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 2005-02-13

00:02:31 Join muesli- [0] (muesli_tv@
00:04:29 Join muz [0] (
00:04:58muzhey guys any more progress from yesterday on iriver?
00:05:24*preglow Hellfish and Producer - Toilet Wars
00:06:02rashernot really
00:06:32muzis the flac decoder nearly done
00:06:57sneakumswell, the sound driver isn't done, so.
00:07:09linuxstband the whole audio playback system isn't done...
00:07:23muzoh ok
00:07:37muzthe sound part of this project is gonna take the longest imo
00:07:51linuxstbbut I have got the FLAC decoder decoding data on the iRiver - at about 8% of real-time (but the iRiver itself is only running at about 10% of full capacity anyway)
00:08:03preglowit's gonne be a while until we make sound
00:08:18preglowso don't expect anything very soon
00:08:40muzokay, its just its sooo exciting
00:08:50muzi got myself some shure e2c's today
00:08:51muesli-yeah :D
00:08:54preglowlinuxstb: ten percent is very probably an understatement
00:09:22muesli-better to say to less than 2 much ;)
00:09:37sneakumsyou mean about the memory thing?
00:09:50preglowwell, 10/140 < 0.10
00:09:55preglowand the memory can be faster as well
00:09:56sneakumsoh, ah.
00:10:07preglowso i at least hope we can expect more than that
00:10:14muzis that what linus is working on now, the clock speed?
00:10:20preglowi don't know
00:10:30preglowdon't think so
00:10:41muzhe aint been on today
00:10:48preglowclocking the cpu higher should be pretty easy, but it's not that important right now
00:10:50linuxstbI don't know when he's working on it, but I think he said he will do the clock speed before sound.
00:11:00preglowwe'll see
00:11:03preglowgive the guy a break ;)
00:11:12preglowhe's been working like mad for a week now
00:11:13muzyeah sorry, hes a legend
00:11:27muesli-do i understand it right that youre gonna overclock it!?
00:11:34preglowmuesli-: it's underclocked right now
00:11:43preglowmuesli-: it's more like we're going to take it to its full potential
00:11:59linuxstbI'm sure someone will try it - lots of insane people around here...
00:12:10muesli-*arms up*
00:12:24preglowdepends, might not be possible, heh
00:12:28preglowwithout raising the system clock, that is
00:12:32amiconnHCl: How large (in terms of weight and size) is an original gameboy?
00:12:48HClamiconn: um. not sure.
00:12:55HCland not big
00:13:02thegeekI think linus said he would work on usb and then clock, since usb is more or less done, he is prob. working on clock now
00:13:06muzis the iriver running at the archos' clock speed of 12 mhz? is that why u said 10% full speed?
00:13:08preglowusb _is_ done
00:13:16thegeekbut there were some bugs
00:13:17preglowat least i can't find any problems with it
00:13:19thegeekmight be more;)
00:13:23preglowand i've been using it this entire day
00:13:29amiconnHCl: If I succeed to get rockboy running on archos, it would probably run on a unit smaller than the original...
00:13:30preglowi found a bug yesterday, and that's been fixed
00:13:37HClamiconn: heh
00:14:23amiconnHCl: Speaking about Ondio...
00:14:24preglowamiconn: isn't the archos cpu like 10 mhz?
00:14:56amiconnArchos player & Ondio are 12 MHz, Recorders are ~11 MHz
00:15:09preglowthen i'd pretty much say farewell to the interpreter core
00:15:52amiconnWe'll see. Do you know the original gameboy clock?
00:15:54preglowfive/four real clocks per interpreted clock is nigh well impossible
00:15:58preglowamiconn: 3-4 mhz
00:16:43preglowamiconn: yup, 4 mhz
00:16:43amiconnThe original uses a Z80 derivate. Z80 can't do one instr/clock, shortest instructions need 4 cycles
00:17:03preglowamiconn: you'd still be cutting it _very_ close for an interpreter core
00:17:18preglowyou still need to emulate the display hardware
00:17:32preglowi guess sound is pretty much out of the question on the archos, unless you have support for pcm
00:17:41amiconnYeah, I don't say it'll run realtime as-is. However, I expect it to run faster than on current iRiver rockbox+
00:18:00preglowamiconn: that'll probably be true, i think the sh1 is faster clock for clock than 68k derivatives
00:18:12muzwhats the video plugin like on rockbox?
00:18:23muzis it like watchable?
00:18:33amiconnPlus the ram isn't set up conservative, and the lcd driver is asm optimized like hell
00:18:52amiconnmuz: I think so, apart from the tiny screen
00:19:06preglowamiconn: what about display resolution?
00:19:11muzso it should be good on iriver with the bigger screen and greyscale
00:19:42amiconnIn fact, the video plugin (and the other grayscale plugins) use a smart trick to show grayscale on an lcd that does only b&w in hardware
00:20:18amiconnpreglow: Yes, display resolution is a problem on the archos
00:20:36preglowyou could downsample, if it's very much smaller
00:20:50thegeekhow hard is it to write a dynarec core compared to an interpreter core?
00:20:50amiconnThe first version will simply use every other pixel from the original buffer
00:20:51thegeekmuch harder?
00:21:05preglowthegeek: yes, if you want to do it efficiently
00:21:14amiconnThis gives the same visible area as the iriver version
00:21:46preglowamiconn: yes, but you'll have to do resampling by a whole number of pixels, everything else'll look like shit on a screen that tiny
00:21:52preglowand that really limits your options
00:22:08preglowso you might be stuck with skipping every second pixel
00:22:34amiconnYes, that's what I mean (and will do first)
00:22:38preglowthegeek: as a matter of fact, it might not be that hard for a z80, it's got really few registers, that simplifies things a lot
00:22:51thegeeknot that I have the skills to do it
00:22:59thegeekjust wondering here
00:23:01preglowi've never tried it myself, but i think i should be able to
00:23:08thegeekI know a little c/c++
00:23:11preglowi'd have to be forced
00:23:11thegeeknot much
00:23:13preglowand paid
00:23:18preglowbut i might be able to do it
00:23:23thegeekI find it quite interesting
00:23:33thegeekI'm wondering if I'll learn it
00:23:40preglowyou'll have to learn assembly first at least
00:23:49thegeekI know assembly
00:23:54preglowthen lucky you!
00:24:00thegeekI don't know how to code it very well
00:24:07thegeekbut I know it well from debugging
00:24:14thegeekI was a 0day cracker for some time;)
00:24:20thegeekbut I quit
00:24:23thegeekso don't shoot me
00:24:25preglowthen learn z80 assembly and tel hcl you want to code a dynarec engine
00:24:32HCl yay! :p
00:24:34preglowi've cracked things myself
00:24:37preglowso i won't shoot you
00:24:45thegeekreally don't want to be shot
00:25:34jypHow/where are the plans to merge the lcd drivers common code ?
00:25:48thegeeksurely there must be a z80 dynarec core that can be built upon?
00:25:57HCli tried to search for it
00:26:01HClbut i guess i sought gameboy specific
00:26:02jyp(I might actually start gmini 220 lcd)
00:26:04HCllet me search again
00:26:29thegeekeven if it is not gameboy specific it should help?
00:27:00preglowamiconn: so, having a hand coded asm lcd does speed things up?
00:27:13preglowlcd driver, yes
00:27:50amiconnpreglow: Well, it depends. Iirc, the iriver (main) lcd is connected parallel, so it might go fast enough without asm.
00:28:16amiconnHowever, the archos lcd (and the iriver remote lcd) are connected serial
00:28:24preglowahh, yes, it's put on the bus
00:28:27preglowthat's right
00:28:54amiconnThe archos lcd serial requires bitbanging (not connected to real serial port hardware)
00:29:28 Quit funkymonkey (" HydraIRC -> <- State of the art IRC")
00:29:29amiconnOur driver does right above 1 MBit/s, that's even slightly above specs for the lcd...
00:30:42amiconnYes. The lcd serial data line is connected to a universal i/o port pin of the cpu
00:30:56amiconnSo we have to clock out each bit individually, in software
00:31:09HClcan't find a z80 dynarec engine
00:31:21preglowthat's sounds like a plain pain in the ass to code
00:31:34preglowHCl: i could have told you that :P
00:31:40HClyes yes.
00:31:42HClshush you :P
00:31:46amiconnpreglow: Look at firmware/drivers/lcd.S if you want to get an impression...
00:31:51preglowi just did
00:32:11HClwhile we're talking about the underlying lcd
00:32:24HClany chance we could make up an interface to have direct access to it? at least on the iriver?
00:33:19preglowwhy'd you want that?
00:33:33preglowif it is connected directly to the memory bus, that would be A Bad Idea
00:33:45preglowyou need to copy the framebuffer to iram first anyway
00:34:12preglowbut that's just if my assumption is correct, yes
00:34:47HClso i wouldn't have to pass the internal rockbox buffer
00:35:20*HCl tries a slight optimization that should make rockboy 12% faster
00:35:28HCloh wait, no
00:35:30HClmuch less :/
00:35:32HClbut still
00:35:34HCla bit faster.
00:35:57muesli-boy *g*
00:36:12preglowwhat kind of optimization would that ne?
00:36:27 Quit mecraw ()
00:37:02HClpreglow: not making it do all the scanline graphic calculation for the scanlines that don't get drawn anyways
00:37:32muzhey would an engineering degree be useful in coding in stuff like rockbox
00:37:45amiconnHCl: Did you check whether gnuboy outputs the scanlines sequentially?
00:37:53 Join gromit^ [0] (
00:38:42linuxstbOh dear... Unless I've done something badly wrong, libmad is decoding at about 4% of real-time.
00:38:48HClamiconn: it does.
00:38:59HClamiconn: i'm not sure whether i can guarantee it, but so far, it does.
00:39:04amiconnHCl: So there is another possible optimisation...
00:39:05preglowmuz: depends on what engineering degree you're talking about
00:39:27HClamiconn: i'm more looking at the build-scanline-data routine, see if i can turn it into a more compatible format
00:39:45amiconnHCl: That'd be an even better idea.
00:39:59preglowmuz: and depends on how low level you wanna get, i'm in electric engineering myself, and this is just the thing for me
00:40:04preglowHCl: well, that's to be expected
00:40:06muza general one
00:40:09preglowlinuxstb: that's to be expected
00:40:16amiconnIf you can make it produce 'scancolumns' instead of scan lines, this should gain much speed
00:40:21HCli'm also looking at the cpu core to see how adaptable it is
00:40:25HClamiconn: hmmmmm.
00:40:30HCli'm not sure if i can do that
00:40:34preglowmuz: i don't really know what you mean about 'a general one', i probably don't know the education system in where you're from
00:40:46 Join ashridah [0] (
00:41:49muesli-g'day mate :)
00:42:00HClthere is 0 documentation on this
00:42:06HCland i have absolutely no clue what they are doing here
00:42:16ashridahcool. i managed to lock up my iriver while shutting it down in rockbox
00:42:28ashridahps, if you insert the usb adapter while it's shutting down, it locks pu :)
00:42:29ashridahup even
00:42:40preglowamiconn: the lcd is addressed by column first?
00:43:02amiconnpreglow: No, but the bytes are oriented vertically
00:43:19preglowreally, funky
00:43:29amiconnThat's the hardware format, see lcd-h100.c (and lcd-recorder.c) for comparison.
00:43:35preglowall the lcds i've progammed have been the other way around
00:43:45 Join Stryke` [0] (~Chairman8@resnet-241-86.resnet.UMBC.EDU)
00:44:00amiconnYou see, both the archos lcd and the iriver lcd in b&w mode have that format.
00:45:02amiconnLinus told that the iriver lcd in 4-gray mode uses a similar format as well, only that one bytes does no longer represent an 1x8 pixel column, but an 1x4 pixel one.
00:46:37amiconnThe gmini lcd (jyp ?) seems to use the same bitmap format, so I'd consider this a common format.
00:46:51preglowlcd's are generally compatible in a lot of ways
00:46:59thegeekhmm ,the old spectrum used a z80 cpu I think , there is a dynarec emulator for it called Spectrem-Dr
00:47:08preglowit's not unusual to find some types are completely compatible
00:47:24amiconnthegeek: ZX Spectrum == Z80 @ 3.5 MHz
00:47:39thegeekbut it uses the same instruction set?
00:47:50preglowthegeek: and bear in mind that the gb cpu isn't a z80, just a z80 lookalike, might still be a good starting point, tho
00:48:09thegeekI'm just doing whatever I can to help here
00:48:21preglowdidn't tell you to shut up, did i? ;)
00:48:36preglowtime for coffee and cigarette
00:48:40thegeekeven if the int. core is fast enought with rockbox fully optimized for iriver, dynarec would rock since you could perhaps do mp3+gamebou
00:48:48jypamiconn, gmini 220 uses a different format
00:48:51*linuxstb thinks of Chuckie Egg...
00:48:55*HCl sighs
00:49:22amiconnjyp: Really? How does that look like?
00:49:33jypActually many modes are possible
00:49:57amiconnIs there a datasheet in the wiki already?
00:50:12jypIn Archos firmware, bpp = 4, 2 pixels per byte
00:50:13thegeekwhat cpu does the palm use?
00:50:18jyporganized as lines
00:50:27preglowthegeek: arm, i think
00:50:29DMJChow is audio decoding coming?
00:50:31jypfully horizontal
00:50:39preglowmost portable things use arms
00:50:45jypDatasheet isn't in wiki yet ...
00:50:47preglowno pun intended whatsoever
00:50:48jypI'm adding it
00:50:48amiconnold palms == dragon ball (68K derivate, much like coldfire)
00:51:19thegeekgameboy emu for palm
00:51:22thegeekthere is a comment there
00:51:33thegeekabout how the 68000 is only running at 16mhz
00:51:36HClis it open source?
00:51:39thegeekno idea
00:51:42HClno, thats the commercial one
00:51:44HCli already saw it
00:51:48thegeekbarf that then
00:53:02*HCl tries to understand this stuff.
