Remove an unused variable
01:05:49jhMikeSwell, I get a "0" at boot with "patched"
01:06:58CIA-22r30905 build result: All green
01:07:06jhMikeSwith patched2 a "1"
01:07:39JdGordon|jhMikeS: did you try loading a .cfg manually also?
01:07:55JdGordon|and i guess your config.cfg is pretty empty?
01:09:04jhMikeSit probably is pretty empty. so far, I can't do anything manually since the batt is so low it shuts down and I don't think there's a "charging only" mode on 3g :\
01:09:19JdGordon|ah nuts
01:10:12jhMikeScould try x5 or h10 5GB, or gigabeat F/S, H100 as well
01:10:20***Saving seen data "./dancer.seen"
01:10:38jhMikeSthe battery on my 20GB H10 is completely useless now
01:10:39TheSeven*PANIC* Stack underflow... viewport manager
01:10:44JdGordon|anything with a hard disk would be great :)
01:10:55JdGordon|TheSeven: good one! how'd you manage that?
01:11:13TheSevensome combination of unplugging USB while holding the power buttion
01:11:16JdGordon|TheSeven: wanna try ?
01:11:20TheSevener, play button, attempting to power it off
01:11:32jhMikeSJdGordon|: well, take your pick from the list...I haven't the patches anyway
01:11:55jhMikeSH100/x5 would probably show it best
01:12:01JdGordon|jhMikeS: the patch changes read_line() to fast_readline() and does some other smartness to hopefuly speed cfg loading up
01:13:00jhMikeSyou know, I don't get why the e260 battery seems practically new after so many years but everyone else's seem to go flat in just a few
01:13:02JdGordon|I dont have the coldfire gcc :(
01:13:23jhMikeSok, the gigbeat f? It has a fairly slow hard drive
01:13:53*TheSeven could test on a classic with a crippled hard drive
01:14:07JdGordon|how abot i put the patches online? :)
01:14:36jhMikeSgo ahead. I though you had like some 6000 core machine or something :)
01:14:55TheSevenjhMikeS: is mixer_reset() irq-safe?
01:14:57JdGordon|yeah, but really slow internet
01:15:11JdGordon|I tried installed sh-elf earlier in the week and gave up after na hour of downloading
01:15:23jhMikeSTheSeven: I don't believe it is
01:16:07*TheSeven tries to move that around to find the spot that needs to happen before it to trigger the crash
01:16:20jhMikeSnone of the weightier mixer routines were meant to be
01:16:41JdGordon|jhMikeS:"> is the first one
01:16:52JdGordon|please manually load a .cfg from the browser/menus also
01:16:58TheSevennow this is odd
01:17:33TheSeveni made usb_drv_send call mixer_reset() on its first invocation
01:17:44TheSevenand it seems to completely fix the issue
01:18:00TheSeveneven if i try it multiple times
01:18:53JdGordon|jhMikeS: and"> is the other one
01:18:53jhMikeSJdGordon|: and the second?
01:19:12TheSeventhis is driving me nuts...
01:20:40*TheSeven has a suspicion
01:25:34TheSevenhm, nope, seems like purely random behavior
01:26:59 Quit keyb_gr (Ping timeout: 240 seconds)
01:27:35TheSeventhis smells more and more like memory corruption
01:28:17 Quit BHSPitMonkey (Read error: Connection reset by peer)
01:29:09jhMikeScache line stuff maybe?
01:29:52TheSevennah, much worsse stuff
01:30:27TheSevenadding an udelay() in the usb code caused the lcd to slowly fade to white when plugging USB
01:31:16jhMikeStry adding things of specific sizes, like nops
01:33:40TheSeventhese really odd things aren't reproducible
01:34:26jhMikeSso, it's not every time with a specific mod?
01:34:45TheSevenadding this code into usb_drv_send fixes it, even for multiple usb connections:
01:34:46TheSeven if (!hit) mixer_reset();
01:34:46TheSeven hit = true;
01:35:15TheSevenadding the same code into usb_drv_init fixes it only for the first connection
01:35:29TheSevennow what the hell does that mean?
01:36:05TheSevenlogical conclusion: the "hit" variable is being zeroed by something in between
01:36:15TheSevenit's probably close to a USB transfer buffer in memory
01:36:23 Join Sethman [0] (
01:36:31jhMikeStransfer buffer is cache aligned?
01:38:22 Quit Sethman (Client Quit)
01:38:31pamaurydoes this usb problem only happen on the nano2g ?
01:39:15TheSeventransfer buffer is USB_DEVBSS_ATTR, whatever that means
01:40:22jhMikeSthat just allows each target to allocate it as what is best for it
01:40:29TheSevenok, i think i have evidence of memory corruption
01:40:29pamauryI seem to remember that we were very suspicious about memory corruption on the clip+ with the same usb controller
01:40:39TheSevenmoving the "hit" variable to iram breaks it
01:41:10TheSevenpamaury: you mean that vector corruption bug?
01:41:36TheSeveni think we got rid of that by pointing its dma address registers into nowhere when doing zero-byte transactions
01:42:03TheSevenfirmware/export/config/ipodnano2g.h:#define USB_DEVBSS_ATTR __attribute__((aligned(16)))
01:42:08TheSeventhat seems to be safe
01:42:09pamauryno, memory close to transfer buffers, like if the dma was writing more than needed
01:42:32TheSevenwhich would be kinda similar to "dma writing something when 0 bytes are needed"
01:42:33pamaury(which can be the case, it can write up to 3 bytes more iirc)
01:43:02pamaury(or more if doing burst ?)
01:43:27jhMikeStry making the buffer bigger (just to check)?
01:44:33JdGordon|TheSeven: I tihink I've fixed keyclick
01:45:18JdGordon|yep :)
01:45:48pamauryPerhaps you can try to disable dma bursts for this controller ? I think there are a register bits for this usb controller
01:47:08TheSevenuh oh, I think I know what's going on
01:47:24jhMikeSJdGordon|: (on GigF) Settings1: boot - 0, load - 1
01:48:13TheSeventhat explains the usb to pcm relationship
01:48:25TheSevenctrlreq is probably overflowing
01:49:01JdGordon|TheSeven: I merged my fix with fs#5111 if you want it
01:49:02fs-bluebot Ipod piezo driver (patches, unconfirmed)
01:51:39 Join chrooted_mother [0] (
01:51:53jhMikeShrm...same reading with the second patch
01:51:55chrooted_motherI just installed rockbox on my Nano 2G. Thanks guys!
01:52:01JdGordon|jhMikeS: damn, ok thanks
01:52:34jhMikeSmaybe i should make a ridiculous config? can I dump even non-defaults somewhere?
01:52:49chrooted_motherusing the current build as well. any cool tricks besides playing doom?
01:52:49jhMikeSer defaults + non-defauls
01:53:16 Quit domonoky (Read error: Connection reset by peer)
01:53:21 Quit bertrik (Read error: Operation timed out)
01:53:23JdGordon|jhMikeS: settings2.diff the values dont matter, dumping a default config and loading it should give a pretty close to 0 time on 1 and large on 2
01:53:57jhMikeSI'll try the settings dumper output
01:54:16JdGordon|no, jsut save a config
01:55:05JdGordon|anyway, that patch wont reload the themes (causing a audio restart) if it doesnt change so its worth doing anyway
01:55:16JdGordon|I'm surprised that readline isnt slower than fast_readline houhg
01:55:36JdGordon|is there disk block caching?
01:55:39 Quit Keripo (Quit: Leaving.)
01:56:08pamauryTheSeven: indeed, 0 bytes of padding seems dangerous with this controller :)
01:56:12 Quit n1s (Quit: Ex-Chat)
01:56:21jhMikeSTheSeven: wait, what am I supposed to spot?
01:56:27CIA-22New commit by theseven (r30906): usb-s3c6400x.c: Fix memory corruption bug (closes FS #12310)
01:57:24jhMikeSguess I'll see in a second
01:57:46pamauryTheSeven: that's a rather hacky fix ! ;)
01:57:54TheSevenjhMikeS: the PCM driver's global variables were trashed by a memory corruption bug in the USB code
01:57:59CIA-22r30906 build result: All green
01:58:30TheSevenpamaury: well, that was the least hacky variant that I could think of :)
01:59:05funmanTheSeven: can you add a comment about that 64 bytes?
01:59:05pamauryit would be interesting to understand why it "overflows" and of how much
01:59:24pamauryand see if it can be fixed
01:59:36TheSevenpamaury: it technically doesn't overflow
01:59:54TheSeventhe maximum size for a USB control request is 64 bytes
02:00:03jhMikeSJdGordon|: according to the 3g, the more modded version was faster
02:00:10TheSevenwe only use the first 8 bytes because we don't support any commands that need more
02:00:32pamauryah, the dma is setuped to received 64-bytes ?!
02:00:42TheSevenso our struct only contains those 8 header bytes, but those can in theory be followed by payload data
02:00:48TheSevenand windows probably sends some "write descriptor" or something command
02:01:30TheSevenso the proper fix would probably to add the missing bytes to the struct itself, but I don't really want to touch that before 3.10
02:02:16 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
02:04:20TheSevenjhMikeS: thanks for making that nasty bug surface :)
02:04:49TheSeventhis could have (or even did?) caused much more subtle and intermittent trouble
02:05:55*jhMikeS was always considered a trouble-maker, but never knew it would be so valuable later in life :)
02:06:03pamauryit might have, and might have caused some bugs on the clip+ too (for those who tried the usb patch)
02:07:38JdGordon|jhMikeS: great, though not really a difference worht worrying about?
02:07:43JdGordon|that number is ticks
02:07:55funmanpamaury: tell me you'll fix AMSv2 USB soon :)
02:08:58pamaurynot sure, we might have made some progress with the OF voltage settings that bertrik found and I'm still puzzled by the fact that with sd inserted it works reliably where is fails randomly without
02:09:18pamaurythis "bug" could explain some things but not all
02:09:21jhMikeSJdGordon|: loading that the big cfg only took two ticks on the F, not like anyone would notice. one config loader that should be better is the one used by plugins. it makes mpegplayer slow to switch between files.
02:11:54jhMikeSit always has to save the last resume but the disk thrashes around while it does it
02:12:48funmanpamaury: do you test with windows or linux?
02:13:04pamauryI tested with both
02:13:06JdGordon|thats completly different code though isnt it?
02:13:30jhMikeSJdGordon|: in the plugin lib iirc
02:14:43pamaurywith windows is always less reliable and does some lockup on enumeration quite often (that might have to do with this bug); but with linux it usually doesn't work unless you have sd inserted (in my case and last time I tried)
02:16:16jhMikeSJdGordon|: it uses readline/parseline to read but fdprintf to write
02:19:11JdGordon|it reads and writes at the same time?
02:24:18 Join Keripo [0] (
02:26:54jhMikeSif modifying a line
02:27:25jhMikeSfor some reason, the settings dumper config failed to properly load as a cfg (should it?)
02:28:20jhMikeSit loaded but some settings like the backlight brightness were some crazy value
02:30:06jhMikeSha, I see, it give mix/max/step but doesn't actually provide values
02:30:22JdGordon|i dont tihn it is supposed to write something loadable
02:30:33JdGordon|anyway, im heading out to enjoy the day
02:34:12jhMikeSTheSeven: It's curious that dblbuf is nothing more than char aligned in the PCM driver
02:36:43pamauryTheSeven: do you plan to add a comment about the 64-bytes in the usb driver ?
02:49:15CIA-22New commit by pamaury (r30907): sbtools: rename to imxtools, move imx_hid_recovery to imxtools/sbloader, fix tools to correctly handle/free memory, properly return error codes
02:50:49CIA-22r30907 build result: All green
02:51:36jhMikeSare the imxtools for all imx?
02:53:54pamauryI think so, although they don't handle things of the more advanced ones
02:54:23pamauryAt least they should handle the imx2y and imx5y for which I'm pretty sure it's the same format
02:55:30pamaurybased on the freescale sbloader, that should work on mx23/28, mx50/6x
02:56:19 Quit GeekShadow (Remote host closed the connection)
02:57:23 Quit Reimu_Hakurei (Ping timeout: 240 seconds)
03:01:27 Quit pamaury (Remote host closed the connection)
03:04:27 Quit chrooted_mother (Ping timeout: 240 seconds)
03:10:22***Saving seen data "./dancer.seen"
03:11:20 Join chrooted_mother [0] (
03:38:48chrooted_motherwhat can rockbox do? i installed it and I know about the codecs, I use FLAC and needed to read flac on my ipod
03:38:53chrooted_motherbut are there any cool tricks?
04:01:44 Join guymann [0] (
04:06:20 Quit TheSeven (Disconnected by services)
04:06:46 Join [7] [0] (~TheSeven@rockbox/developer/TheSeven)
04:11:57 Join ageis [0] (
04:17:15 Quit amiconn (Disconnected by services)
04:17:16 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:17:21 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:18:15 Quit pixelma (Disconnected by services)
04:18:17 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:18:19 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:28:00 Join [Saint] [0] (~Saint]@unaffiliated/saint/x-8516940)
05:01:04 Join Reimu_Hakurei [0] (
05:01:43 Quit robin0800 (Quit: Leaving)
05:10:23***Saving seen data "./dancer.seen"
05:21:00 Join Rob2223 [0] (
05:24:32 Quit Rob2222 (Ping timeout: 245 seconds)
05:27:11 Join Rubyiris [0] (
05:33:06 Join robin0800 [0] (~robin0800@
05:33:31Rubyiriscirca-2009), Rockbox has been reporting the wrong battery life, and my battery life has dropped significantly. Less than that of the original Sansa firmware.
05:36:04 Quit chrooted_mother (Ping timeout: 240 seconds)
05:43:38 Quit Rubyiris (Quit: CGI:IRC (Ping timeout))
05:51:26 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
06:07:04 Quit zchs (Ping timeout: 252 seconds)
06:25:44 Join zchs [0] (
06:34:55 Quit robin0800 (Quit: Leaving)
07:10:27***Saving seen data "./dancer.seen"
07:25:29 Quit BHSPitMonkey (Remote host closed the connection)
08:23:05 Quit ageis (Quit: laters)
08:31:17 Join Horschti [0] (~Horscht@xbmc/user/horscht)
08:34:45 Quit Horscht (Ping timeout: 260 seconds)
08:40:53 Join bertrik [0] (
08:40:53 Quit bertrik (Changing host)
08:40:53 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
09:04:06 Quit Keripo (Read error: Connection reset by peer)
09:04:48 Join Keripo [0] (
09:06:48 Join stoffel [0] (
09:10:30***Saving seen data "./dancer.seen"
09:28:56 Join n1s [0] (
09:28:57 Quit n1s (Changing host)
09:28:57 Join n1s [0] (~n1s@rockbox/developer/n1s)
09:32:35 Nick Guest45092 is now known as froggyman (~seth@
09:32:43 Quit froggyman (Changing host)
09:32:43 Join froggyman [0] (~seth@unaffiliated/froggyman)
09:59:59 Quit [Saint] (Read error: Connection reset by peer)
10:00:14 Join [Saint] [0] (~Saint]@unaffiliated/saint/x-8516940)
10:10:44 Quit [Saint] (Read error: Connection reset by peer)
10:10:47 Join [Saint_] [0] (~Saint]@unaffiliated/saint/x-8516940)
10:12:12 Join caramou1 [0] (
10:24:18 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
10:27:53 Quit [Saint_] (Remote host closed the connection)
10:32:56 Join chrooted_mother [0] (
10:35:57 Join [Saint] [0] (~Saint]@unaffiliated/saint/x-8516940)
10:49:53 Join ender` [0] (
10:53:43 Quit bluebrother (Disconnected by services)
10:53:45 Join bluebrother^ [0] (~dom@rockbox/developer/bluebrother)
10:55:40 Quit fs-bluebot (Ping timeout: 240 seconds)
10:57:20 Join fs-bluebot [0] (
11:01:34 Quit mystica555 (Quit: Ekkusu Chatto)
11:06:38kugelalright, I think I've fixed voice
11:07:25[Saint]What actually broke it?
11:09:00kugel[7]: "unsigned char data[sizeof(struct usb_ctrlrequest)]" would also work wouldn't it?
11:10:27kugel[Saint]: several negligences
11:10:31***Saving seen data "./dancer.seen"
11:11:25[Saint]Oh, fancy that, in Rockbox too of all places ;)
11:11:30JdGordon|kugel: why not just make data a sturct usb_ctrlrequest* ?