00:54:00***Saving seen data "./dancer.seen"
00:54:27HCli think i somewhat get it.
00:54:36HClit might be possible to make it do scancolumn
00:55:41ashridahheh. i've still got a palm m100 lying around here someplace. haven't used it in ages, since i tried to make sense of the prc build environment under linux, and got hopelessly lost in the resource stuff.
00:56:06preglowdoes anyone know how the hell to stop windows from scanning my h120 whenever i connect it by usb?=
00:56:13preglowit's irritating the fucking hell out of me
00:57:33 Quit ashridah ("Reconnecting")
00:57:35 Join ashridah [0] (
00:57:49ashridahpreglow: tweakui should be able to
00:58:02preglowi'll try it
00:58:21ashridahit's under 'my computer'->'autoplay'
00:58:35ashridahyou should be able to disable it for the drive the iriver usually shows up as
00:58:52ashridahor for 'removable devices' in general, although i'm not sure if that's worked for me
00:58:58muesli-i have tweakui installed
00:58:58muesli-if anyone finds a solution i'd like to know it as well
00:59:15*ashridah pokes
00:59:27preglowashridah: which tweakui do you speak of, btw? there are quite a few programs by that name
00:59:42ashridahpreglow: 'powertools for windows xp'
00:59:43muesli-i only know the one from m$
00:59:43*HCl sighs
00:59:47ashridahit's an ms-made app
01:00:17preglowfound it
01:00:55ashridahyeah. turning off the drive letter that the device shows up as here (i left a gap between my cdrom drives and my hard drives so that it'd always show there) did the trick for me
01:01:07ashridahdoesn't even pop up an explorer window
01:01:23preglowashridah: thanks, man, you cut my stress level in half
01:01:33ashridahyeah. i hate the autoplay crap
01:01:36HClthis is confusing x.x
01:02:06preglownow if i could only get xp to not require me to do that clicking shit before i can unplug my h120
01:02:18preglowi probably don't have to, since they have disabled delayed writes for portable disks
01:02:21preglowbut still
01:02:28sneakumsit sure does like to whine
01:02:51sneakumsalthough you haven't lived until you've seen a machine cripped by ~1500 'device disconnected' dialogs when a storage array has gone offline
01:02:57amiconnjyp: That's a pretty complete system-on-chip. USB, lcd controller, SmartMedia, RTC...
01:03:04sneakumsi think we had to power-cycle it
01:03:34ashridahsneakums: could have been worse. could have been nfs :)
01:03:40jypamiconn, indeed. It still amazes me ;):
01:04:21 Quit muz ("CGI:IRC")
01:06:28HCli don't think i can rewrite this to scancolumns
01:06:35HClits too much confusing code thats written without comments
01:07:20jypHCl, it's still time to re-target to gmini 220 ;))
01:07:42thegeekdon't give up on the gbemu
01:07:49thegeekwe all want gb on our irivers
01:07:50DMJCiriver pwns j00
01:08:32jyphahahaaaa you all iRiver n00bs pity me! :P
01:08:52TangAbout wavpack great news:
01:09:06TangRoberto amorim (thanks to him)
01:10:00Tangtold me he mailed the wavpack author
01:10:18DMJCyay doom3 addon
01:11:00*jyp melts himself
01:11:03 Quit jyp ("poof!")
01:11:51preglowhaha, i just noticed the optical output is enabled with rocbox
01:13:59Tanglol will drain more battery power
01:14:35linuxstbTang: Thanks for the update. It's good to know that the wavpack author is approachable in case of problems.
01:15:00TangIndeed it's very cool
01:15:36Tangmoreover seems this guy has skill in embedded development)
01:16:00linuxstbI'm sure lots of codec authors will be interested in this project - apart from FLAC and OGG, I can't think of any "non-mainstream" codecs with hardware support.
01:16:26HCl*NO ONE* has ever attempted to optimize the cpu core of gnuboy u.u
01:17:16linuxstbHCl: What % of real-time do you think gnuboy is at the moment?
01:17:23thegeekbut there is a x86 asm core?
01:17:31thegeekfor gnuboy I neab
01:17:46thegeeksurely that = optimizing?
01:17:46Tangindded linuxstb that was why i wanted to try approaching the HA community with Rbx... :)
01:18:03HCllinuxstb: it'd need to be near 60 times faster
01:18:19HClthere's an extreme amount of loss in the lcd driver
01:18:29HClmostly because the iriver lcd/rockbox require you to do bitsets
01:18:32HClrather than byte
01:20:02HClthere's an x86 cpu asm core written entirely in asm..
01:20:07HCli'd almost call it... dumb.
01:20:16HClbut i guess dynarec wasn't really invented back then
01:20:43DMJChow long ago?
01:20:50DMJCfx!32's been around for ages
01:20:53preglowdon't give up yet, probably won't be long till the coldfire is running at full throttle, then you can sigh and say 'ITS HOPELESS'
01:20:55HClgnuboy was written in 2001
01:21:02HCli'm not giving up
01:21:03HClat all.
01:21:18HCli'm just saying, that i don't see an easy way to make it produce scancolumns
01:21:22preglowaren't there any other emulators out there, btw?
01:21:26DMJCwhat speed does the coldfire normally run at when it runs the iriver firmware?
01:21:30HClunless you like
01:21:35HClcapture the scanlines in a buffer
01:21:37preglowDMJC: probably 140mhz
01:21:39amiconnHCl: Dynarec was already invented back then...
01:21:42HClthen when you receive scanline 143
01:21:47HClpush the buffer to rockbox
01:22:00DMJCso why does rockbox run it so slowly?
01:22:02HClwhat would be ideal, is to get the lcd to accept a byte array with the lower two bits being its value for each pixel
01:22:12HClin the current format
01:22:13preglowDMJC: because the cpu isn't clocked at more than 10mhz
01:22:15HClone byte
01:22:26HClis for 8 pixels / scanlines
01:22:29preglowDMJC: it's software controllable
01:22:50DMJCand it has a default slow state until it receives a "fast" command
01:22:51HClso it has to set bits individually
01:22:59HClwhich is expensiveish
01:23:00preglowDMJC: exactly
01:23:29linuxstbIsn't the Rockbox LCD driver about to change anyway to support 2-bits/pixel? Maybe 1-byte per pixel can happen then.
01:23:33HClthe idea is that it can be sped up a lot
01:23:42HCllinuxstb: its more about the lcd hardware format..
01:23:57HCli'm gonna make an attempt to write a "better" driver
01:24:05preglowHCl: well, you should be able to hack gnuboy to do it a bit more efficiently, but again, scanning in the y direction will be a bit more expensive on the coldfire part of things
01:24:07HClit might be faster, who knows.
01:24:25preglowsince it's probably set up to scan x first internally
01:25:04amiconnHCl: You could buffer 8 scanlines in the driver, then convert at the end.
01:25:51amiconnThis would simply change to 4 once the rockbox lcd driver does 2 bit
01:28:35HClthats my plan
01:32:40HCli'm fearing this'll actually be slower, but who knows.
01:34:12amiconnWell, you'll still need bitsets no matter what. But as they can be put in a tight loop then, they're done in a register.
01:34:39amiconnAsm would help to speed it up even more (utilizing the carry bit)
01:37:04preglowHCl: getting a bit more than you bargained for, ehh? ;)
01:37:11HClpreglow: not really
01:37:26preglowno, i forgot, you initially wanted dynarec
01:37:39HClamiconn: i'm getting the feeling this'll be more efficient once rockbox has a 2bit buffer
01:39:03HClold routine took 30 seconds to get from the start of mario to display output
01:39:08HClnow to time my new routine :x
01:39:24*preglow holds breat
01:40:51HClgotta make it compile first >.o
01:40:57HClvariable name conflict
01:41:18*HCl compiles
01:41:29HCli think it'll actually be slower
01:41:32HClbut who knows.
01:43:59HClthat actually crashed it
01:44:13preglowand there was much rejoicing
01:45:02HCl :p
01:45:10HClwell ofcourse
01:45:12HCli'm dumb
01:45:14HCland stuff.
01:45:21HCli kind of forgot to increase the variable
01:45:26HClthat determines whether i get out of the loop
01:45:38preglownot entirely unessential
01:48:37HCli don't see how it actually makes code smaller
01:48:38HClbut ok
01:51:07HCli'm not sure if its any faster
01:51:24HCli still have to map 4 bits to 1 color
01:51:30HClso thats fairly expensive
02:00:14*HCl reads about the lcd driver and sighs.
02:00:23 Quit Hohoman ("( :: NoNameScript 3.81 :: )")
02:00:30HClcan't this thing scan horizontal instead of vertical?
02:02:31amiconnWhat difference would this make? If you buffer 8 scanlines, it should no longer matter.
02:03:21HCli guess
02:04:33*HCl ponders
02:04:59 Join webguest61 [0] (
02:05:08 Quit webguest61 (Client Quit)
02:05:18 Join toolmanwill [0] (
02:07:57 Quit toolmanwill (Client Quit)
02:16:16Tangi've to go
02:16:19 Quit Tang ("Chatzilla 0.9.66 [Mozilla rv:1.7.5/20041108]")
02:17:41preglowahh, great, fireworks, sleep will come easily tonight
02:25:05*HCl scratches his head
02:25:08HCli think its a little faster
02:25:09HClbut not much
02:26:43 Quit edx (Read error: 110 (Connection timed out))
02:29:51*HCl goes to sleep..
02:30:45amiconnNite HCl
02:31:26preglowi'll do the same
02:31:40preglownite all
02:31:52 Quit preglow ("rofllolmao")
02:33:10HClamiconn: i'll put my new implementation version on my ftp.. hold on
02:33:45amiconnArgh! Every time I fetched that zip blob, you're going to update afterwards...
02:33:58HClsorry :X
02:34:09HClits one of the reasons why its handier to have it in cvs :x
02:34:44HClanyways, thats the new version, its not that much faster, though the adding of the -O2 to the compile options helped a bit
02:35:01HClonce we have 2 bit color it should be a bit faster
02:35:41amiconnI need to cleanly integrate that into a build dir tree
02:36:07amiconnYou always distribute your whole dir, with all the junk in it...
02:36:13HClsorry :X
02:36:22HCli can zip up just my gnuboy dir from now on
02:36:27HClassuming you have all the other changes
02:36:44amiconnNot yet... have to figure what these are
02:38:00amiconnBtw, the gnuboy dir also contains some junk. There is a .tar.gz ....
02:38:12HClyou can do cvs update in the dir and it'll say what files are modified compared to the cvs
02:38:24HClthe .tar.gz is mostly for if you need to compare to the original sources
02:38:29HClit can just be removed
02:38:44HClso can the asm dir in gnuboy/ since it contains x86 asm anyways
02:38:55HCloh wait, i already deleted that
02:39:21HCli can clean it up a bit tomorrow
02:39:55HCland give you a dir with a clean cvs + changes and no other files
02:40:22HCli'm gonna sleep now
02:40:27amiconnI'll try a diff agains my cvs dir
02:40:49HCli can give you a list of modified files
02:40:49amiconn(hoping that you updated recently)
02:40:51HClin whisper
02:40:54HClif you want
02:54:04***Saving seen data "./dancer.seen"
03:12:57amiconnI should go to sleep now
03:13:13 Part amiconn
03:18:10 Join gromit`` [0] (
03:20:33 Quit ashridah ("out")
03:23:16 Join elinenbe [0] (
03:23:46 Part Quelsaruk ("Ooh, look, a shiny object...")
03:29:15 Join muesli|tarn [0] (
03:31:44 Quit gromit^ (Read error: 110 (Connection timed out))
03:31:49 Quit linuxstb (Read error: 60 (Operation timed out))
03:34:24 Quit Stryke` ("Friends don't let friends listen to Anti-Flag")
03:38:14 Quit muesli- (Read error: 60 (Operation timed out))
03:46:05 Join orbital [0] (
03:47:04 Part orbital
03:52:28 Quit gromit`` ("Leaving")
04:01:21 Join diprospero [0] (
04:01:48 Quit diprospero (Client Quit)
04:01:55 Quit cYmen ("dying ....slowly")
04:01:56 Join webguest78 [0] (
04:03:23 Part webguest78
04:06:30 Join QT [0] (as@area51.users.madwifi)
04:08:03 Join edx [0] (
04:14:38 Join toolmanwv [0] (
04:19:21 Quit QT_ (Read error: 110 (Connection timed out))
04:20:11 Part toolmanwv
04:20:40 Join toolmanwv [0] (
04:23:40 Part toolmanwv
04:23:43 Join toolmanwv [0] (
04:46:02 Join Strath [0] (
04:46:11Strathhey guys ;)
04:48:01Strathi was just looking around, i think there may be an error on the page reguarding rel="nofollow"
04:49:25Strath(the last four links)
04:54:08***Saving seen data "./dancer.seen"
05:18:10 Join webguest99 [0] (
05:49:50 Quit webguest99 ("CGI:IRC (EOF)")
05:51:27 Quit DMJC ("Leaving")
06:54:12***Saving seen data "./dancer.seen"
07:28:23 Join methangas [0] (
07:33:03 Quit XShocK (" HydraIRC -> <- Go on, try it!")
07:37:18 Join midk [0] (
07:38:17 Join R3nTiL_ [0] (~zorroz@
07:46:38 Quit pill (Read error: 110 (Connection timed out))
07:56:47 Join DMJC [0] (
07:57:48 Quit R3nTiL_ ()
08:16:47 Join pill [0] (
08:24:09 Quit midk ("Leaving")
08:54:15***Saving seen data "./dancer.seen"
08:57:03 Join amiconn [0] (
09:36:22 Nick Lynx_awy is now known as Lynx_ (HydraIRC@
09:45:09amiconnrasher: I had a look at your logo plugin
09:45:25amiconnLooks pretty clean, just some comments:
09:46:12amiconn(1) HAVE_LCD_BITMAP is used as a boolean define, so line 21 should read #ifdef HAVE_LCD_BITMAP
09:46:51amiconn(2) The variable button definitions are not really variable, so you could just use BUTTON_OFF directly
09:47:20rasheroh right, I thought about that, and then forgot
09:47:21amiconn(3) There is no reason why it should not run on gmini (simulator for now)
09:47:45amiconn(This would be solved automatically if (2) is resolved.