11:12:04 Quit chrooted_mother (Ping timeout: 240 seconds)
11:16:15 Join TheLemonMan [0] (
11:19:27kugelJdGordon|: that was before
11:19:36kugelthat caused memory corruption
11:19:58JdGordon|where is the naughty code?
11:20:17JdGordon|is this more bad pointer arithmatic?
11:20:32kugelsee r30906
11:21:06[Saint]There's no such thing as a bad pointer, just good pointers with a bad upbringing.
11:21:40kugelamiconn: ping
11:21:42JdGordon|what am I looking at?
11:21:55JdGordon|is this completly unrelated to buflib?
11:23:42JdGordon|oh ok
11:31:41kugelenglish.voice is getting really close to audio buffer size on hwcodec, on the sim at least
11:32:21kugelonly 10k room
11:33:00JdGordon|doesnt it keep a seek table if its too big so the whole file isnt loaded?
11:33:13kugelnot on this one
11:33:34kugelwell it does on all targets
11:33:43JdGordon|seriously time to consider forking hwcodec
11:33:43kugelthe voice file has a seek table
11:34:03kugelbut only on lowmem swcodec the case where a voicefile doesnt fit is actually handled
11:34:24JdGordon|wow, really?
11:34:33JdGordon|I thught that was there specifically for hwcodec?
11:34:56kugelyou thought wrong :)
11:38:15 Join chrooted_mother [0] (
11:43:33 Join Buschel [0] (
11:44:02BuscheljhMikeS, [7]: thanks for fixing the nano2g USB issue −− works as expected now
11:45:23[7]kugel: no, that wouldn't work, because the size of that struct being too small is the root cause of the problem :)
11:46:19[7]jhMikeS: why should the dblbuf be more than char aligned?
11:47:23Buschelbut: I again see unexpected re-buffer activities on the nano2g. use case: buffering is finished, then skipp to the next song. often I see flash read-accesses even though everything has been buffered −− those rtead-accesses have a short duration (so, they do not re-fresh the full buffer)
11:47:50[7]that's probably another bug :)
11:50:14kugel[7]: ah ok, I thought it was an alignment problem
11:50:38Buschel[7]: yes, looks like FS #12309 strikes back
11:50:38fs-bluebot font buflib: mass storage activity when skipping buffered tracks? (bugs, closed)
11:52:03[7]kugel: the proper fix for this would probably adding 56 payload bytes to the struct - or possibly renaming the struct to usb_ctrlreq_header or something, and wrapping that into another struct which will get used in this place
11:55:50[7]Buschel: can't reproduce that with the couple of testing mp3s that are currently on my nano2g
11:59:26bertriktbh, I see lots of little weird stuff (including mass storage activity when I don't expected), but in the past week I've been using mostly the clip zip, which is so new that I guess I can't really report bugs (although I strongly suspect these are general bugs, not target-specific ones)
12:02:12*user890104 sees [7]'s nickname and remembers about FS #12233
12:02:12fs-bluebot iPod 6G/Classic: LCD not displaying rockbox correctly if using emCORE >= r746 (patches, unconfirmed)
12:03:01[7]aw right....
12:03:15[7]another thing that needs to go in before 3.10
12:04:21[Saint]It does?
12:04:31[Saint]Is the classic being promoted?
12:06:39user890104[7]: is the so-called "usbfix" in the svn, or it was merged in your changes from last night?
12:07:58[7][Saint]: I don't think so, but I want to have a somewhat consistently working state in the 3.10 branch
12:08:01Buschelbertrik: I could not reprodcue this on my iPod Video as well. somehow I suspect this is depending on the theme
12:08:56[Saint]Well...theme affecting the core in ANY way is a bug for sure.
12:09:02Buschelunneeded rebuffering or flash/disk access will reduce the battery runtime −− especially for the HDD targets where the spinup takes a lot of power
12:09:14[Saint]And not technically possible, as I understand it.
12:09:39[7]user890104: I'm a bit hesitant about the usbfix
12:10:01Buschel[Saint]: we had issues with the font cache that resulted in rebuffering and disk spin up
12:10:16[7]there are reports of a regression (that I can't reproduce) in flyspray, and I don't want to take that risk
12:10:58[7]Buschel: your theme on nano2g => can't reproduce
12:11:21[7]possibly the metadata of the files being played is related here? (assuming it's a font cache issue)
12:11:29Buschelhmm, you are using enmbedded album art?
12:11:32[Saint]Buschel: there was a specific theme that managed to screw up playback order a whole ago too iirc.
12:11:40 Quit chrooted_mother (Ping timeout: 240 seconds)
12:11:50[Saint]*while ago
12:11:50[7]Buschel: yes
12:12:00Buschelok, me too
12:12:27[7]i have just 4 aac files on my ipod right now
12:12:27JdGordon|[Saint]: ah yes, there was some truly wierd shit happening back then. I dont remember what the fix was
12:12:59[Saint]gevaerts found it I think.
12:13:07[Saint]Both the bug and the fix.
12:15:19 Join pamaury [0] (
12:15:32 Quit pamaury (Changing host)
12:15:32 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:15:39[Saint]Buschel: have you tried stripping the tags from your test files?
12:15:51[Saint](Just a thought)
12:17:45Buschel[Saint]: no, I am trying to bisect now
12:18:21*Buschel thinks we have too many regressions these days
12:18:27caramou1Hello, can someone add me (OlivierKaloudoff) to WikiUsersGroup (changed from CaraMou to OlivierKaloudoff as requested), would like to update a broken link for Sony port
12:18:37JdGordon|fire the QA team!
12:20:17[7]caramou1: looks like you're in that group
12:20:32[7]so someone apparently beat me to it :)
12:21:41[Saint]caramou1: I did that for you days ago...
12:22:26[Saint]First as CaraMou, then as your actual name when it was discovered that it wasn't your real name.
12:23:30caramou1something must be wrong then, cause I still get this message "Access check on Main.SonyNWA808Port failed. Action "CHANGE": access not allowed on web."
12:23:56caramou1trying to purge my browser settings
12:24:51[7]caramou1: fixed
12:24:51[7]there was a typo :)
12:24:52 Join Poodlemastah [0] (
12:25:32 Quit Poodlemastah (Client Quit)
12:25:48[Saint][7]: oh...whoops, what did I mess up? I was fairly confident I did I straight c/p
12:26:42[7]pretty subtle
12:26:53caramou1sniff still the same .. Still appers as OlivierKaloudouff in WikiUsersGroup
12:27:16[7]sure? (cache?)
12:27:32caramou1great ! perfect, thank you
12:30:16 Join Poodlemastah [0] (
12:33:58Buschel[7]: the rebuffer is *not* happening anymore when I have dircache switched on
12:34:23 Quit Poodlemastah (Remote host closed the connection)
12:34:41[7]that means that it's unlikely to be said font cache bug hitting back :)
12:35:22CIA-22New commit by theseven (r30908): iPod Classic: Set LCD controller configuration register during initialization and don't rely on the bootloader doing that (closes FS #12233)
12:35:51Buscheltrue, but the effect looks similar
12:36:56kugelerr, voice is even exceeding the audiobuffer already
12:37:09CIA-22r30908 build result: All green
12:37:16 Join liar [0] (
12:37:40[Saint]kugel: on which target(s?)
12:37:44kugelhwcodec (sim)
12:37:55[7]now that's odd: getting "All green here" before svn ci even terminates :)
12:38:01kugelenglish.voice is 1.7MB, audiobuffer is less than 1.5MB
12:38:02[Saint]Time to fork hwcodec :p
12:38:31[7][Saint]: and while we're at it, fork charcell as well :)
12:38:41kugelthat's automatic
12:39:14[Saint]All the charcells are hwcodec, no?
12:39:18kugelseriously, forking hwcodec/charcell/sh would allow for a *massive* cleanup, since much of the old and ugly code is just for those
12:43:07user890104[7]: while you're at it: FS #12209
12:43:07fs-bluebot iPod 6g/classic usb product id (patches, unconfirmed)
12:43:44[Saint] that a release required fix?
12:44:00[Saint]*s/were/errr /
12:44:24[Saint]Silly me thought we were in a freeze.