09:47:52rasherdid I exclude it?
09:47:55rasheroh right, button defs
09:48:08rasherI'll make a newer version
09:48:11amiconnNo, but you didn't include the variable button definition for it
09:48:38rasherwell that'll come as a result of (2)
09:48:45rasherassuming gmini has a BUTTON_OFF
09:48:53amiconnThe gmini also hast LEFT/RIGHT/UP/DOWN/OFF, and the lcd is almost the same size as the other archos units
09:49:12amiconn(128x64 instead of 112x64)
09:49:40amiconn(4) Do you plan to make the speed variable? I saw that 'timer' variable...
09:50:00rasheroh, that was just a leftover from mosaique on which I based it :)
09:50:20rashernot really
09:50:55rasherso.. remove button definitions.. change to #ifdef
09:51:00amiconnFor (1), the 2 #if conditions could be combined to one.
09:51:13rasherAh, how does that look?
09:51:25amiconn#if defined(HAVE_LCD_BITMAP) && (LCD_WIDTH > 90)
09:51:38amiconn>= 90
09:51:45amiconnof course
09:51:55rasher91, I guess
09:52:05rasher90 was from a previous version of the small bitmap
09:52:13rasherwhich was slightly wrong-looking
09:52:19amiconnHmm, then correct thios as well ;)
09:52:39amiconn> 90 would be correct then
09:52:50rasherI think >= 91 makes more sense
09:52:58rashersince 91 is the number in question
09:53:45*rasher makes iriversim and recorderv2sim
09:55:07 Join midk [0] (
09:55:15 Join midk_ [0] (
09:55:15 Quit midk (Connection reset by peer)
09:57:16rasherlooks fine on those sims
09:57:27rasherI'll make a new patch then
09:57:42 Join midk [0] (
09:57:43 Join midk__ [0] (
09:57:45 Quit midk (Read error: 104 (Connection reset by peer))
09:57:47 Quit midk__ (Read error: 104 (Connection reset by peer))
09:57:54rasheror maybe I'll just edit the original patch by hand
09:58:16rashershouldn't be too hard
09:58:48amiconnFor a new plugin, just submitting the .c file is okay.
10:00:27amiconnA .ptach only makes sense if you change existing files, imho.
10:00:33amiconn.patch even.
10:05:13*rasher uploads the new version
10:05:53rasheror rather, deletes old version
10:05:56rasherTHEN uploads new version
10:10:27rasherI wonder if I got the PluginIndex page right
10:10:36rasheror rather, if it catches all of the %Y%s
10:12:06rasherI wonder if there are a lot of orphan pages
10:16:35rasheramiconn: do you know about the stopwatch plugin?
10:18:47rasherinstead of "Recording" on the main menu, I have "Max range"
10:18:57rasher"Maximum of range"
10:19:08rasher(the Danish translation, that is)
10:19:54rashermust be a glitch in the build I have on my device, looks right in the sim
10:20:21rasherhuh, I get it for English as well
10:20:24rasheroh well
10:20:50rasherat least it's consistent
10:25:35amiconnThis means your language strings are off by one (just looked up in english.lang)
10:26:12amiconnI have no idea why this is. It's working correctly here.
10:26:31 Join webguest36 [0] (
10:28:14 Join Nibbler [0] (
10:28:18rasherYes, I don't know.. I don't remember fiddling with the languages
10:28:25rashermaybe I have
10:28:38rasherWorks in the sim, and probably would work if I built and installed again
10:28:55 Part webguest36
10:29:09*rasher tries
10:31:26 Join webguest73 [0] (
10:32:39 Quit midk_ ("Leaving")
10:33:08rasherah, I might have an explanation
10:33:21rasheror at least, I might know what I did to make it happen
10:34:05 Join midk [0] (
10:34:35rasherlet me check this..
10:36:11rasherI chose '4. dansk' as default language in configure
10:36:40amiconnHmm, that's strange.
10:37:32amiconnIt should work correctly. There was a bug in the script that generates lang.[ch], but I fixed that. At least I think that I did...
10:37:48rashercould it be that I've simply messed up dansk.lang?
10:38:20amiconnI don't think so. The ID order is always taken from english.lang
10:38:30amiconnCould you please send me the generated lang.c and lang.h?
10:38:35rashersure"> and .c
10:49:58amiconnHmm. There is indeed a missing id. Need to check why; this shouldn't happen
10:50:32rashercan you see which one?
10:50:57amiconnBut it should get included from english.lang instead...
10:51:16amiconnLooks like a bug in one of the scripts, or this one really messed up.
10:51:30rasheroh, it's not in dansk.lang
10:51:37rasherI wonder how that managed to get away
10:51:41 Nick Strath is now known as StrathAFK (
10:51:46 Join loopz [0] (
10:51:51rasheroh, it's there
10:51:58rasherjust not where you'd expect it to be
10:52:05rasherwhat the..
10:52:12amiconnYes, and the order messed up in lang.h as well
10:52:27amiconnThis should never happen, so there is a bug in the script...
10:52:53rasherit could be that I messed up the original?
10:52:55rashermanually, that is
10:53:05rasheror should it cope with that?
10:53:38amiconnThe ID order in other .lang files than english.lang shouldn't matter.
10:54:19***Saving seen data "./dancer.seen"
10:55:04*amiconn is building a dansk recorder simulator...
10:56:19rasherthat should prove amusing :)
10:57:20amiconnHmm, looking correctly here.
10:58:22amiconnHmm, not really.
10:58:46amiconnWhile the menu is looking right, the lang.h order is wrong (same as yours)
10:58:56rasherthe only place I've noticed a fault is in the main menu
10:59:09rasherwhere "Recording" is changed to "Maximum of range"
10:59:18rasherall other places look fine
10:59:28rasherI think
10:59:34rasherhaven't noticed other places at least
10:59:35amiconnThat's clear; I think you have a dansk.lng on your unit as well (with the 'official' id order)
11:00:05amiconnSo if this is loaded into the build which expects the wrong order, it'll mess up
11:08:04amiconnHmm. This cannot work without a major rework of the involved scripts.
11:09:42amiconnI didn't run into this, because I always select english as the build language, and then load deutsch.lng
11:09:46rashermaybe We should just fix dansk.lang and look the other way :)
11:10:07rasheryes, that's how I usually do it
11:10:21amiconnIt will definitely work if the IDs are in the same order as in english.lang
11:10:39amiconnGerman, french and spanish are
11:11:00rasherso it's just a matter of moving LANG_WEEKDAY_THURSDAY up between WEDNESDAY and FRIDAY ?
11:11:09amiconnHowever, there's a number of other languages, which are probably in the wrong order as well
11:11:21amiconnNo, there are other IDs shuffled as well
11:12:24amiconnIt shouldn't be that hard to change uplang to sort the IDs, but this would make it difficult to preserve comments.
11:12:38amiconn(and it's perl, grr)
11:13:10rashermaybe it should only do the sorting when compiling?
11:13:21rasherSuch that the order will be fixed at that point
11:13:24rasherinstead of in the .lang
11:13:32amiconnIt would have to work the other way round than now.
11:14:26amiconnCurrently, it first reads english.lang and stores it in a hash. Then it reads the selected.lang, updates/checks with the english IDs, then writes the new file, ID by ID
11:14:27 Quit loopz ("CGI:IRC (EOF)")
11:15:05amiconnThis way, the order of the generated file is always in the selected .lang order, except for IDs that are missing there
11:15:48amiconnIt would have to work the other way round, i.e. first reading the selected .lang, then going through english.lang ID by ID
11:16:10amiconnHmm. As I think it over, this is the more logical way...
11:21:24 Join linuxstb [0] (
11:31:06amiconnGotta go now.
11:31:16 Part amiconn
11:32:31 Quit webguest73 ("CGI:IRC (EOF)")
11:33:03 Join cYmen [0] (
11:48:05 Join Quel|out [0] (~kvirc@
11:48:18 Nick Quel|out is now known as Quelsaruk (~kvirc@
11:57:24muesli|tarng'day mates
12:01:14rasherHrm, shouldn't just link to the wikipage of ircnicks?
12:01:21rasheror at least
12:01:30rasherbe updated with the nicks from ircnicks
12:04:18Quelsaruki suppose
12:04:24Quelsarukask Bagder :)
12:05:55 Join ashridah [0] (
12:06:14 Quit xen` (Read error: 110 (Connection timed out))
12:06:31rasherBagder: shouldn't just link to the wikipage of ircnicks, or at least be updated with the nicks from IrcNicks?
12:06:50 Join alxcm [0] (
12:07:07alxcmhey all
12:07:27alxcmwhat's up?
12:13:29 Quit alxcm ("leaving")
12:16:16rasher"the latest release of libFLAC has a compiler define (FLAC__INTEGER_ONLY_LIBRARY) that builds the whole library (decoder, encoder, etc.) with integer only"
12:25:01Bagderrasher: turning that into a link to the wiki
12:25:48linuxstbrasher: Yes, saw that (and am using it).
12:26:10rasherstill, the 64-bit ints :\
12:26:14 Join jyp [0] (
12:30:00rasherhi there
12:30:12jypho ho ho
12:48:23HClits santa!
12:48:52HClwhatcha gonna get me for christmas :x i've been good! i built a gameboy emu for rockbox :x
12:50:35 Join GnagelRam [0] (
12:50:59 Quit Quelsaruk (Read error: 104 (Connection reset by peer))
12:53:07linuxstbI've got libmad and libFLAC running as viewer plugins on the iRiver, but I'm getting some strange crashes - during read/write accesses and after the plugin returns control to Rockbox.
12:53:27linuxstbDoes anyone have any debugging suggestions?
12:53:32rashersounds good (and not so good)
12:53:55linuxstbThey both work perfectly in the X11 Simulator.
12:54:23***Saving seen data "./dancer.seen"
12:54:32rasherCould be ata strangeness perhaps?
12:55:10linuxstbrasher: Maybe, or maybe my plugins are overwriting memory used by Rockbox - but I would expect that to crash the Simulator as well.
12:56:25linuxstbIf you recall, my FLAC viewer crashed (never returned from a call to rb->read()) if I was reading from the file as I was decoding it. But when I changed to reading the whole file into memory in advance, then all the reads worked fine.
12:56:33 Join webguest93 [0] (
12:57:08linuxstbNow the same happens with calls to rb->write if I am trying to write the output of the decoders to a file on disk - the LED stays on forever and the write() never returns.
12:57:41linuxstbIt does all seem to be disk related - I've reformatted my iRiver, but that didn't change anything.
12:58:21linuxstbHCl: Did you have any problems when rockboy was writing debugging information to a file?
13:01:10 Quit webguest93 (Client Quit)
13:01:21HCllinuxstb: yes, actually
13:01:23linuxstbThe crash in the rb->write() call happens on about the 3rd or 4th call - possibly when the write buffer is full and Rockbox actually needs to access the disk.
13:01:32HCli never even managed to get it to write to a file
13:01:39HClprobably because i messed up the first 3 times
13:01:51HCland after that it proved unneeded
13:01:59HClsince i had gotten the lcd driver to work
13:03:00linuxstbHCl: And I assume your only reads are to read the entire ROM into memory at startup?
13:03:10HClpretty much, yes.
13:03:21HClbut eventually i'm gonna want to store savegame data on disk
13:04:36linuxstbWould you mind putting some test writes (e.g. writing about 512KB of random data to disk) into your program. Ideally these should run about 30 seconds after the reads finished to duplicate my behaviour.
13:05:12HClmaybe once i'm awake :X
13:06:19linuxstbHCl: I'll download your source and try it myself.
13:10:40 Join amiconn [0] (
13:20:04rasherlinuxstb: another interesting fact - the batterytest plugin crashes as well
13:20:08 Join Bagder_ [0] (
13:21:33amiconnrasher: Battery_test writes a big file, then reads it at regular intervals. Plus it writes a .log file...
13:21:53amiconnsort.rock does work?
13:22:34rasherDoesn't happen here afaics. When I start battery_test, the iRiver stops responding. It creates a battery.dummy in the root-dir, this file is empty.
13:23:06rasherHm, I don't know about that actually. I just added it in the wiki because I assumed it did
13:24:27rasherHow do I check?
13:24:58amiconnsort.rock should sort the all lines of a file alphabetically.
13:25:32amiconnIt works as a viewer plugin, i.e. "Open with.." -> "sort" on a file should sort that file
13:25:48rasherlooks like it worked
13:26:03rasheralthough for the file I tried it with, the lines were already in order
13:26:04amiconnMaybe it depends on the filesize
13:26:13rasher(and only 2 lines)
13:26:27amiconn(assuming there are problems with file read/write on iriver)
13:26:56rashersort went trough reading/writing just fine
13:27:42amiconnBattery test should create a big file (battery.dummy) in the root. This file will be the same size as the free RAM, i.e. ~31 MB on iriver.
13:28:11amiconnThen it opens a log and writes out the battery status and current time both to the log and the screen
13:28:37rasherthat never happens
13:28:51rasherit only gets as far as to create battery.dummy afaics
13:28:59rasheror maybe I should let it try for a while
13:29:01amiconnAfter the time has passed that the buffer runs empty if it were a 128 kbps MP3, it re-reads the file and writes a new log entry
13:29:29*rasher starts batterytest
13:29:29amiconn(Thereby simulating the harddisk load pattern of mp3 playback)
13:29:49amiconn31 MB should take a long time to drain on iriver...