12:46:12kugelamiconn: how to handle this?
12:46:49kugel[Saint]: no but it also doesnt affect a release target
12:47:29[Saint]I didn't think that mattered.
12:49:35JdGordon|there is a forum thread which has turned to discusison about forking
12:49:49JdGordon|it should have happened around 3.6~
12:49:55JdGordon|or earlier even
12:50:11kugel[Saint]: non-release targets aren't affected by freeze, see recent clip zip work
12:52:15 Quit liar (Quit: hallowed are the ori!)
12:52:41 Join liar [0] (
13:01:39CIA-22New commit by kugel (r30909): Fix broken voice when .talk files are used and voice buffer calculation.
13:01:44CIA-22New commit by kugel (r30910): hwcodec: fix shrink_callback resuming playback wrongly, and add checks ...
13:01:52[Saint]Dragged from community:
13:02:37[Saint]JdGordon|: wouldn't the line height patch thing detecting the fallback case be the "right" way to do this?
13:02:49*JdGordon| thinks he needs to bite the bullet and rework the skin engine to only deal with offsets and a single buflib alloc, which grabs the full thing and resizes after load
13:03:04CIA-22r30909 build result: 0 errors, 198 warnings (kugel committed)
13:03:11JdGordon|[Saint]: I said ignore it because I'd fix that in the code
13:04:00JdGordon|which is also why I said one specific screen size
13:04:24JdGordon|I'm more interested in what can't be done currently than getting it perfect
13:04:27 Quit Keripo (Quit: Leaving.)
13:04:27CIA-22r30910 build result: 0 errors, 209 warnings (kugel committed)
13:05:15[Saint]Right. I'm with you now. What about icons? If the user theme fails to load, isn't it safe to assume that cabbie wont be there too?
13:05:33[Saint]Well...hmmm. no.
13:05:41JdGordon|do it as :)
13:05:58JdGordon|so its not so much the fallback as a flly loaded skin
13:06:21CIA-22New commit by theseven (r30911): Commit FS #12209 by Vencislav Atanasov: iPod 6g/classic usb product id
13:06:43[Saint]That just made me think...why do we go to the fallback if the user theme fails, as opposed to trying cabbie (if the user theme isn't cabbie)
13:06:56 Join GeekShadow [0] (
13:07:12JdGordon|cabbie isnt hardcoded
13:07:35CIA-22New commit by kugel (r30912): Fix warning and typos.
13:07:45[Saint]Right...but can't it check if all of cabbie is present and able to be loaded?
13:07:50CIA-22r30911 build result: 0 errors, 209 warnings (theseven committed)
13:08:06JdGordon|sure, maybe
13:08:46[Saint]If *that* fails...were definitely in trouble, so use the fallback :)
13:08:56JdGordon|also, I tihnk the viewport tag is going to need some extension to to relative alignment to other viewports
13:09:11JdGordon|and then magic to resize to X lines
13:09:24JdGordon|maybe not
13:09:32CIA-22r30912 build result: All green
13:09:35kugelsomeone needs to check voice on hwcodec, before 3.10 preferably
13:10:00[Saint]I've nfi how far he got, or with what, but I know zagor has a skin resizong branch...
13:10:34***Saving seen data "./dancer.seen"
13:10:44JdGordon|not so much skin resizing as screen resizing iiuc
13:11:04JdGordon|and svn branches suck ass, so wait for git to start looking into it :)
13:11:12JdGordon|fingers crosses is Soon :)
13:11:27kugelwith proper git-svn setup they're git branches :)
13:11:33[Saint]Right...or, relative positioning. Something, anyway :)
13:12:07JdGordon|but yeah, if you could start (and put it on github please) then I'd be really greatful
13:12:33*JdGordon| would love to nuke apps/gui/bitmap/list.c and make the list skin handling code nicer
13:12:42JdGordon|nicer, less crash prone, better
13:13:22JdGordon|then once the entire ui is drawn by the skinengine, rip it out and use something better :) sdl perhaps!
13:13:33JdGordon|(I kid)
13:13:39JdGordon|(or do I?)
13:15:01JdGordon|kugel: is there any management on things calling core_alloc_max()?
13:15:55kugelyou don't get anything with _maximum() if something else has the maximum
13:16:29JdGordon|I tink what I want to do is for each skin, alloc the entire buffer, load the skin file, shrink the buffer to the used size, then move onto the next one
13:16:37JdGordon|so obviously playback would have to be shutdown
13:16:52kugelbut you can do audio_buffer_available() and ask for a specific size in which case the shrink callback will be used
13:17:20JdGordon|maybe, maybe not... I want to move everything to offsets so could even load into the plugin buffer perhaps and memmove to a handle
13:17:40JdGordon|this is going to be a massive branch :/
13:18:11kugelit affects just a handful files no?
13:19:16*[7] wonders which openid provider he did use to register in gerrit
13:19:18JdGordon|sure, but those files are huge
13:19:39JdGordon|and we arent talking about single lines to change, there are probably hundreds of pointers that need fixing
13:19:57amiconnJdGordon|: Partial voice file is out of question on anything disk based. It could (and maybe should) be used on the Ondios though
13:20:09JdGordon|amiconn: oh? why?
13:20:24amiconnBattery usage and response time
13:21:30amiconnkugel: If the voice file is that large, the parameters need to be adjusted. 1.7MB would have been too large even years ago
13:21:49amiconnJdGordon|: What do you mean?
13:22:21amiconnMy own voice files are much smaller
13:22:21JdGordon|why would they increase?
13:22:48JdGordon|when the voice file isnt kept in memory the whole time anyway you have to do a open() and seek() already no?
13:22:53amiconnIf you're browsing the menu it would (nearly) constantly access the disk to buffer more clips
13:24:38amiconnAlso if you stopped browsing for a while, spinup time would cause a huge delay
13:24:58amiconnThat delay already exists for .talk clips and is rather annoying, but unavoidable for them
13:25:52amiconnOn hdd swcodec the voice file is only loaded once at boot (for hdd hwcodec it happens once each time you stop playback)
13:26:49 Join caramou [0] (
13:27:18amiconnFor flash based stuff partial loading is nice because you don't need to keep an eye on voice file size, the extra battery consumption is much smaller, and there is no spinup delay
13:27:32amiconnBtw, there's some optimisation potential in that code
13:28:18amiconnThe partial load caches 64 clips, and it allocates a buffer of 64 * max_clip_size for this. It finds the max clip size from the table in the header
13:29:03 Quit caramou1 (Ping timeout: 258 seconds)
13:29:03amiconnBut max_clip_size is a waste, because in the worst case it would cache the 64 largest clips - which most likely are smaller than 64 * the largest one
13:29:46amiconnIf we change the voice file format a bit so that it includes a field for the necessary buffer size, the code would become (slightly) smaller *and* it would allocate less ram
13:31:37 Join fyre^OS [0] (
13:34:24 Quit fyrestorm (Ping timeout: 258 seconds)
13:36:43JdGordon|is useing offsets any slower/worse than using pointers directly?
13:37:02JdGordon|I suppose it is irrelevant if we want a long term solution to skin buffers though
13:49:56JdGordon|how do we feel about these macros and typedef to replace every pointer in the skin engine?
13:50:48JdGordon|struct foo *bar = SKINOFFSETTOPTR(skin_base,, struct foo*)
13:52:12amiconnkugel: From a quick test it seems that voice is fixed now on swcodec (tested H1x0, mini2g and Color)
13:52:25amiconnHwcodec still doesn't play .talk clips
13:52:52amiconnFurthermore I already hd two occasions for the MMC directory vanishing
13:53:38amiconn(testing on Ondio FM)
14:07:28 Join ewloni [0] (
14:07:43kugelamiconn: check how big the voice file and the audio buffer is
14:08:08amiconnAudio buffer is 1.23 MB, voice file is 1.15 MB
14:08:41kugelclose one
14:08:47kugelperhaps .talk doesnt fit?
14:09:05amiconnThe it wouldn't talk anything, would it?
14:09:20kugelvoice and talk is loaded separately
14:09:44amiconnThe effect atm is that rockbox pretends the .talk clips aren't there. It spells all folders
14:10:18amiconnI just tried VOICE_PARTIAL_LOAD on Ondio. It works, but even then .talk clips don't
14:11:13kugelthat still allocates the entire voicefile size
14:11:34kugelah that was PROGRESSIVE_LOAD
14:11:55amiconnSince partial load works on Ondio, we can drop the separate progressive load
14:12:25 Quit Xerion (Read error: Connection reset by peer)
14:12:45amiconnThat was introduced way before swcodec, and was designed that way to keep it more similar to full load while avoiding the long load time of the entire voice file (since Ondio storage access is slow)
14:13:43amiconnPartial load does basically the same, with the additional benefit of not needing to take much care about voice file size
14:15:04kugelamiconn: does that work ?
14:15:25amiconnThe disadvantages are more storage accesses when browsing around a lot (because it throws away clips) and a slight binsize increase (~400 bytes)
14:15:51kugelhwcodec needs to be PARTIAL_LOAD sooner or later, binsize and/or english.lang is going to get smaller
14:18:00[Saint]Is *not going to get smaller?
14:18:49amiconnkugel: For hdd based targets partial load is out of question
14:20:44 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
14:21:14*JdGordon| is very releived to realise that converting the skin engine to offsets isnt actually a difficult task, just huge
14:22:31amiconnkugel: Seems to work; I've done a simplified implementation though
14:23:02amiconnPlease don't commit, I'll clean up my partial load patch (maily fixing comments and cleaning up the ifdefs) then commit at once
14:23:15kugelokay :)
14:26:23 Join keyb_gr [0] (
14:27:32 Quit zchs (Ping timeout: 260 seconds)
14:27:35CIA-22New commit by pamaury (r30913): fuze+: scramble bootloader file
14:28:34 Join zchs [0] (
14:28:39 Join lebellium [0] (
14:28:44pamauryI'm now close to have a working a mkimxboot, hopefully just a few lines of code before it's ready
14:29:14CIA-22r30913 build result: All green
14:30:11CIA-22New commit by pamaury (r30914): imx233: don't include debug-imx233.c in bootloader
14:30:14CIA-22New commit by pamaury (r30915): elftosb: add static
14:31:50CIA-22r30914 build result: All green
14:33:45CIA-22r30915 build result: All green
14:39:26 Join y4n [0] (y4n@unaffiliated/y4ndexx)
14:44:10 Quit sinthetek (Ping timeout: 258 seconds)
14:48:32 Quit zchs (Ping timeout: 260 seconds)
14:48:45 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
14:48:55 Join zchs [0] (
14:58:18CIA-22New commit by amiconn (r30916): Voice related fixes and cleaup ...
14:59:50CIA-22r30916 build result: All green
15:00:08kugelamiconn: is all mmc so slow or why do you use mmc instead of ondio for the define?
15:00:56amiconnI kept it as it was. Using target specific ifdefs in apps/ is (or at least was) frowned upon
15:01:04 Quit Buschel (Ping timeout: 260 seconds)
15:01:36amiconnMMC isn't that slow in general, it just happens that the Ondios are the only targets with STORAGE_MMC
15:03:51amiconn(Standard) MMC uses a transfer clock of 20 MHz and an 1 bit interface (very similar to standard SD which uses 26 MHz and an 1 bit interface)
15:04:46amiconnThe problem on the Ondios is that MMC is hooked up to the CPU's SPI, and the highest possible clock is CPUCLK/4, i.e. 3 MHz
15:04:59 Join Horscht [0] (
15:05:00 Quit Horscht (Changing host)
15:05:00 Join Horscht [0] (~Horscht@xbmc/user/horscht)
15:05:21amiconn3 Mbps is 375 KBps
15:06:44amiconnThe driver is actually quite efficient, i.e. we get very close to this theoretical limit, but that still means a complete voice file would take ~4 seconds to load
15:07:35 Quit Horschti (Ping timeout: 258 seconds)
15:07:52 Quit Horscht (Read error: Connection reset by peer)
15:07:52[Saint]Both the ondio users in the world will be impressed if that was sped up someway... ;)
15:08:59 Join Horscht [0] (
15:09:00 Quit Horscht (Changing host)
15:09:00 Join Horscht [0] (~Horscht@xbmc/user/horscht)
15:09:26[Saint]I wonder if it would be sane to load voice files in batches based on the current screen?
15:09:55amiconnkugel: It seems using a target specific define is actually better. See Fuze+ delta... it's certainly not wanted to do partial load there, is it?
15:10:29amiconn(meaning Fuze+ had progressive load before)
15:10:37***Saving seen data "./dancer.seen"
15:15:40 Join sinthetek [0] (
15:15:41 Quit sinthetek (Changing host)
15:15:41 Join sinthetek [0] (~sinthetek@unaffiliated/sinthetek)
15:17:00 Nick [sanyi] is now known as [Sanyi] (~sanyi@unaffiliated/alexanderc)
15:25:34CIA-22New commit by amiconn (r30917): Make usage of TALK_PARTIAL_LOAD actually Ondio specific. MMC is not always slow.
15:26:47amiconnHmm, the commit seems to hang here?!
15:27:06bluebrother^it will return. Give it a while :)
15:27:18bluebrother^it's been happening for me sometimes lately
15:27:20CIA-22r30917 build result: All green
15:27:34bluebrother^(though annoyingly sometimes even the build server was faster)
15:28:09amiconnYes, that is the case right now
15:28:33 Join Xerion [0] (
15:28:39amiconnOh it returned
15:37:04bertrikpamaury, did you figure out if the fuze+ has a GPIO to indicate reception of an RDS packet?
15:40:33 Join Guest5141 [0] (
15:41:44 Quit Guest5141 (Client Quit)
15:44:01 Quit Horscht (Ping timeout: 260 seconds)
15:46:30 Join Horscht [0] (
15:46:31 Quit Horscht (Changing host)
15:46:31 Join Horscht [0] (~Horscht@xbmc/user/horscht)
15:48:00 Quit Horscht (Client Quit)
15:52:16[Saint]kugel: s/freeze/release/?
15:52:42[Saint](Wrt: mailing list)
15:53:25 Part ewloni ("WeeChat 0.3.6")
15:58:31 Quit caramou (Quit: Leaving.)
15:59:45 Join Horscht [0] (
15:59:45 Quit Horscht (Changing host)
15:59:45 Join Horscht [0] (~Horscht@xbmc/user/horscht)
16:36:57 Join chkktri_ [0] (
16:36:58 Quit chkktri (Read error: Connection reset by peer)
16:47:20 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
16:48:07kugelp[Saint]: a/freeze/branch/ actually
16:49:07[Saint]*almost* the same ;) I just certianly didn't think you meant to say freeze.
16:53:24 Part Llorean
17:10:41***Saving seen data "./dancer.seen"
17:18:25funmanhm what happens if i write to rockbox-committers with a different address from the one i'm subscribed?
17:18:54funmanmy mail is silently discarded and i should send it again with the right From: address?
17:20:12Bagderbut we saw your mail
17:20:27funmanhm weird
17:20:48funmanperhaps anyone can reply to a mail sent to the list?
17:21:41 Nick [Sanyi] is now known as [sanyi] (~sanyi@unaffiliated/alexanderc)
17:25:09Bagderllorean did ~10 hours ago
17:29:46[Saint]He's (technically) a committee.
17:30:08[Saint]Yeah...colour me shocked too :p
17:30:51[Saint](Note: humour)
17:35:59[Saint]The committers list is a bit weird, though. As there's some very non-active (past) commiters on it, and some rather recently active (current) commiters that aren't.
17:37:02funmanit's only used to mention RSB (re)election and propose new committers
17:38:10 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
17:41:01Bagder[Saint]: we invite _all_ committers to it
17:41:21Bagderif someone isn't on it, that's his/hers decision/fault
17:41:39pamaurybertrik: no, I haven't looked at it, I'm working on mkimxboot right now but I can give it a try if you have some code
17:54:54 Join hilbert [0] (
17:55:53gevaerts[Saint]: I only reproduced it
17:59:16 Join dfkt [0] (dfkt@unaffiliated/dfkt)
18:00:29 Join thomasjfox [0] (
18:00:31 Quit thomasjfox (Changing host)
18:00:31 Join thomasjfox [0] (~thomasjfo@rockbox/developer/thomasjfox)
18:17:36 Quit liar (Remote host closed the connection)
18:23:58 Join nosa [0] (
18:25:31 Quit nosa-j (Ping timeout: 252 seconds)
18:25:31 Nick nosa is now known as nosa-j (
18:27:18 Join mystica555 [0] (
18:31:31 Join Thra11 [0] (
18:36:37n1s[7]: seems r30905 included the usbfix patch, was that intentional?