13:31:10rasherit exit
13:31:31amiconnIf you press off, it should exit, otherwise not
13:31:41rasherI didn't
13:31:58rasher-rwxr-xr-x 1 rasher trusted 32M Aug 1 2003 battery.dummy
13:32:00linuxstbWell, I can't make rockboy crash with disk writes. I'm going to try to write a standalong plugin based on helloworld.c and see if I can make it crash with disk accesses.
13:32:02rasherwell at least that's correct
13:32:16rashermaybe it crashes in the read?
13:32:48amiconnNo battery.log?
13:33:01linuxstbIt could just be a bug in my plugins (or my build commands).
13:33:25amiconnrasher: Maybe it exits because the battery state isn't properly detected yet.
13:33:38rasheramiconn: ah, that might very well be it
13:33:40amiconnThere is a check that makes it exit if <5% battery left
13:33:51rasherthat's probablly it then
13:34:00amiconnbattery_test.c, lines 79..81
13:34:35amiconnOkay, so battery_test simply does not work (yet), because the battery state check is wrong
13:35:18*rasher turns off that check and builds again
13:38:03*rasher starts battery_test
13:39:37rasherit's working now
13:39:50rasherafter 1:40 - battery - 1% :)
13:42:10rasher31mb 128kbps mp3..
13:42:17rasherI could be waiting for that for a while
13:42:49Bagder_about 30 mins or so I'd guess
13:43:00linuxstbI'm not sure, but I think my plugins have decided to work now...
13:43:48linuxstbCould it be something to do with the battery indicator, causing problems with the disk accesses?
13:44:41amiconnrasher: ~33 min
13:45:20rasherlinuxstb: shorten is both lossless and lossy apparently
13:45:25rasherbut it looks like the license is horrible
13:45:47rasherlet's see about free implementations
13:45:47linuxstbrasher: Yes, and so is WavPack....
13:46:43sneakumsah well, we'll always have paris^Wflac
13:47:56*rasher stares blankly at sneakums
13:49:43linuxstbYes, my libmad decoder is writing to disk OK now. I've just played (on my PC) two seconds of audio decoded by libmad on the iRiver.
13:50:29rasherlooks like the shorten source is under some silly non-commercial license, and no Free implementations exist
13:51:15rasherand a horribly vague license at that
13:51:37linuxstbShorten is rapidly being replaced by FLAC as the de-facto lossless standard (with Monkey's and WavPack being minority formats). But for example, still use Shorten for their downloads.
13:52:10linuxstbSo Shorten support would be nice, but if the license doesn't allow it, then it's no big loss.
13:52:25Bagder_got a URL for the shorten license?
13:52:26rasherWell we have to use a gpl compatible license don't we?
13:52:32DMJCwhat about matroska audio support?
13:52:33Bagder_yes we do
13:52:38linuxstbUnless we decide we want to make a GPL exception for audio codecs.
13:52:38Bagder_we distribute binaries
13:52:58rasherthat's not the primary shorten site, but the license is correct
13:53:38Bagder_its short at least ;-)
13:53:39rasherlinuxstb: relicensing to get shorten support.. sounds like a poor choice
13:54:18linuxstbrasher: iwe may have problems with other audio codecs as well. I'm not saying I'm in favour, I'm just mentioning the option.
13:54:58rasherlinuxstb: doesn't that all fall down the moment we include outside gpl code?
13:55:00Lynx_DMJC: istn't matroska just a container format?
13:55:07rasheras in, don't we lose that option
13:55:31Bagder_rasher: that is indeed a grey area
13:55:40linuxstbI'm not an expert, but I'm sure the Linux Kernel has an exception to allow non-GPL'd kernel modules.
13:55:40DMJCnot sure, but there is a matroska audio format
13:55:50Bagder_it isn't much researched, as people tend to avoid that thinking
13:56:02Bagder_linuxstb: actually no
13:56:12Bagder_linuxstb: its just Linus who says he doesn't care
13:56:26Bagder_Torvalds that is
13:56:49linuxstbBagder: So, strictly speaking, binary modules infringe the kernel license?
13:56:50rasherAlright. I'm no expert by far, it was just my understanding that by including other people's GPL code, you give up the ability to decide your own license (because then you'd be in violation with the GPL of the included code)
13:56:59Bagder_linuxstb: yes, according to FSF
13:57:14Lynx_DMJC: from the site: "Matroska, the extensible open standard Audio/Video container"
13:57:18linuxstbrasher; Yes, that's true, I think every contributor would need to agree.
13:57:58rasherlinuxstb: and contributors might be something like "anyone involved in libmad" (assuming libmad is GPL, which I've no idea if it is)
13:58:03ashridahrasher: that doesn't mean you can't license your own code with two different licenses
13:58:19Bagder_rasher: no, only the ones owning the copyrights in libmad
13:58:33linuxstbI'm sorry I mentioned this now...
13:58:40Bagder_the copyright owner can relicense at will
13:59:17*linuxstb goes to read GPL faqs
13:59:44rasherashridah: but for the non-GPL licensed code, you'd have to rip out everything GPLd
13:59:50rasherwhich would be a pain :\
13:59:54Bagder_but I strongly think we should avoid adding exceptions to the GPL license
13:59:54rasherI see why people avoid this
14:00:35DMJCjust do what nvidia do
14:00:51Bagder_they provide a binary
14:00:55ashridahBagder: particularly since the GPL code you link in needs to have an exception to YOU
14:00:56Bagder_which violates the GPL
14:00:58rasherand some Free glue
14:01:10Bagder_ashridah: yes
14:01:11DMJCthey use lgpl code the bridge afaik
14:01:18 Join preglow [0] (
14:01:33Bagder_DMJC: that doesn't help, it is still technically a violation
14:01:40Bagder_according to FSF's interpretation
14:01:51Bagder_but Linus, as copyright owner does't persue
14:01:55DMJCbah, screw stallman, go with the practical option
14:02:27DMJCdon't the rockbox developers own the copyright to their own code?
14:02:35Bagder_DMJC: yes of course
14:02:45preglowDMJC: you always own the copyright on your own code
14:02:50preglowDMJC: unless you sign it away
14:03:01DMJCso just choose not to pursue it?
14:03:20rasherWell, say we include a GPL codec
14:03:25preglowDMJC: practical is not always enough, i'm afraid, some people have to be forced to cooperate
14:03:25rasherand a non-gpl codec
14:03:42rasherthe authors of the GPL codec would need to agree to this
14:03:54Bagder_rasher: yes
14:04:04linuxstbrasher: What if the GPL'd and non-GPL'd codec were not linked to Rockbox at the same time?
14:04:04DMJCmyself, i'd have bsd licensed it, I mean it's not like someone could ever make this stuff worth billions of dollars heh
14:04:09Bagder_it would mean pain, trouble and more crap
14:04:23preglowDMJC: with a codec, i agree with you
14:04:26rasherDMJC: that doesn't change anything
14:04:36rasherDMJC: assuming you mean that rockbox should be bsd
14:04:51rasheroutside (GPL) code could still create license problems
14:04:53Bagder_if Rockbox were BSD, we'd avoid some problems
14:04:58preglowbut why is this an issue, we aren't going to go into some licence trickery, are we?
14:05:14Bagder_it isn't really an issue
14:05:18rasherpreglow: read the irclog for 30 minutes before you entered :)
14:05:25preglowyes, i did
14:05:26Bagder_until we want to use codecs that aren't incompatible with GPL
14:05:33preglowbut that's not an option anyway, is it?
14:05:47rasherit is.. given a number of things happen
14:05:49Bagder_well, that's what we discuss :-)
14:05:56rasherbut it's all a big mess
14:06:02preglowyou're opening a can of worms, that's for sure
14:06:11Bagder_the thumb rule must be: no GPL exceptions
14:06:14rasherand the only codec right now is shorten, which isn't that interesting
14:06:35preglowthere are great big lists of which licenses are compatible with gpl
14:06:40rasherSo I think the conclusion is: Rockbox doesn't care right now.
14:06:40amiconnpreglow: An unrelated question, out of curiosity: Is your norsk translation nynorsk or bokmål?
14:06:46preglowamiconn: bokmål
14:06:58linuxstbMonkey's Audio iis interesting, but I think the processor requirements make the license problem a non-issue.
14:07:14preglowamiconn: though it probably isn't that big a difference
14:07:58linuxstbShould I mention patents? What's the law in Sweden?
14:08:05Lynx_linuxstb: well, practically support for minor lossless codecs is not that important anyway, as they can be converted to flac for example without loss
14:08:19preglowahhh... patents....
14:08:43rasherLynx_: Very good point :)
14:08:44preglowby the way, when you buy an mp3 player, who has the license to play back mp3s? the company that made the player or you?
14:08:56linuxstbBut if the worse happens, we are still lieft with OGG and FLAC - not that bad a situation.
14:09:09preglowlinuxstb: heck, that would cover all my streaming music needs
14:09:21 Quit Bagder_ ("Leaving")
14:09:29preglowi'm almost running vorbis exclusively as it is
14:10:44Lynx_i'd guess out of all iriver owners less than 10% use ogg, and way less would use flac
14:10:59rasherI think ogg usage is a bit larger
14:11:23Lynx_actually i have not met anyone that has his/her main music library in flac, unless they need high quality stuff because they are musicians or such
14:11:51rashermainly because iRiver is one of the few players with ogg-vorbis, so there'd be a rather large concentration of ogg users compared to general population
14:11:53preglowi'd use flac mainly for music production
14:11:54Lynx_rasher: hmm, it may be on the iriver, because that is a selling point?
14:11:58linuxstbLynx_: I do.
14:12:04ashridahLynx_: i know someone who converted all his stuff to flac when the company who made his player added support for it (ie, he reripped all his music)
14:12:05rasherlinuxstb: exactly :)
14:12:06preglowi also keep my ripped vinyls as flac files
14:12:31Lynx_hmm, ok, my sample is probably too small ;)
14:14:05Lynx_googling for file:.flac gives 1390, for .ogg 6950 and for .mp3 351000 results
14:14:38linuxstbGoing back to my iRiver problem - I'm still getting crashes sometimes. But other times, the disk writes are working perfectly. Either it will crash on the first disk access, or it will work fine for them all.
14:14:52rasherOh dear
14:14:57rasherbattery_test crashed
14:15:20Lynx_linuxstb: does the false battery status go up and down, or stay more or less the same?
14:15:22rasherat 00000002
14:15:42linuxstbIt may have something to do with USB - it seems to always crash if I run it immediately after disconnecting. If I power down, then restart, it seems to work.
14:15:45preglowlinuxstb: i didn't do anything but write files yesterday, never saw a crash
14:16:12preglowlinuxstb: and a LOT of usb connects
14:16:30preglowlinuxstb: that is, never saw a crash once i took my huge file buffer off the stack :P
14:16:40linuxstbLynx_: At the moment, the indicator says ?
14:16:55linuxstbLynx: But it is now flashing empty.
14:17:07rasherBattery_test said 1%
14:17:40rasherhrm, maybe even -1%
14:17:52rasherwhich is a stunning number
14:17:54Lynx_linuxstb: i meant: is it determined at the beginning, and then if it is 'high' all your disk writes work, and if it is determined 'low' at the beginning, the first one crashes already. also, after usb connect the battery may be especially low.
14:18:28amiconnbattery == -1% means "not yet known" in rockbox
14:18:59amiconnThe power thread first measures for a while until it sets the real status
14:19:33*amiconn remembers something...
14:19:40Lynx_does rockbox shutdown the box with low battery, or does it just physically run out of juice?
14:21:04linuxstbAre there any battery-level checks in the ATA code that I could remove?
14:23:23amiconnNope, I just checked
14:24:01amiconnI remembered that there might be such check, as we have a function for that: bool battery_level_safe()
14:24:32amiconnImho the battery state computation should be fixed. Shouldn't be that hard.
14:25:00amiconnBtw, amyone here who is familiar with X11 programming?
14:25:15*HCl quietly reads along
14:28:00linuxstbamioconn: Not for about 15 years - what do you want to do?
14:28:18amiconnI'm currently looking into the X11 simulator
14:28:47 Join R3nTiL_ [0] (~zorroz@
14:28:48amiconnThere are button handling problems which can imho only be solved if we use an interval timer
14:29:17amiconnCurrently the X11 simulator is completely non-realtime
14:30:16 Quit jyp ("poof!")
14:30:24amiconnI already found XtAppAddTimeOut(), but I'm not sure whether I'll get this to work. I also don't know yet where to put it.
14:30:41amiconnThe whole X11 sim source looks a bit messy to me...
14:31:18amiconnI would then also need to disable X's keyboard auto-repeat
14:31:49linuxstbamiconn: I can't help with any of that. I think I've forgotten anything I ever knew about X11.
14:32:24 Quit DMJC ("Leaving")
14:35:35preglowand i just plain hate x11
14:37:22amiconnlinuxstb: Maybe your random crashes are because of a stack overflow? I just saw that the stack overflow check in the thread scheduler is disabled for iriver & gmini
14:37:31amiconn(Btw, I wonder why this is)
14:38:30preglowgood question
14:38:37preglowhow large is the stack allowed to be?
14:38:40preglownormally, that is
14:39:07linuxstbCould I increase the stack to a huge size to check if that changes anything?
14:39:28preglowsmashing the stack boundary sure does weird things
14:39:36preglowmy h120 would never come out of usb mode when i managed that
14:39:44preglowand it would write partly garbage
14:40:05linuxstbYes, I don't know what the stack usage of libFLAC and libmad is like, so that's a possibility.
14:40:10amiconnThe standard thread stack size is 1024 bytes. The main thread (where plugins run in) has a bit more - 8 KB iirc
14:41:11amiconnThe USB thread get 3 KB - it did overflow before...
14:42:06linuxstbamiconn: Should I try changing the 0x2000 in line 61 of firmware/ (is that the right place?)