18:39:08 Join Buschel [0] (
18:40:16[7]i did actually check the diff of the following commit, but apparently didn't realize I had a months-old local change when committing that cleanup for a warning that only existed because of that local change :/
18:40:58[7]when are we going to branch? I'd like to commit that to the trunk, but not to the release branch for now
18:41:15[7]should I go for reverting it in the branch or reverting it now and then recommitting it once we've branched?
18:41:41n1si've been uncertain of haow to go forward with the usb patch since people reported less reliable usb on nano2g
18:42:05[7]I've never been able to get hold of one of the people affected by that
18:42:35[Saint]My transfer speeds dropped dramatically with an early version of it.
18:42:48[7]can you try with current svn head vs. r30904?
18:43:14[7]current head has the patch applied, r30904 doesn't
18:43:42[Saint]I can, yep. Not immediately though.
18:46:18[Saint]Android has seen my nanos don't get a lot of love these days.
18:47:22funman[7]: please add a comment for the char [64] added in r30906
18:48:26 Join The_Seeker [0] (
18:51:18 Quit domonoky (Quit: Leaving.)
19:00:58 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
19:10:20 Quit The_Seeker (Quit: The_Seeker)
19:10:45***Saving seen data "./dancer.seen"
19:27:31 Quit n1s (Ping timeout: 244 seconds)
19:27:56 Quit Buschel (Ping timeout: 260 seconds)
19:33:44*jhMikeS wants to destroy that autoscan nag on the FM screen (and it not remembering the scan/preset setting between entries into the screen)
19:33:59 Quit ReimuHakurei (Quit: Leaving...)
19:33:59 Nick Reimu_Hakurei is now known as ReimuHakurei (
19:35:06 Join n1s [0] (
19:35:07 Quit n1s (Changing host)
19:35:07 Join n1s [0] (~n1s@rockbox/developer/n1s)
19:35:19[Saint]jhMikeS: I seem to recall the fms recalled the scan/preset selection in the past.
19:36:52[Saint]It definitely *should*.
19:37:11 Quit stoffel (Remote host closed the connection)
19:37:32jhMikeSIt keeps jumping me back to preset
19:39:14[Saint]I've only played with it in the sim, which may give a different result. But I recall it working as intended. This was some many moons ago though, and my recollection may by broken.
19:39:46*jhMikeS always remembers it being sucky and naggy
19:40:20[Saint]I certainly remember wanting to strangle that nag screen.
19:40:52*pamaury has always been unsatisfied by the fm screen, it never worked the intuitive way for, and never remember my last freq
19:40:58jhMikeS"hello tuner enthusiast, who you like to purchase some presets today? if you act now, you get two for the nag of one."
19:42:32*[Saint] always thought the FM frequency should be saved as a config setting.
19:43:04jhMikeSthe frequency is saved, but my refusal to scan is not, nor is my preferred tuning method
19:46:28 Join ReimuHak_ [0] (
19:48:14jhMikeSThis RDS stuff probably needs to mind error correction if it doesn't already since it gets corrupt strings at times
19:49:45 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
19:50:10pamauryjhMikeS: the si7043 chip already does error correction I think, but clearly if the chip doesn't do it, one absolutely needs to do it
19:51:04jhMikeSso maby the 4701 doesn't? sometimes I see oddball garbage chars
19:51:13 Join The_Seeker [0] (
19:51:28pamauryjhMikeS: 470x doesn't do RDS when x<3
19:51:37jhMikeSWhy's it working?
19:51:46pamauryand we don't have rds support
19:52:00jhMikeSI modded the patch to work on gigabeat S
19:52:19pamaurywhich patch ?
19:52:19jhMikeSFS #12370
19:52:20fs-bluebot RDS support for Si4701/Si4703 (Sansa Clip Zip proof-of-concept) (patches, new)
19:52:50pamauryare you sure you have si4701 ?
19:53:55jhMikeSMaybe I should double-check that.
19:54:01pamaurybertrik: I think you need to discard rds packets with too much errors corrected
19:54:03 Join Buschel [0] (
19:54:12pamaurythat's written somewhere in the datasheet
19:55:05 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
19:55:06jhMikeSThere is a contradiction in the datasheet, showing lack of regs/bits yet showing supporting RDS features
19:58:23jhMikeSshowing up as 4702 in the Probably I misremembered?
19:58:26pamauryjhMikeS: perhaps try something like this: in si4700_rds_read_raw, read register 0xA and if bits 9:10=0b11, then discard packet
19:58:46 Quit ReimuHak_ (Quit: Leaving...)
19:58:57 Join ReimuHak_ [0] (
19:59:06pamauryeven then, the si4702 is not supposed to have rds support :-/
20:00:30 Join GermanMushroom [0] (
20:02:37jhMikeSeven the hardware info shows 4701 and I believe even the IC physically showed that on two boards
20:02:38 Quit Thra11 (Ping timeout: 248 seconds)
20:02:59jhMikeSso, it's chips with different firmware apparently
20:04:14jhMikeSregardless of theory, it's reading it :)
20:04:58[Saint]Nice discovery.
20:05:00pamauryinteresting, it means some other chips might have rds
20:05:25pamauryjhMikeS: can you try what I suggested for rds, to see if the weird characters go away
20:06:43*jhMikeS wonders if it was a different model that failed certain test and was sold as "4701'
20:07:49jhMikeShmmm...we're not using "verbose mode"
20:08:12pamauryjhMikeS: when reading register 0xA, you want to read it and not use the masked value probably
20:08:36pamaury(stating the obvious but we never know :))
20:09:05[Saint]Find another gigabeat s owner? Should be pretty easy to tell if you got a random substituted chip then.
20:09:37jhMikeSpamaury: well, it said "verbose mode only"
20:09:55jhMikeS[Saint]: don't know any except here :)
20:10:13jhMikeSso, they should come here, rather that I go to them :)
20:10:36pamauryhum true, then perhaps you need to enable it, let me reread it
20:10:44[Saint]Well...I meant via here, not irl :)
20:12:02pamaurybut the datasheet suggest that in non-verbose mode an invalid block is not notifie
20:13:19 Part The_Seeker
20:13:36pamauryyou can always have a try at enabling verbose mode and see what happen when checking for errors
20:16:13bertrik4701 and 4703 support RDS, 4700 and 4702 do not
20:16:19 Join Thra11 [0] (~thrall@
20:18:55*jhMikeS enabling and compiling
20:22:16jhMikeSnot seeing bad chars right away. it's not a constant problem but occasionally they show
20:23:26bertrikright now, missing any packet in a message, means missing the entire message
20:23:53bertrikbut either is not a big problem IMO
20:24:32jhMikeSyeah, I am seeing bad packets and having it rejecting but now caught a few bad chars forcing the reception to be weak
20:25:29jhMikeSare you saying it should wait for the next loop to try again?
20:26:34bertrikat first I thought I'd just wait until all characters were received, even when not in order, but in practice station change the station name and radiotext very quickly
20:27:17bertrikso I'd get a half old/half new string
20:28:11bertrikThe spec states specifically that stations should not change the station name, but in practice many of the commercial stations cycle traffic info in the station name.
20:29:46jhMikeSit's not that. I some plainly garbage strings.
20:30:06jhMikeSlittle boxes and odd chars
20:31:28bertrikI haven't seen bad characters so far, maybe that's a difference between si4701 and si4703
20:31:47jhMikeSit happens when reception is near the threshold for RDS
20:32:16jhMikeSthe verbose mode info seems to be available
20:32:34bertrikI thought we wouldn't need it
20:32:44bertrikbut I have no objection to enabling it
20:33:31bertrikhm, garbage characters in the strings mean that the segment field survived in all packets, but that the text field did not
20:35:44bertrikJust thinking freely, but I think we could nicely do "album art" based on the PI code, since the PI code should be specific to the station, and even stay the same on different frequencies of the same station.