14:42:58amiconnI think it is. You could also try to enable the overflow check in thread.c , but I'd ask Linus first why this is disabled
14:43:32linuxstbI'll try adding a zero to the stack size and see what happens.
14:44:26HClrockboy crashes now and then, i wonder whether its related
14:45:16preglowas a matter of fact, it would not surprise me at all if this is true
14:46:12linuxstbWell, my first test after increasing the stack size seems to be working. I'll try some more...
14:46:36HClstack size of 1024 bytes? O.O;
14:46:44preglownot for plugin
14:47:02amiconnHey, you're programming an embedded system.
14:47:06preglowbut yes, i quite happily smashed my stacks with a not very large number of kilobytes
14:47:23preglowand it did give rise to quite strange things, heh
14:47:33HClamiconn: shall i clean up rockboy's code some more?
14:48:16amiconnI did not get it to compile on the sim so far. Didn't investigate yet...
14:49:05amiconnIt complains about conflicting types for vsnprintf() - that's only a warning. It errors out on not finding widgets.h
14:51:11linuxstbNo, increasing the stack size to 0x20000 didn't seem to solve the problem, but it's possibly working slightly better. I'll try increasing it even more.
14:53:41linuxstbI've enabled the stack-overflow check in thread.c, let's see what happens.
14:54:10HClwhat does the stack overflow check do?
14:54:27***Saving seen data "./dancer.seen"
14:54:57linuxstbHCl: It causes my iRiver to crash immediately on startup.... :-(
14:55:02HCl :X
14:55:09HClbricked? :X
14:55:17HClor wait, you didn't reflash..
14:55:29linuxstbNo, iRiver firmware is loading fine.
14:55:47linuxstbBest not to touch what I don't understand....
14:57:59HClamiconn: shall i clean up the code of rockboy some? or..?
14:58:18amiconnI think leave it as-is for now.
14:58:50HCli was just hoping it could be allowed in cvs since it makes it easier for more people to work on it if they want to :X
15:00:07preglowHCl: have you modified the gnuboy code much?
15:00:38HCla fair deal, yea, i can make you a .patch if you want to?
15:00:55HClmostly some speed optimizations and a completely different way to do screen updates
15:00:58preglownono, i have no time for coding right now
15:01:00preglowi was just wondering
15:03:39muesli|tarnpreglow: disabling autostart for removable devices in tweakui didnt help btw
15:04:24preglowmuesli|tarn: did for me
15:04:43preglowmuesli|tarn: that is, i just removed autostart for j:\, which is the drive my h120 always shows up as
15:04:50sneakumsthere's only one way to settle this - a duel!
15:05:12*preglow fetches his dueling pistol
15:05:16HClautostart settings is just when you have your iriver plugged in
15:05:18*rasher fetches 2 banjos
15:05:24HCland you select it
15:05:27HCland click properties
15:05:34HClautoplay -> always take this action: no action
15:05:39preglowdoesn't work here
15:05:43preglowthat option isn't here
15:05:46sneakumsi don't think there's a way to stop windows from automatically mounting the disk
15:05:50muesli|tarnyepp, here too
15:05:51preglowit's here for cdroms and such, but not for the h120
15:05:54sneakumseven if you disable autoplay
15:06:07preglownono, mounting = okay, the pesky screen that asks me what i want to do is not
15:06:17preglowi want it to mount, but not nag
15:06:19HClyea, thats autoplay
15:06:22HClit works here for h140
15:06:38preglowi'm okay anyway, it doesn't appear here anymore
15:06:52muesli|tarni have disabled my typical drives
15:06:55sneakumsis this specific to some version of windows or something?
15:06:57muesli|tarnlet see what happens
15:07:05sneakumswin2k never bothered me when i plugged in my h120
15:07:08muesli|tarni am running xp
15:07:30rasherxp is a bitch with the autoplay
15:07:39muesli|tarnyeah, 2k is quit uninteresting in an new plugged ihp
15:08:06preglowxp is a bitch, period
15:08:08Lynx_muesli|tarn: home or pro?
15:08:34linuxstbDoes anyone here understand firmware/ ?
15:09:13muesli|tarndisabling the drives worxxx
15:10:09Lynx_or gpedit.msc -> computer config -> system -> turn off autoplay -> enabled
15:11:26linuxstbAah. Just realised I was changing the stack size for the Gmini build.....
15:12:24muesli|tarnanother anoying thing is that the iriver starts everytime with the same song after i pluged it to the computer
15:12:25amiconnGood that you found it.
15:12:28linuxstbWho agreed that line 61 was correct :-)
15:12:47linuxstbI'm now changing line 185 of
15:13:44sneakumsmuesli|tarn: the iriver firmware isn't too smart
15:14:26sneakumsi have a feeling they might be storing the current track by number, and of course the numbering is invalidated if files are added or removed
15:16:21rasherso they just start by number 0001 :)
15:16:26 Join Quelsaruk [0] (~kvirc@
15:16:27*rasher punches iRiver
15:18:20preglowrasher: they do store current file by number, yes
15:19:32Quelsarukamiconn: did you check Omnipage pro SAPI voices?
15:20:09amiconnUnfortunately they don't appear in the list, as you found as well
15:22:51Quelsarukand i suspect there's no way to make them appear in the list
15:23:03amiconnYay! Disabling X keyboard autorepeat works! Hacky, but that is fixable...
15:23:20Quelsaruk? ein?
15:27:04linuxstbamiconn: I think that's done the job :-) Increasing the stack to 0x20000 fixed my libmad plugin, but I had to increase it to 0x50000 to get libFLAC working.
15:27:52amiconnlinuxstb: That means some more work on them, changing some variables to static...
15:28:27linuxstbamiconn: Yep, but at least they seem to be working reliably now. That's a big step.
15:29:23linuxstbIs there a way I can display the current stack usage?
15:29:53linuxstbSorry, I've realised that doesn't make much sense.
15:29:55amiconnHmm. I dunno whether the stack fill & check does work correctly.
15:30:22amiconniRiver crashing with the check in thread.c enabled suggests that it doesn't
15:30:48amiconnIf it works, you can check the peak stack usage of each thread in the debug menu.
15:42:05linuxstbAt least I can now happilly say that both libmad and libFLAC are correctly decoding the data on the iRiver. So technically, the iRiver now supports FLAC - convert to WAV using Rockbox (in about 8% of real-time), then play it using the iRiver firmware :-)
15:45:39 Part amiconn
15:47:20Lynx_linuxstb: does libmad work in realtime?
15:47:39linuxstbLynx: No :-( Even slower than FLAC - about 4% of real-time on my 192kbps test files.
15:48:02linuxstbBut I'm not going to worry about that until Linus makes the iRiver run at full-speed.
15:48:26preglowmad should be even slower than flac
15:48:45preglowno surprise there
15:51:01Lynx_why is the iriver firmware so much faster? it's all software decoding, too, isn't it?
15:51:41preglowhardware isn't properly utilized yet
15:52:19linuxstbAlso, I'm sure the iRiver MPEG decoder will be a proprietory version, highly optimised for the coldfire.
15:52:29Lynx_ah, so the processor runs faster with the iriver firware than it does currently with rockbox?
15:52:32linuxstbI think Motorola have written such a MP3 decoder.
15:52:52linuxstbLynx_: Yes, probably at least 10x faster.
15:53:00preglowthe mp3 decoder is very probably motorolas own
15:53:09preglowi doubt very much we'll be able to make ours that efficient
15:53:31linuxstbI forget the details, but a google for emac and mp3 will probably find the press release.
15:53:45preglowyes, it's on the 5249 page
15:54:09Lynx_well, do we know how efficient the iriver one is?
15:54:27preglowand we probably never will
15:54:43preglowthe closest we can do is compare battery usage, but that depends on a lot of other factors as well
15:54:58preglowbut why should we care? we'll optimize our mp3 decoder as much as we can anyway
15:58:15Lynx_an then we can always think that maybe the iriver decoder also just gets by with the available processor power ;)
16:00:37 Quit R3nTiL_ ()
16:04:27preglowhadn't it been for the fact that i know the motorola decoder just uses 20% cpu ;)
16:07:31Lynx_hmm, is it correct that if the we can't get libmad to use only like 50% (or less) there will be no crossfading?
16:08:02Lynx_unless of course it is somehow calculated in non realtime while the first file is still playing...
16:08:09linuxstbLynx_: I don't know how much CPU cross-fading will need, but surely it will be less than the actual MP3 decoding itself.
16:08:25Lynx_but don't you have to decode two files at the same time?
16:08:30sneakumsand as long as you're ahead of playback by a bit, you won't need to decode two files at once
16:10:00 Join Sucka [0] (
16:10:14Lynx_are the files read into ram and then decoded (into ram), or decoded from disk into ram?
16:10:40linuxstbLynx_: My tests are reading the files into RAM in advance, and then decoding them.
16:10:54preglowLynx_: you can just start decoding the other file as soon as you're done decoding the first
16:11:01preglowLynx_: you dont have to run the decoders in parallell
16:11:12sneakumsand buffering lots of undecoded data should be better for battery life
16:11:30linuxstbpreglow: As long as you have enough unplayed data in the PCM buffer fromt he previous track.
16:11:51linuxstbSo that buffer needs to be big enough for that.
16:11:54Lynx_ok, i get it
16:12:45Lynx_it's 32 mb ram in the iriver, right? and pcm is about 8 times more data than mp3?
16:12:51preglowlinuxstb: yes, but that's just a matter of buffering long enough ahead
16:13:00preglowLynx_: that varies
16:13:24Lynx_preglow: i know, but roughly 8 times...
16:13:30preglowmore like nine
16:13:46preglowbut no one uses 128kbs mp3 anymore
16:13:47preglowi hope
16:14:27Lynx_but in mpc to mp3 back to pcm the second pcm file should be smaller, right? or is pcm always the same size per time?
16:14:40sneakumspcm is fixed.
16:14:44preglowhell no
16:14:50preglowpcm is always the same size per second
16:15:17preglowa pcm file might contain perfect silence, and it would still be just as big as a pcm file containing music
16:15:30Lynx_hmm, that's why it's called uncompressed i guess ;)
16:15:44sneakumsa reasonable thesis.
16:16:02 Quit Cassandra_ (Read error: 110 (Connection timed out))
16:16:06 Join Cassandra_ [0] (~christi@
16:16:25Lynx_and that's why the lossless codecs are around...i learn more every day in #rockbox :)
16:16:55Lynx_so flac is just like zip specialized for audio data?
16:17:19sneakumsaltohugh flac tries different schemes for each block
16:17:27rasherIt includes frame information and such so you can seek
16:17:32rasherwhich zip doesn't
16:17:35linuxstbGoing back to my stack problems - libFLAC and libmad both now seem happy with a 32K stack. I can probably reduce it even more.
16:17:37sneakumssometimes it can't save space at all, so some blocks may be uncompressed
16:18:44rasherlinuxstb: might a different stack sice change cpu usage?
16:18:56linuxstbrasher: I don't think so.
16:19:10sneakumsyou either have enough stack, or you're scribbling all over the heap
16:20:36rasherI'll scribble all over your heap in a minute.
16:21:14*Lynx_ looks up heap in wikipedia
16:22:37*Lynx_ notices that it does not make sense to look up single words when you don't have a clue about the general concepts
16:23:37*rasher gives Lynx_ a cookie
16:24:24sneakumsi used the term in a more general sense than what the wikipedia article describes
16:24:42*Lynx_ greedily grabs the cookie, almost biting rashers hand
16:25:02sneakumsand by "more general" i mean "wrong"
16:25:54Lynx_sneakums: so it's a good thing i didn't understand the article? ;)
16:33:28 Join jyp [0] (
16:33:40 Join webguest93 [0] (
16:33:44 Join R3nTiL_ [0] (~zorroz@
16:34:34 Quit ashridah ("sleep.")
16:37:39 Quit webguest93 (Client Quit)
16:37:51 Quit R3nTiL_ (Client Quit)
16:46:55 Quit Diway ()
16:50:23Lynx_another question: if pcm is about 9 times as much as mp3 in data/time, and the iriver has 32 mb ram, we effectively get only the same buffering time as on the 4mb archos?
16:50:49linuxstbIt depends how big the PCM buffer is - it will probably only be a few seconds, at the most.
16:51:27Lynx_ah, right
16:51:45preglowthe buffering time will be lousy with wav and such, yes
16:51:55preglowthat's why you shouldn't use wavs on portable players ;)
16:52:23sneakumsLynx_: the ram will mostly be used to buffer mp3s and such, not decoded pcm
16:52:27Lynx_how is the flac compression ratio?
16:52:42linuxstbWAV is all I've been using with the iRiver firmware - it can't play any of my compressed files :-(
16:53:10linuxstbBut I have to admit, I rarely use it at the moment.
16:53:21preglowflac compression is usually around 1:2
16:53:28sneakumsthe iriver was the first halfway decent player that would play vorbises
16:53:28Lynx_linuxstb: and now you can, with if you wait a litte ;)
16:53:36sneakumsbefore that i just suffered in silence
16:53:43preglowlinuxstb: why couldn't you?
16:54:03preglowlinuxstb: do you exclusively collect protected wma's or somethinmg?
16:54:05rasherwe loves them vorbises.. preciousssesssss
16:54:22sneakumshe mentioned mp2 at some point
16:54:31***Saving seen data "./dancer.seen"
16:54:33Lynx_wasn't he the flac guy?
16:54:35preglowwell lucky him, mad supports all mpeg layers
16:54:42rasherflac and mp2, I think
16:54:49linuxstbMy files are either MP2 (recordings from European digital radio/TV), AC3 (US digital TV) or FLAC.
16:55:06preglowlinuxstb: are there tools to rip the mp2 straight from the broadcast?
16:55:20linuxstbpreglow: Yes -"> :-)
16:56:24linuxstbIt's trivial using the Linux DVB drivers from - you just call a few ioctls to tune the card, another one to set "PID filters", and then read the MPEG Transport Stream direct from a device file.