20:37:20 Join robin0800 [0] (
20:39:33jhMikeSI'm not really seeing any improvement with verbose mode
20:41:26CIA-22New commit by pamaury (r30918): imxtools: add const and set pointer to NULL on free to avoid double free
20:41:30CIA-22New commit by pamaury (r30919): imxtools: add function to free instruction
20:42:15pamauryjhMikeS: perhaps the chip doesn't do error correction correctly ?! ;)
20:42:36bertrikjhMikeS, you're using an si4701, right? I don't see the verbose mode stuff in the si4701 datasheet (but it could still be there anyway)
20:43:16CIA-22r30918 build result: All green
20:43:36pamaurybertrik: is the si4701 datasheet available on the internet ? I can only find the 4703 one
20:44:04CIA-22New commit by pamaury (r30920): Introduce mkimxboot to build bootloader images for the i.MX (only the fuze+ currently). Still not integrated in rbutil
20:44:07 Quit GermanMushroom (Ping timeout: 252 seconds)
20:45:00CIA-22r30919 build result: All green
20:45:24bertrikpamaury, google for si4700-B15rev1_0
20:45:29jhMikeSwell, it is in fact showing up as 4701. I read again and bit 9 is set.
20:46:28CIA-22r30920 build result: All green
20:46:38jhMikeS3 bits were allocated away from FIRMWARE and moved to REV beween 0/1 and 2/3
20:47:08 Quit y4n (Quit: PANTS OFF!)
20:47:14bertrikyes, I noticed that, I inferred that bit 9 was the indicator for RDS or not
20:47:33bertrikperhaps a slightly dirty hack :)
20:48:14jhMikeSor a clean one :)
20:48:21 Quit robin0800 (Ping timeout: 253 seconds)
20:51:12jhMikeSbsi4700-b15rev1_0 appears like it's some preliminary junk, partly missing things for 4701
20:53:44 Join robin0800 [0] (
20:55:01jhMikeSmaybe sometimes it just plain can't detect errors? what range of characters is legal in a string?
20:55:16 Join GermanMushroom [0] (
20:55:40bertrikthe si4702/03 datasheet mentions something about a RDS high performance mode flag RDSPRF 06h[9]=1
20:55:53bertrikit's not mentioned later in the register description though
20:56:22bertrikactually, I think there can be a few encoding, I just assumed plain ASCII for now
21:01:15pamauryjhMikeS: error detection is always possible normally
21:01:27pamauryat least knowing if what you've got is correct or not
21:01:47 Quit robin0800 (Ping timeout: 258 seconds)
21:07:01 Join robin0800 [0] (~robin0800@
21:09:28jhMikeSis it possible the parser is putting together bits that don't belong next to one another?
21:10:49***Saving seen data "./dancer.seen"
21:11:30 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
21:11:38pamauryhum in fact, it's possible that some invalid data is considered valid but this is extremely unlikely
21:13:41bluebrother^ok guys, I'm going to release Rockbox Utility 1.2.11 unless someone speaks up in the next couple of minutes.
21:13:41 Quit dfkt (Quit: -= SysReset 2.55=- Sic gorgiamus allos subjectatos nunc.)
21:13:46jhMikeSsince I don't have a freak tuner afterall, something else seems to be going. it says it does all the error handling.
21:14:29saratogabertrik: higher PCLK did decrease runtime on the clip+, from about 17 hours to 15 hours
21:14:39saratoganext i'll try higher PCLK and lower FCLK
21:15:27bertrikbluebrother^, let me check clip zip support again please
21:16:18bluebrother^bertrik: sure.
21:16:26jhMikeSmaybe weird things are being encoded here and ascii isn't a valid assumption
21:18:41bertrikbluebrother^, ok, it's detecting the clip zip correctly (when "disabled" targets are enabled)
21:22:34bertrikjhMikeS, it's supposed to patch only packets together if they come in the right sequence, if not, it should start waiting for the first packet of a new message
21:23:24bertrikCould you just remove the existing rds_event without side effects? It was already there from the ipod fm remote tuner, so I didn't want to touch it too much.
21:23:52bertrikor perhaps I don't understand how it's supposed to work
21:24:35jhMikeSbertrik: it doesn't seem the radio screen updates if it's not there :\
21:25:05bluebrother^bertrik: ok, so I'm going to push the release then.
21:25:12jhMikeSI also have a suspicion polling the buffers without an "ok" from the parser probably leads to this
21:26:28n1si thought we already knew the beasts tuner was capable of rds and someone did a dirty proof of concept a while ago
21:26:38jhMikeShmmm...ah, seems it waits before placing data there
21:27:14 Quit robin0800 (Ping timeout: 256 seconds)
21:27:19jhMikeSn1s: I probably wasn't here. I remeber seeing the registers in action, but that was it.
21:29:49n1syeah i saw that too when hacking on the driver but i think someone whipped up something that decoded the messages in the debug menu or something, i might have dreamt that though, anyways, not important
21:33:16n1shah, i think i wasn't entirely wrong
21:34:24bertrikI did make something in the past and looked at it together with AlexP on his beast, but I didn't have an RDS capable target myself
21:34:50n1sbertrik: yes, i fond that in the logs
21:34:59n1sfebruary 2009 :(
21:35:36 Quit Farthen (Excess Flood)
21:36:18n1si think i was planning on looking into it but never got around to it, i don't use the radio and here rds is only used for stupid ads anyway
21:36:22 Join Farthen [0] (~Farthen@2a01:4f8:101:2a4:0:bc28:b2e1:9)
21:36:28CIA-22New commit by bluebrother (r30921): Rockbox Utility: bump version to 1.2.11.
21:36:59pamauryare we releasing a new version of rbutil ?
21:37:09bluebrother^pamaury: yes
21:37:19*bluebrother^ spots yet another copy of bin2c
21:37:34bluebrother^I need to get that cleanup finished I was working on months ago
21:38:02pamauryyes I introduced a copy of bin2c but realized a few minute ago there was one in tools/
21:38:09CIA-22r30921 build result: All green
21:38:59bluebrother^I wanted to merge bin2c with ipod2c, and we have some bin2c versions that aren't identical (the mkamsboot version behaves differently)
21:39:19pamaurydifferently in which way ? (I copied this one)
21:40:57CIA-22New commit by pamaury (r30922): mkimxboot: fix bootloader load address, replace call by jump on final call to have a clean boot
21:41:02bluebrother^the mkamsboot version can process multiple input files, while others can't.
21:41:19pamauryah yes
21:41:59 Join [Saint_] [0] (~Saint]@unaffiliated/saint/x-8516940)
21:42:26CIA-22r30922 build result: All green
21:42:44bluebrother^nothing that couldn't be combined via a few command line arguments.
21:42:55bluebrother^and the ipodpatcher variant adds a different header iirc.
21:43:07 Join [Saint__] [0] (~Saint]@unaffiliated/saint/x-8516940)
21:45:01 Quit [Saint] (Ping timeout: 252 seconds)
21:45:32 Quit Buschel (Ping timeout: 259 seconds)
21:46:11CIA-22New commit by bluebrother (r30923): Tag Rockbox Utility 1.2.11 release.
21:46:30 Quit [Saint_] (Ping timeout: 256 seconds)
21:49:24 Join robin0800 [0] (~robin0800@
21:50:30bertrikjhMikeS, one tricky thing is that the si4700_detect function is called before si4700_init, so the mutex is not initialised yet during si4700_detect
21:53:35jhMikeSoops, that's bad :)
21:55:40 Join kugel_ [0] (~kugel@rockbox/developer/kugel)
21:55:41 Quit kugel (Remote host closed the connection)
21:55:42 Nick kugel_ is now known as kugelp (~kugel@rockbox/developer/kugel)
21:55:57bertrikit's needed for the clip+ which can have either an rda5802 or a si4702 fm tuner, and the difference is detected by reading the device id
22:03:01pamauryZoiah: ping
22:03:01 Quit robin0800 (Quit: Leaving)
22:03:10 Quit Thra11 (Ping timeout: 252 seconds)
22:04:06jhMikeSwhere's the code for that?