16:56:34Lynx_linuxstb: do you have to split the radio mp2 into songs by hand? If you want to have one file per song in the first place, that is.
16:56:39preglowwhich is how it should be
16:57:36linuxstbLynx_: Yes. What I normally do is create a .cue file for shows that I want to listen to more than once. Otherwise, I just listen to the whole broadcast (John Peel, RIP) from start to end.
16:58:04preglowrip indeed
16:58:26Lynx_linuxstb: depending on the show maybe silence detection would work somehow...
16:58:36preglowvery few shows have silence
16:58:42preglowdepends on the channel, probably
16:59:02Lynx_preglow: and classic music channel have too much silence, probably ;)
17:00:31Lynx_yawn, reading 6 gb textfiles into mysql takes longer than i thought...
17:02:15*linuxstb wonders why the .wav format has 16-bits to store the number of channels...
17:03:17 Join Tang [0] (
17:03:27TangHello :)
17:03:40preglowi wonder more why they store 8 bit samples as unsigned numbers and 16 bit samples as unsigned numbers
17:03:54muesli|tarnhi tang
17:04:03TangHi Museli
17:04:06 Nick muesli|tarn is now known as muesli_ (
17:04:06linuxstbevening all.
17:04:11TangSorry Muesli
17:04:25TangGood evening linuxstb
17:04:26muesli_for what?
17:04:34Tangi've watched the codec part
17:04:53Tangcool you 've seen the last corrections by HA members
17:05:04Tang(for libFLAC specially)
17:06:23linuxstbTang: Yes, libmad and libFLAC are now working (but outside the non-existent new Rockbox audio system)
17:06:54TangAh okay linux
17:07:06Tangi didn't know that they were ported yet
17:07:43preglowlinuxstb: as for tremor, btw, you don't think we'd need the lowmem branch?
17:07:52Tangi'm gonna relay the info on my usual boards
17:08:21preglowTang: just make it clear we're not making sound yet, please
17:08:26preglowonly wav writing
17:08:32linuxstbpreglow: I've only very briefly looked at Tremor, I haven't started coding yet. Where would I find the lowmem branch?
17:08:40Tangyes no matter i've well understood
17:08:43preglowlinuxstb: it's in cvs, but i don't think we'll need that
17:08:56Tangand i'm not guy who make false hope
17:08:58preglowlinuxstb: it trades off memory requirements for cpu
17:08:59Tangdon't worry
17:09:03linuxstbpreglow: Actually jsut raw PCM writing, I'm adding WAV headers as we speak.
17:09:31*rasher sets out putting together a wiki page with links interesting for blind users
17:09:33preglowlinuxstb: remember to swap the endianess :P
17:10:04linuxstbpreglow: Yep!
17:10:41preglowaiff writer would be neat as well
17:11:37linuxstbpreglow: I'm not going to worry about AIFF yet, this is just for my own testing. In the long term, we should probably add AIFF support to Rockbox though - add it to the Codecs page.
17:11:42 Join webguest58 [0] (
17:12:28preglowno, just meant you might as well see if that lib's usable, if you haven't gotten too far in the wav stuff
17:12:59sneakums i found a lowmem tremor branch in xiph's svn
17:13:01preglowit uses f* file class :PP
17:13:27sneakumsthe changelog doesn't say much though
17:13:34linuxstbsneakums: Thanks, I'll look at then when I start work on Tremor.
17:13:39preglowi haven't been able to find much info
17:13:55preglowthough i did read a nice post recently, i'll see if i can find it
17:14:03sneakumsthe trunk remor has the same changelog
17:14:15sneakumsguess you'd need to have a look at the commit history with xvn
17:14:33linuxstbpreglow: Writing WAV is trivial, it's reliably reading them that can get difficult.
17:15:14preglowlinuxstb: i know, i've made about three wav libs
17:15:18preglowone for each hard disk crash
17:15:22sneakumswow, those numbers look pretty decent.
17:15:34preglowsneakums: i agree, i don't think we'll be needing the lowmem branch
17:16:14sneakumsthey weren't kidding when they said lowmem
17:19:31TangHave you some contact with Xioh guys or any ogg coders?
17:20:39linuxstbTang: No, there's been no need to yet.
17:21:07Tangindeed it 'd very hard
17:21:11preglowi don't expect we'll need it either
17:21:47Tangok that's fine really
17:25:59TangSo i'm gonna correct the comparison chart abut iHP and mail this
17:28:18Tang(mmm i think about something
17:28:30Tangwhen working on codec
17:28:47Tangare you working the metadata part in the same time
17:29:17Tangor would it be left for later?
17:30:08linuxstbTang: I'm thinking about the metadata part at the same time... i.e. I'm not writing any code yet, but I'm sure it will be part of the first real implementation.
17:30:16preglowthe metadata functionality would almost certainly have to be separate from the codecs, but i don't know
17:30:25preglowthere might be clever ways of handling it
17:30:36Tangin fact i do know some METADAT are "transcodec"
17:30:43preglowTang: exactly
17:30:44Tangexemple: PA can be used for both
17:30:56TangMPC,PAEtag for
17:31:05TangmonkeyAudio but also MPC and MP3
17:31:15preglowyes, and that's why i think they need to be separate
17:31:32Tangokay i was wondering about
17:31:47preglowwe know of the issue, so it'll be in our design decision
17:31:48Tangbut indeed sperate coding may be suitable
17:32:42Tangokay so if needed later i would search for info about metadata part at HA place
17:35:23rasherI started an index for blind users, but I don't really know where this info is collected, so if anyone could add "CategoryBlindUsers: <description>" to pages with information for blind users, or add links to that'd be great
17:40:39 Join Hohoman [0] (
17:40:39 Join xen` [0] (
17:54:08 Join amiconn [0] (
17:55:04rasherEvening Hohoman, xen`, amiconn
17:55:31 Quit webguest58 ("CGI:IRC")
18:06:35Hohomangood evening
18:09:48 Join elinenbe_ [0] (
18:09:49 Quit elinenbe (Read error: 54 (Connection reset by peer))
18:09:52 Nick elinenbe_ is now known as elinenbe (
18:21:49linuxstbNot sure if this is considered a bug, but I created a .wav file from my libFLAC plugin, and the 8.3 MS-DOS filename Rockbox gave it was "FLACTEST.A4L", and not "FLACTEST.WAV", as requested.
18:22:28linuxstbThe iRiver then refused to recognise it as a WAV file - even though the long name is "flactest.wav"
18:22:30rasherthat sounds very strange
18:22:56 Join XShocK [0] (
18:23:07amiconnlinuxstb: That's expected rockbox behaviour
18:23:22amiconnThe 8.3 extension is generated randomly
18:23:57amiconnEasier handling of duplicate shortnames
18:24:07linuxstbIt looks like the iriver firmware ignores the long names when checking for filetypes, and just uses the short name. Hence .flac support can't happen :-)
18:24:57HClamiconn: how about only doing that when flactest.wav already exists?
18:25:29HClsounds more sensible to me...
18:27:01linuxstbIt could cause problems if people want to switch between Rockbox and the original firmware (heretics!)
18:33:52linuxstbLooking at the code in fat.c, it already checks if the proposed short name is used, and if so, changes the extension to something random and then retries. So I think the only necessary change would be to make "create_dos_name" more intelligent.
18:35:58linuxstbWe could also change the "retry" logic to randomise the last three characters of the main part of the name, and leave the extension as it is.
18:37:45 Quit toolmanwv (Read error: 104 (Connection reset by peer))
18:38:02preglowi think that's a better idea
18:38:04 Join toolmanwv [0] (
18:39:42rasherdoesn't windows do it by adding ~X to the main name?
18:39:46rasherwhere X is a number
18:40:20rasherWhy not use that strategy?
18:41:55 Join Nuxator [0] (
18:44:20Nuxatori just made a google search on coldfire mp3
18:44:24Nuxatorand found
18:44:39Nuxatorit's an coldfire mp3 decoder
18:45:18preglowwe can't use that
18:45:27preglowyou have to bloody jump through hoops to get it
18:45:34preglowand once you have it, there are licensing restrictionms
18:45:38NuxatorSOFTWARE - MP3 Decoder – Using the MCF5249 feature set, the advanced MP3 decoder is engineered to require only 37 kB of memory and 19 MHz. The decoder can run from on-chip memory and is designed to require fewer CPU cycles than typical decoders. In turn, this design helps to provide longer product battery life for consumers.
18:45:39sneakumsand libmad is ported and working anyway, thanks to linuxstb
18:45:50preglowNuxator: we know of this
18:45:56Nuxatorok sorry
18:46:01preglowno need to apologize
18:46:05preglowbut we can't use it
18:46:11Nuxatorbut wow only 19mhz to decode mp3
18:46:21preglowyup, it's a very fast decoder
18:46:32muesli_how many mhz does irivers cpu have?
18:46:39Nuxator140 max
18:46:52Nuxatorbut it rns at 10 by now
18:47:09amiconnNuxator: This is just a question of the implementation & dsp features in the cpu
18:47:17linuxstbHas anyone read the license? Unless I've missed something obvious, we have the right to use it and redistrbute it in source and binary form.
18:47:49preglowlinuxstb: then i propose someone be our "contact person" and get us this thing so i can study it :)
18:47:50muesli_how much are normally those fees?
18:48:02amiconnThe hw codec used in the archos (which is a dsp after all) runs at only ~18 MHz - and it is able to *encode* mp3 in realtime
18:48:34rasherIt doesn't matter if the fee is $0, the point is that there are licensing restrictions that don't go well with the GPL
18:48:46linuxstbrasher: For example?
18:49:09rasherlinuxstb: oh.. :\ I haven't actually *read* the license :)
18:49:15amiconnBagder: r u around?
18:49:49linuxstbI'm sure it isn't GPL compatible (that would be too good to be true), but I can't immediately see why not.
18:49:56preglowrasher: it actually allows you to change and redistribute the source
18:50:09sneakumsthe revocation clause is an additional restriction, i think
18:50:15preglowbut it's not gpl compatible, no
18:50:32preglowbut still, i'd love to study it
18:50:53Nuxatorthey are talking about a emac unit in coldfire wigh help decode algrithms
18:50:55rasherthen you're in the horrible gray-zone area of derived works
18:51:22sneakumsthe emac is being investigated, i believe
18:51:25preglowrasher: if they allow redistribtion of their changed source anyway, i don't really think they care
18:51:28linuxstbI'm sure for legal reasons no-one involved in Rockbox should even download it unless we're sure about the license..
18:51:45preglowit's in their best interest that this source gets studied
18:52:02rasherMaybe someone could ask them about it somehow
18:52:14rasherbut extreme caution should be taken
18:52:26sneakumsno sudden moves
18:52:29preglowworst case, the person that touches it should just not develop the mp3 part
18:52:34linuxstbrasher: Yes - that seems sensible. Ask if the code could be used in a GPL'd project.
18:52:48preglowlinuxstb: i don't think motorola cares, the gpl people cares
18:53:28rasherMaybe someone could even ask about them releasing it as gpl
18:53:30Tangcan't help this time sorry
18:53:33sneakumsthe indemnity clause also looks problematic
18:53:43preglowmotorola will not gpl it
18:53:48preglowthat's pretty certain
18:54:00preglowbut hey, let's screw this for now
18:54:01preglowwe've got mad
18:54:10preglowi'd like layer 1 and 2 support anyway
18:54:12muesli_dudes, what is a gpl ? :-)
18:54:25preglowmuesli_: a software license
18:54:26linuxstbpreglow: Motorola could license it to Rockbox under the GPL, but still keep their current license for other users.
18:54:26rasherlet's see how much damage it does to battery life
18:54:27sneakumsmuesli_: the gnu general public licence
18:54:34Suckahow much processor power does MAD use?
18:54:34amiconnpreglow: I don't care about layer 1, but layer 2 is also important for me
18:54:35***Saving seen data "./dancer.seen"
18:54:45sneakumswell, other users could then just get it from rockbox
18:54:50sneakumsso they might not go for that
18:54:50rasherwhat sneakums said
18:55:07linuxstbYes, we have to use MAD for layer 2, but if a better Layer 3 decoder can be found, we should use it.
18:55:14sneakumsonce you gpl something it can be redistributed indefinitely
18:55:16preglowwell, if this is interesting, someone should mail them
18:55:18preglowsimple as that
18:55:30linuxstbsneakums: But what are Motorola losing? They already give it away for free.
18:55:38preglowpersonally, the only thing i want from it is to see how they utilse emac instructions
18:55:45sneakumsi'm not saying they win or lose anything
18:56:00sneakumsi'm just saying it doesn't make sense to gpl something for only a subset
18:56:07Nuxatorooops i didn't expect to make people arging
18:56:18sneakumsthis is a discussion
18:56:23 Quit toolmanwv ("—I-n-v-i-s-i-o-n— 2.0 Build 3515")
18:56:30zesneakums: there's plenty of dual-license stuff
18:56:37rasherWhat, I came here for an argument!
18:56:41zelike reiserfs
18:56:43preglowso, anyone willing to mail them? :P
18:56:53linuxstbrasher: No you didn't.
18:56:54sneakumstrue, but in that case you decide yourself which licence to use
18:56:57TangYou want someone exterior to rbx?
18:56:57rasherthis should probably be a non-developer as well
18:57:10rasherjust to be on the safe side
18:57:16amiconnIf motorola releases it as gpl, it can only be used in open source projects (or they dual-license it)
18:57:17rasherlinuxstb: Yes I did.
18:57:23linuxstbrasher: No you didn't.
18:57:34preglowok, now it's an argument!
18:57:39Tangi do'nt mind doing
18:57:47rasherpreglow: No it isn't!
18:57:56*linuxstb must stop that, it's silly.
18:58:01Tangbut i think it'd be betetr someone who speak english better
18:58:12Tangtoa voide misunderstanding
18:58:14*rasher nominates sneakums
18:58:18preglowgo sneakums!
18:58:20rasherJust Because.
18:58:30linuxstbLet's wait and see what others (Bagder?) think iof it.
18:58:56Tang(question: will the ragument figure in the IRC log? :lol: )
18:59:17rasherIt certainly will
18:59:32amiconnrasher: Iirc you run linux?
18:59:41rasheramiconn: absolutely
18:59:45 Part GnagelRam
19:00:14amiconnCould you perhaps test my X11 hacks for me, as I can only test on cygwin?
19:00:31rashersend a patch or directions my way
19:00:53amiconnThis is not yet the full rewrite of the X11 sim kb handling, it's a proof of concept
19:01:07amiconnOne sec, generating patch...
19:03:28amiconnThis applies to 3 files in /uisimulator
19:03:53amiconnAfter applying, build an X11 sim (preferably Ondio, but it shouldn't matter)
19:04:15amiconnThen start this sim, press some keys and watch the console output
19:04:25rasherokay, checking
19:04:47amiconnTwo things should happen if things work as intended:
19:05:02amiconn(1) Keyboard repeat should be off
19:05:20rasheruh oh
19:05:21amiconn(2) The tick value printed should increase
19:05:23rasherdidn't build
19:05:33rasher../x11/screenhack.c: In function `main':
19:05:33rasher../x11/screenhack.c:588: warning: implicit declaration of function `signal'
19:05:33rasher../x11/screenhack.c:588: error: `SIGALRM' undeclared (first use in this function)
19:05:48amiconnUrgs... missing #include
19:05:56amiconnStrangely cygwin doesn't care
19:06:47rashercygwin lacks discipline
19:07:17amiconnPlease add #include <sys/time.h> somewhere at the top of /uisimulator/x11/screenhack.c and retry
19:07:23 Join hubble [0] (
19:07:37NuxatorI found a pll wizard may help linusN with setting coldfire speed
19:07:52NuxatorNo registration needed for this one
19:08:02rasherhaha, it's based on xscreensaver hacks?
19:08:03Bagder#include <signal.h> for signal()
19:08:11Bagderrasher: yeps
19:08:23*amiconn notices bagder around
19:08:29BagderI'm not really here
19:08:33hubbleI've got the iRiver audio chip to initialize, mute and demute and it works =)
19:09:06Bagdercool progress hubble
19:09:32amiconnrasher: I think both includes are necessary
19:09:35Bagderjyp: you working on fixing the builds?
19:09:36hubblenext is trying to write a sin-wave and see if I can get any sound out of it =)
19:09:48rasheramiconn: I added them both, it compiles
19:10:25preglowhubble: cool
19:10:32preglowhubble: that's all i2c, yes?
19:10:46hubblehehe.. it took some time reading the documentation, reversing the original firmware and getting it to work..
19:10:53hubblepreglow: yes
19:11:03preglowhubble: easier to write a sawtooth, heh, no sin() on 5249
19:11:21hubblepreglow: precalc table
19:11:24amiconnBagder, jyp: There is an 'e' at the very start of thread.c
19:11:30preglowsaw requires no table :P
19:11:33 Quit thegeek (Read error: 104 (Connection reset by peer))
19:11:36rasherpreglow: I get a keypress at the beginning, nothing in between
19:11:45rasheramiconn: I get a keypress at the beginning, nothing in between
19:11:57rasheramiconn: the tick count is larger at the end
19:11:58amiconnDo you get the release as well?
19:12:12amiconnNice, so this is working....
19:12:29rasherit didn't bring up the menu, but I'm guessing it's not tied into the sim yet?
19:12:30amiconnI can go ahead reimplementing the keyboard handling based on that
19:12:48preglowhubble: so you're porting i2c.c as well?
19:12:52amiconnrasher: No, as I said, I merely wanted to know if it works on real X
19:13:19rasheramiconn: alright, that's fine then :)
19:13:22amiconnThe timing of the X11 sim wil become much more similar to the target...
19:13:52hubblepreglow: not really, iriver has i2c interface in hw.. you basicly just write byte's to the CPU and it will take care of the bit-fiddeling etc
19:14:01amiconnBagder: What I wanted to ask - I need a place where to put the timer routine, but I'm undecided in which file it should go
19:15:16amiconnThe Win32 sim timing is handled in uisw32.c, so the equivalent would be uibasic.c
19:15:29preglowhubble: but yes, so you think you have i2s going as well?
19:15:44amiconnHowever, the Win32 sim doesn't use a separate routine, but the timer is part of the overall event loop
19:15:55hubblepreglow: i hope so.. going to try that tonight
19:16:19preglowi can write you sin generation code in a jiffy
19:16:39Bagderamiconn: I don't have any strong opinion
19:17:11linuxstbIs down at the moment, or is it just me? I can ping it, but get nothing via http
19:17:32rasherWorks For Me.
19:17:39linuxstbYes, and it's now working for me too.
19:17:51*Bagder is away again
19:18:23amiconnrasher: Not ® ?
19:18:50rasherThat too, I guess.
19:27:06 Join Stryke` [0] (~Chairman8@resnet-241-86.resnet.UMBC.EDU)
19:28:03 Join webguest17 [0] (
19:28:35 Quit webguest17 (Client Quit)
19:28:44 Join nrjazz12 [0] (
19:30:18 Quit nrjazz12 (Client Quit)
19:30:58 Quit Nuxator ("Chatzilla 0.9.67 [Firefox 1.0/20041108]")
19:34:50Lynx_see you guys, i'm going hooome
19:35:11 Nick Lynx_ is now known as Lynx_awy (HydraIRC@
19:40:46linuxstbProbably only Linus can answer this, but could the start_iriver_fw() function from bootloader/main.c be copied into Rockbox and then used to start the original firmware without needing to reboot?
19:43:04preglowprobably, depends on how much the iriver firmware depends on the hardware state
19:44:06amiconnSomething like RoLo on the archos units...
19:46:02linuxstbWould anyone else find that useful?
19:46:39sneakumspersonally, i hope to never have to use the iriver firmware again
19:47:00 Part amiconn
19:47:38rashermaybe if we don't get mp3 recording/encoding going quickyl
19:48:10Tangor the tuner
19:48:24linuxstbor WMA playback
19:48:30sneakumsnone of which i use ^_^
19:48:36rashersneakums: how deviant
19:48:48TangSomeone can tell me the CPU complete name and his real speed for an HA memeber?
19:49:13 Join Patr3ck [0] (
19:49:34sneakums CPU: Motorola SCF5249 140MHz coldfire
19:49:39sneakumsor so says
19:49:42Tangthanks :)
19:49:54Tangi'll give the info and the link
19:50:09preglowTang: they're welcome in the channel to ask themselves, seems kind of laborious to have to go through a board all the time
19:50:27Suckawhats HA?
19:50:39Tangyes indeed i'll send a link to the channel do you know the syntax for an IRC link in BBcode?
19:50:58preglowjust say #rockbox on
19:51:01preglowthat should do it
19:51:07Suckajust as a url
19:51:17linuxstbOr link to the IRC page on the Rockbox site - there's an archive there as well.
19:52:21Tangokay will use the IRC page access
19:55:49jypOops, I needed to go afk before the build finished
20:02:31 Nick StrathAFK is now known as Strath (
20:04:08TangNB: of course you can register and login at HA board too :)
20:11:05 Quit Patr3ck ()
20:12:24 Join JeanLuc [0] (
20:17:41 Part JeanLuc
20:20:16linuxstbIf anyone is interested in testing my libmad and libFLAC plugins, a file is here:">
20:21:17linuxstbI'm preparing a source zip file now - I've made a few changes in various places in Rockbox, so I'll just upload the whole of the apps, firmware and uisimulator directories (they work in the X11 sim, but I haven't modified the Makefile(s) for win32).
20:24:46xen`linuxstb: its approx the final codec api format ?
20:25:11hubblejust have to love the MCF5249 manual.. there are typos everywhere =)
20:25:25xen`and I would just have to do the same way in case I want to do another codec ?
20:25:27linuxstbxen`: No, it's just my test programs. There isn't really any API format at the moment.
20:27:48linuxstbI've implemented them as "viewer" plugins that convert the selected file to WAV - "/flactest.wav" or "/madtest.wav" respectively. You can then reboot and play that WAV in the iRiver firmware.
20:29:37hubblehow is the mp3 player performing? realtime?
20:30:09 Join ep0ch [0] (HydraIRC@
20:30:12linuxstblibmad is currently running at about 3.8% of real-time. libFLAC is about 8%
20:30:36hubbledoh! :)
20:30:42linuxstbBut we don't know how big the problem will be until Linus fixes the CPU speed.
20:31:20linuxstbI may also have made mistakes with my implementations - so if people could seriously look at my code, it could help.
20:31:43linuxstbEspecially the build commands (compiler flags, -D options etc).
20:33:30izzyI'll try to find some time to look at them
20:34:48 Join mrelwood [0] (
20:34:57 Quit mrelwood (Client Quit)
20:35:29 Join mrelwood [0] (
20:37:01 Quit methangas (" HydraIRC -> <- IRC for those that like to be different")
20:39:05mrelwoodsince the rockbox for iriver is still on hold, have you guys found a suitable replacement for a md-recorder? amongst mp3-players that is.
20:39:23sneakumsit's not hold
20:39:26sneakumswork is ongoing
20:41:45coobmrelwood: wouldn't any mp3 player with recording abilities suit your needs?
20:46:09TangThe iHP1xx seems suitable it can record yet and will do better with rbx
20:46:21Tangbut hard to find an iHP1xx now
20:46:32preglowmrelwood: what makes you think it's on hold? things have never gone quicker than right now
20:47:41hubblehow good are the archoses recording quality? v2 recorder? ondio?
20:49:28linuxstbhubble: I don't know about the quality of the ADC, but the v2 recorder (and others with the MAS chip) can only record at a maximum quality of about 175kbps VBR MP3
20:51:35hubblelinuxstb: ok.. I was thinking on getting one just to record some voice.. Is the build-in mic useful? I know my H120 gets interference from the HDD if I use the internal mic..
20:52:19Tangit's not interfernce
20:52:36Tangit's just that the int mic record the HDD activity sound
20:52:42Tangcause tey are too near
20:52:56Tangno electrical or magnetic phenomenon
20:53:14hubbleTang: ok :), does the archoes have the same problem?
20:53:16linuxstbMore info on the Archos:
20:53:48linuxstb(yes, the link does end with "cordi")
20:54:04Quelsarukhubble: the archos has the same problem with the internal mic
20:54:32hubbleQuelsaruk: ok
20:54:37***Saving seen data "./dancer.seen"
20:54:44linuxstbThere is a 140kbps VBR mono mode, which should be more than good enough for voice.
20:55:42mrelwoodcoob, i need atleast md-quality, so 320kbps the least i would accept
20:56:42coobwell will plain uncompressed recording do you?
20:56:46 Nick Quelsaruk is now known as Quel|out (~kvirc@
20:58:48linuxstbI think the Nomad Jukebox 3 is gaining respect from DAT tapers for uncompressed WAV recording.
21:02:23Tangat linuxstb:
21:02:49Tangnomad are very famous in this domain
21:03:38preglowi hate this fucking emac shit
21:05:49rasherthe nomad jukebox looks like a fucking joke though
21:07:03linuxstbAnd I don't think it's acts as a USB mass-storage device - so you need Nomad's software to access it. I think that's what stopped me from buying one.
21:07:36linuxstbThat, and the news that Rockbox was in the early stages of a port.
21:08:52coobheh and the archos recorder doesn't :P ugggly :)
21:09:23 Join toufn [0] (
21:12:44 Join Aison [0] (
21:14:31preglowdoes anyone know of any other coldfire disassemblers?
21:15:04 Join methangas [0] (
21:15:39hubblemrelwood: md quality? isn't atrax3(?) compression pretty pad?
21:17:22hubblepreglow: i'm using IDA (old version).. it isn't lovely but it works.. have to go through and manually tell it like.. this is code, disassemble please..
21:17:48preglowbut swell, i have ida
21:17:51preglowdidn't think of that
21:17:59hubblepreglow: but it's great because it's support for cross-references, commenting, etc etc :)
21:18:31hubblepreglow: i can send you my database of the original firmware so you get a jump-start
21:21:01Tangi didn't say i like the Nomad
21:21:11Tangbut it's recognied for recodrding usage
21:21:44preglowi USED to have ida pro
21:21:54*preglow pats his crashed hard disk
21:24:02hubblepreglow: oh :(
21:26:37hubblesaw-tand working!
21:26:51hubblemagic :)
21:27:28preglowyou get sound????
21:27:57preglowgoddamn, man
21:28:01preglowcommit this as fast as you can!
21:28:15 Join rjamorim [0] (
21:28:26rjamorimOMG hello Tang :)
21:28:36hubblei dont have write access to cvs but i'm definitely going to continue with this.. very fun
21:28:40TangHello Roberto
21:28:59Tangi presnet you to the Rbx team,
21:29:15Tangit's Roberto Amorim from HA
21:29:25Tang(admi of
21:29:39preglowhow timely
21:29:39rjamorimThe CPU you mentioned is more than enough to decode and encode WavPack in real time, even at high mode, dude :)
21:29:45preglowwe just made our first sound with rockbox
21:29:51Tangoups sorry
21:29:55rjamorimoh, my
21:29:58Tang(of course mùore than one rareware ;))
21:30:09rjamorimheh, true
21:30:59Tangin fact LinusM is not here now
21:31:10Tanghe's the HW specialist here
21:31:19rjamorimOh, too bad
21:31:23Tang(he has released the bootloader)
21:31:30rjamorimI was going to ask if Coldfire is ARM based or 68K based
21:31:46Tangnot ARM
21:31:48preglowwith dsp extensions
21:32:01Tangi let you with the specialists
21:32:09TangPreglow hubble and so on
21:32:10rjamorimI think David has experience with ARM assembly (in case you guys need optimizations), but I don't know if he knows 68K assembly
21:32:14rjamorimAh, OK
21:32:41preglow68k assembly is very nice, he'll surely be willing to learn :P
21:33:01rjamorimI wonder why the heck the CPU is so underclocked, and then comes iRiver and whines that some Vorbis bitrates can't be decoded in real time :B
21:33:08rjamorimHaha, ok
21:33:34rjamorimHe's working for a DVD controller manufacturer right now. He's very much into embedded stuff
21:33:41rjamorimUsed to work for a printer maker before...