22:04:46pamauryI've finally put the instructions to build the Fuze+ bootloader on the wiki
22:06:11 Quit [Saint__] (Read error: Connection reset by peer)
22:06:15 Join [Saint] [0] (~Saint]@unaffiliated/saint/x-8516940)
22:06:48jhMikeSnvm, got it
22:07:24 Quit kugelp (Remote host closed the connection)
22:09:07 Join kugel [0] (~kugel@rockbox/developer/kugel)
22:09:09 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
22:09:42 Quit guymann (Ping timeout: 276 seconds)
22:10:20 Quit TheLemonMan (Quit: WeeChat 0.3.6)
22:11:01 Join guymann [0] (
22:11:36jhMikeSI think RADIO_EVENT will have to sneak back in for now too
22:12:05TorneDoh, we have a new committer since i last caught git up from svn
22:15:40 Quit thomasjfox (Remote host closed the connection)
22:16:18jhMikeSreally, perhaps radio events should use the event system much like playback does
22:17:46 Part Zagor
22:20:04 Join liar [0] (
22:22:21 Quit hilbert (Quit: Computer has gone to sleep.)
22:22:28saratogapamaury: since the install for the Fuze+ is so similar to AMS, maybe it would make sense to make mkimxboot use the mkamsboot?
22:22:50pamauryso similar ?
22:22:59 Join hilbert [0] (
22:23:23pamaurythe only similar part is putting a file at the root of the fs
22:23:34saratogawell similar from a users stand point
22:23:57pamauryyes but that's not the job of mkamsboot
22:24:04pamaurythat's the job of the code in rbutil
22:25:05 Quit benedikt93 (Quit: Bye ;))
22:25:05pamaurythe problem is that this installation method is really specific to the fuze+, I know if at least one other imx233 target which doesn't work the same way so it would make sense to integrate the Fuze+ with the AMS ones perhaps
22:27:29 Quit GermanMushroom (Ping timeout: 260 seconds)
22:28:08saratogaare you ready to release a bootloader?
22:29:11saratogadid anyone test FS #12352 on a D2 (non-plus)?
22:29:12fs-bluebot Charger detection for newer Cowon D2+ (pcf50635 equipped) (patches, unconfirmed)
22:29:23saratogaassuming nothing breaks i'd like to commit it
22:29:37pamauryI would prefer to wait for some other people to test it
22:31:26Torneargh, how on earth do we have 97 different file extensions in our tree :)
22:32:39[Saint]C/c H/c WPS/wps BMP/bmp etc?
22:32:57Torneactually different ones
22:33:48*Torne is having fun with .gitattributes
22:36:32*jhMikeS considers statically initialized objects like pthreads has
22:40:04 Join Teasp00n [0] (
22:45:31 Quit ReimuHak_ (Quit: Leaving...)
22:45:47 Join ReimuHak_ [0] (
22:50:14 Quit Stummi (Quit: Bye!)
22:50:44JdGordon|are we still in freeze?
22:51:07bluebrother^according to the topic, yes
22:51:57JdGordon|next release we chain alexp to an internet connection so it doesnt drag out :)
22:53:27gevaertsI'm not sure if he can really be blamed for this :)
22:56:07bluebrother^Bagder: can you move Rockbox Utility 1.2.11 to the download server?
22:56:21[Saint]Sure he can! He probably won't deserve it, but we can still blame him.
22:57:55jhMikeSradio screen fails to update RDS lines automatically (because there's no UPDATE_ALL in ACTION_NONE upon a radio event)
23:00:14JdGordon|are the rds tags marked dynamic in libs/skin_parser/tag_table.c?
23:00:21jhMikeSno idea
23:00:38jhMikeSI can certainly look though :)
23:01:06JdGordon|yes they are
23:01:22jhMikeSnot working though :\
23:01:35JdGordon|someone really needs to rewrite the fm screen code :)
23:03:15jhMikeSmono/stereo status updates
23:04:41jhMikeSI wonder why its getting that ok but not the RDS
23:08:21jhMikeSI'm not sure but I think the Canadian station might be making more bad data than the US ones!
23:08:37 Join metaphys [0] (
23:09:19 Quit metaphys (Client Quit)
23:09:19bertrikmaybe they use more weird french characters!?
23:09:30jhMikeSmight be
23:09:49 Join metaphys [0] (
23:10:04jhMikeSI don't have any other RDS tuners on hand to check em
23:10:53***Saving seen data "./dancer.seen"
23:12:55lebelliumOh I see the word French :D
23:14:24 Nick GeekShadow is now known as antoine_t (
23:15:26jhMikeSwhy is the f'ing skin not updating that ?!?
23:16:38jhMikeSok, I totally wiped everything and rebuilt again and it looks like it's doing it now :\
23:18:03Bagderbluebrother^: done now good sir!
23:18:49jhMikeSnope, still not quite right. clicking vol or whatever makes it show it before it's ready, but otherwise it waits for a full line
23:20:30jhMikeSIn the first line, I usually see text fragments that end up in the second line
23:21:18Torneheh. this line ending nonsense is very entertaining
23:21:29Tornewe have 7560 .c files and only 1431 of them have svn:eol-style set
23:21:35Torneand 35 of those are set *wrong*
23:21:53jhMikeS"JoviWant"/"or[]Mliv" :) not "Bon Jovi - Wanted Dead or alive"
23:23:08funmanTorne: not again!
23:23:21Tornefunman: yeah, i'm working out how to make git fix this all for us
23:23:32funmanignore that crap
23:23:44jhMikeSwhuuuu? something git can't fix?
23:23:45funmanmake the 2 or 3 windows files use \r\n and the rest \n
23:23:52Tornefunman: that's what i'm doing
23:23:58Torneit's just kinda fiddly
23:25:58Tornefunman: the files which actually have wrong contents (which is quite a few) need to be changed. the actual svn metadata is not important as it will go away when we convert
23:26:50 Join Scromple [0] (
23:27:10 Quit n1s (Quit: Ex-Chat)
23:28:07bluebrother^Bagder: thanks!
23:33:36bertrikjhMikeS, your rds v2 patch works for me. I get the impression that it takes a bit longer to receive the radiotext message
23:34:14 Quit metaphys (Quit: CGI:IRC (EOF))
23:34:38jhMikeSbertrik: not sure why it would take longer
23:36:00bertrikmaybe the fmradio mutexes give more opportunity to switch away and not allow the RDS data to be read quickly enough
23:44:26jhMikeSwhat are the time constraints on that? they'll only switch away if contended
23:44:35CIA-22New commit by torne (r30924): Bulk convert all DOS line endings to UNIX. ...
23:45:40Torneyes, i am a scumbag :)
23:45:42jhMikeSthing is, it could get quite a delay if in fact the thread that reads it isn't the current one
23:45:48Tornesorry for ruining all your blames :p
23:46:18CIA-22r30924 build result: All green
23:46:19gevaertsTorne: *ruin*? svn blame will now point to you for everything!
23:46:28TorneSure, but I'll ignore you :)
23:46:40TorneI can just assume all accusations that i've touched something are false now
23:46:43TorneEven if I really have :P
23:47:05bertrikjhMikeS, packets come in at about 11.5 packets/s, so I guess about 88 ms
23:49:38 Quit liar (Remote host closed the connection)
23:50:07 Join liar [0] (
23:50:24jhMikeSevent with the bitbanging, the packets could be read in the ISR and queued to the thread for parsing
23:51:14jhMikeSI did up the thread priority, which should let it run asap (it should hardly ever be scheduled at that rate)
23:52:22CIA-22New commit by torne (r30925): Convert one more file from DOS to UNIX. ...
23:52:53Torneright. now everything is UNIX apart from certain windows-specific files that are never edited by humans anyway (visual studio projects, and the cat/inf files for some device drivers)
23:53:16 Quit lebellium (Quit: ChatZilla 0.9.87 [Firefox 8.0/20111102223350])
23:53:52CIA-22r30925 build result: All green
23:54:56funmanTorne: git blame -w
23:55:04Tornefunman: Indeed
23:55:20TorneBut svn is too stupid to do that
23:55:28Torneso it won't work until we switch to git :)
23:55:49JdGordon|which is soon right? :)
23:55:53TorneWorking on it!
23:55:55TorneThis is one of the steps :p
23:55:56 Nick JdGordon| is now known as JdGordon (~jonno@rockbox/developer/JdGordon)
23:56:07TorneHaven't had as much time to work on it these alst couple days as i'd hoped, personal stuff came up
23:56:11Tornebut have done some things :)
23:56:18 Join Keripo [0] (
23:56:51 Join metaphys [0] (

Previous day | Next day