21:33:55rjamorimProgrammed his printer to play a Bach fugue :S
21:34:17preglowrjamorim: it's software controllable
21:34:29preglowrjamorim: you can clock it however you want by adjusting pll registers
21:34:36rjamorimHrm.. I see, but why? Battery saving?
21:34:40preglowrjamorim: it's just that linus hasn't had time to find a good working configuration yet
21:34:53preglowit will run at full throttle later
21:35:03rjamorimWhat compiler are you using? Metrowerks?
21:35:08linuxstbpreglow: or hopefully less than full throttle.
21:35:17rjamorimhrm. Wouldn
21:35:25preglowmetrowerks costs money, afaik
21:35:30rjamorim't matrowerks create more optimized code?
21:35:35rjamorimAh, good point :B
21:35:39preglowwell, probably, but it's not free
21:35:47preglowwe can't exactly tell people to pirate software if they want to develop
21:35:50preglowand gcc is pretty goox
21:35:53linuxstbEverything about Rockbox is open source - including the build tools.
21:36:03preglowthe code it generates is pretty tight
21:36:22preglow68k is a very old target for it, so has had time to mature
21:36:38rjamorimAh, indeed.
21:36:40rjamorimThat's great
21:36:57sneakumsi think one of the first machines the gnu project had was a 68000 box
21:44:17preglowi wish to god i could make this emac unit work
21:44:36rasherhrm, the GPL, LGPL and BSD isn't a problem
21:44:43preglowusing that lib is pretty much out of the question any way you look at it
21:44:56preglowmotorola would have to release it as gpl
21:44:59preglowand that they won't do
21:45:08rasheror any other gpl compatible license
21:45:21preglowyes, bet they won't do that either, heh
21:45:31rjamorimYes. Pretty much unlikely :B
21:45:36rjamorimBetter focus on MAD
21:46:19rasherDear Motorola,
21:46:27rasher We want your shit. Pick one:
21:46:49rjamorimRight away, Mr. rasher
21:46:59rjamorimThank-you for doing business with us
21:47:03rasherI wonder if it'd be worth the bother
21:47:22rjamorimProbably not
21:47:27rjamorimhehe, they even offer an MP3 encoder
21:47:32rjamorimNow TAHT would be sweet
21:47:35linuxstbrjamorim: Yes, we are focusing on MAD - if by some miracle we can use the Motorola decoder, it would only be for Layer III decoding anyway, and we want Layer II as well.
21:47:42rjamorimEven if quality is bad (as it likely is)
21:47:52rjamorimAh, good point
21:48:26Tangokay Roberto i've read it
21:50:03TangBad quality for mad?
21:50:37rjamorimThe MP3 encoder from motorola
21:50:52rjamorimIt's full of constraints, so it can't be properly tuned.
21:51:08sneakumswhat sort of constraints?
21:51:18rjamorimReal time operation at a little more than 100mHz, low battery consumption, no FP values...
21:51:28 Quit toufn ("Leaving")
21:51:36rjamorimMostly RT operation at few CPU power
21:52:03TangAh okay
21:53:11preglowyou can do quite a lot in 140 mhz
21:53:25preglowno floating point really doesn't restrict you that much, you just have to think more while coding
21:53:30 Quit methangas (" HydraIRC -> <- IRC with a difference")
21:55:10rjamorimHaha, yeah, but MP3 encoding at high quality would require more horse power than 140mHz, even with heavy ASM optimizations.
21:55:40preglowwell, yes, if you want lame
21:55:47preglowbut you can probably do pretty good stuff still
21:56:21rjamorimyeah, guess so
21:59:59izzylinuxstb: have you tried how much faster your flac decoder works if you don't write the output to a file?
22:00:11hubblepreglow: why is it a problem to use the binary mp3 lib from motorola? it's almost as the linux kernel (nvidia) thing? as long as rockbox authors (linus?) are fine with it?
22:00:38linuxstbizzy: Yes, that's how it was working yesterday. There's virtually no difference to the % decoding speed.
22:01:26linuxstbhubble: I mentioned the same thing this morning. Best to leave that can of worms closed for the moment.
22:01:37hubblelinuxstb: ok =)
22:02:47preglowhubble: it's not binary
22:03:18preglowthe simplest thing would just have someone ask if it's okay
22:03:21preglowbut it's not urgent
22:04:12hubblepreglow: aha.. it is only available commersially?
22:05:48preglowhubble: you have to contact them, but yes, it seems they expect you to be commercial, since they ask for phone number for your salesperson and stuff
22:06:31 Quit mrelwood ("CGI:IRC (EOF)")
22:06:37hubblepreglow: ok..
22:06:41preglowi don't see why they restrict it like this in the first place
22:06:47preglowbut corporations can be pretty stupid
22:07:20rasherThey have a lot of people working hard on that
22:07:32rjamorimDamned dial-up. BBL
22:07:35 Quit rjamorim ("On the Turning Away...")
22:07:53preglowif they release a lib that does fast mp3s on a coldfire, how the hell can they not benefit from releasing it freely?
22:08:31 Quit Cassandra_ (Read error: 60 (Operation timed out))
22:08:34hubblepreglow: they cant decide if their main business is chips or software :)
22:08:51rasherI imagine it's down to concerned lawyers
22:09:12 Join mrelwood [0] (
22:09:42preglowyes, of coruse
22:09:52preglowlawyers need to be shot more often
22:09:56mrelwoodhey, the iRiver port really has proceeded a lot! I guess there is no knowledge if the glitch will be apparent or not?
22:10:07preglowmrelwood: not yet
22:10:16hubbleif it's hand optimized asm then it won't be any good for any other (compeditors) platform
22:10:22preglowmrelwood: we've barely gotten audio out to work, no recording yet
22:10:31XShocKhi all
22:10:49mrelwoodpreglow, well that is really good news! I suppose You are aware of the glitch?
22:10:59XShocKhmm, as i understand some people had gotten sound working somehow?
22:11:25 Join Diway [0] (~diway@
22:11:58 Quit muesli_ (Read error: 54 (Connection reset by peer))
22:13:05 Join muesli_- [0] (
22:13:43preglowmrelwood: yes
22:13:56preglowXShocK: hubble just made sound out work
22:14:19 Quit muesli_- (Read error: 104 (Connection reset by peer))
22:14:55 Quit midk ("Leaving")
22:15:09 Join Cassandra_ [0] (~christi@
22:15:21 Join muesli- [0] (
22:15:28 Quit muesli- (Read error: 104 (Connection reset by peer))
22:15:38mrelwoodman, I can't wait! I can't stay in my pants! If Rockbox fixes the glitch, it will become the Perfect DAP for me!
22:16:05preglowi think gas is frigging bugging on me here
22:16:14preglowi explicitely tell 'no shifts in multiply', it encodes 'shift left'
22:16:22mrelwoodI took it back to the store because of the glitch, and I'm still without a player/recorder.
22:16:37preglowmrelwood: the glitch is really very easily fixable
22:16:43 Join muesli_- [0] (
22:16:52preglowmrelwood: just open the file with an audio editor and fix the glitch by hand
22:17:51rasherpreglow: so even if it's a hardware problem, it's just a matter of making a "viewer" that fixes that error
22:18:14mrelwoodwell, i'm not ready to do that to all recorded files. besides, the 45 samples are lost anyway, the recorded track is shorter than the original.
22:19:09muesli_-sorry mates..mirc kixxx my butt
22:19:53muesli_-was gayt!?
22:20:03 Quit muesli_- (Read error: 54 (Connection reset by peer))
22:20:22preglowrasher: it can be pretty hard to detect
22:21:13rasherpreglow: ah, I don't know much about this glitch.. I don't really use my iRiver for recording
22:21:37preglowrasher: nor do i, but i checked it out once, it just drops a couple of samples from time to time
22:21:53preglowrasher: it probably occurs periodically, so could probably be easy to filter it out
22:21:55izzylinuxstb: what does the loop do in your flac plugin write-callback function?
22:22:02rasheroh, I thought it was periodical.. ah :)
22:22:21mrelwoodduring my testings it was always 44-45 samples, random interval was 20s - 1min.
22:23:19mrelwoodand it had nothing to do with the time when the hd light was blinking, so what I understand it is not a buffer flushing issue after all.
22:23:45preglowit's pretty hard to say without knowing what the firmware is like internally
22:23:51preglowbut we'll find out soon enouygh anyway
22:23:57preglowi require food
22:24:49 Join muesli- [0] (
22:25:03 Part jyp ("poof!")
22:25:11linuxstbizzy: It converts from flac's "one array per channel" output to an interleaved stereo buffer.
22:25:13mrelwoodI just ate a delicious steak, and brownie with ice-cream. I wish there was something I could do about the fw...
22:28:08muesli-mrelwood r u coming from the states?
22:28:23 Join Patr3ck [0] (
22:28:35Patr3ck /msg NickServ IDENTIFY internet
22:29:21mrelwoodmuesli, no. Finland.
22:29:30muesli-ah, ok
22:29:54muesli-am just watching a documentary about unhealthy fast food ;)
22:30:21mrelwood"Supersize me"?
22:31:03 Join twister [0] (
22:31:14muesli-yeah, something like this
22:31:30muesli-60% of all americans are overweighted ;)
22:32:16 Quit twister (Client Quit)
22:32:48mrelwoodwell, finns are not that much worse... I heard that we are supposed to be the most american population in Europe!
22:33:01mrelwoodworse, I mean better! :D
22:34:35muesli-you can eat as much as you can, but dont forget to move ;-)
22:34:48 Join piratePenguin [0] (
22:35:08 Join GnagelRam [0] (
22:35:10mrelwoodI know. I'm still working on that myself, still recovering from a back surgery.
22:36:07muesli-:-/ hope you make progress
22:36:27TangFinished my correction of the comparison chart
22:36:31Tangi send you
22:36:33Tangby mail
22:36:44Tangit's in a RTF attached file
22:37:07muesli-me too pls ;)
22:37:29mrelwoodmuesli, slowly but unsurely... :P
22:39:19 Quit piratePenguin (Client Quit)
22:41:02muesli-hi tang ;)
22:41:45Tanghhi Muesli
22:43:01hubblehm.. the sinus wave does not sound like a sinus should :)
22:43:15muesli-like a cosinus? ;=)
22:44:33Tangcannot sent the file
22:44:41Tangwill retry tomorrow
22:45:23 Quit Tang ("Chatzilla 0.9.66 [Mozilla rv:1.7.5/20041108]")
22:51:16preglowhubble: no?
22:51:30preglowhubble: i've got a sine wave routine i know works
22:53:02preglowor is the problem the codec? heh
22:53:28hubbleyea, it's probably i2s format or some codec setting
22:53:32 Join amiconn [0] (
22:54:02 Quit ep0ch (" HydraIRC -> <- Go on, try it!")
22:54:37 Join webguest94 [0] (
22:54:40***Saving seen data "./dancer.seen"
22:55:40webguest94question..does anyone use Trillian as an IRC client ? I'm trying to access this room via trillian
22:56:26*rasher bites his lip
22:57:46 Quit webguest94 (Client Quit)
22:59:03 Join webguest07 [0] (
23:00:18 Quit webguest07 (Client Quit)
23:05:26 Join midk [0] (
23:06:24mrelwoodIs any of You aware of a mp3 recorder that would record atleast 320kbps from line in with input metering?
23:06:41mrelwoodand would be small enough to fit a pocket.
23:06:50midkthe irivers do that, don't they?
23:07:33hubblepreglow: okey.. now the sinwave is pretty good.. a few plopps because i'm not using interrupts to feed the audio data yet
23:07:34rasher320kbps mp3
23:07:51midkiriver does 320, right?
23:08:41preglowmrelwood: at least? 320kbps is maximum for mp3
23:08:54preglowhubble: goodie
23:13:23hubblewonder if it's the "analog mixer volume" or "master mixer volume" you're supposed to use interactively
23:13:55 Quit HCl ("Lost terminal")
23:14:34 Join hcl [0] (
23:17:47mrelwoodpreglow, I meant that wav is preferred.
23:18:10mrelwoodmidk, other than iRiver. Just in case the glitch can not be fixed after all.
23:18:18midkwhat glitch?
23:19:30mrelwoodit drops 45 samples every now and then (20s - 1min), which will sound as a "snap"
23:20:43mrelwooddo a search on "iriver recording glitch" with google, or forums or wherever, it is quite well known.
23:21:14 Join amx [0] (
23:23:54 Join jyp [0] (
23:29:16 Quit midk ("Leaving")
23:30:23 Join toolmanwv [0] (
23:35:45 Quit amx ("Client Exiting")
23:51:20amiconnDoes anybody have an idea what "Xlib: unexpected async reply (sequence 0x155)!" means?
23:51:21 Quit toolmanwv (Read error: 104 (Connection reset by peer))
23:51:42preglowhubble: do you have any disassembled code that uses bloody mac instructions handy?
23:51:55preglowi can by no means make mac work in fractional operand mode
23:56:52 Join amiconn_ [0] (
23:57:06 Quit amiconn (Nick collision from services.)
23:57:06 Nick amiconn_ is now known as amiconn (
23:57:07preglowand the docs can't bloody well agree on if 1 means signed or unsigned mode
23:57:15amiconnHmm. That's actually bad, because for the new button handling to work, I need to poll button events from X asynchronously.
23:57:19amiconnHowever, I just tried polling slower. It *seems* to work

Previous day | Next day