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).

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2007-10-18

00:00:03safetydanpreglow, oh yeah, I remember you talking about that at some point. Definitely a good idea to keep the frequencies out of it. They can always go in the manual if someone really wants to know
00:00:57BonBonTheJonmy ipod 4g won't boot rockbox, I get a checksum error
00:01:02 Quit davina ("xchat on Ubuntu 7.04")
00:01:30preglowsafetydan: yeah, and like i said, if we want to change sample rates on the hw side at some point, they're not going to be correct, and i don't know if i can be bothered to supply a full (calculated) list in the manual
00:01:44preglowsafetydan: if you want frequency exact controls, you should be using sweq
00:01:47krazykitBonBonTheJon, sounds like you need to download a new build and update.
00:02:15preglowsafetydan: i'm thinking perhaps something like "1 (lowest), 2, 3, 4 (highest)" for the video, perhaps
00:02:20BonBonTheJonkrazykit: I used the rbutil-qt to do the full install and small install, both times I get the error
00:02:23preglowand just "low, high" for the targets that only have two choices (most)
00:02:48 Quit donutman25 ("Konversation terminated!")
00:03:05safetydanpreglow, sounds good to me. I'll see what I can do
00:03:14preglowsafetydan: sweet
00:03:58krazykitBonBonTheJon, well, how about you try doing it manually? download the latest build and unzip it to the root of your ipod's drive
00:04:19BonBonTheJonkrazykit: I'm busy now, but I'll try it when I get back home
00:05:22preglowamiconn: some eu person is lending his nano away on the ml (hint, hint) ;)
00:07:25 Quit BonBonTheJon ()
00:08:51 Join AppleOSHelpplz [0] (i=463ea1c6@gateway/web/cgi-irc/
00:09:31 Join BigMac [0] (
00:10:07AppleOSHelpplzHI - I have a 5.5g iPod and want to have Apple OS and Rockbox on it and I am using Loader 2.5 and when ever i select Apple OS from the menu it just resets to the loader 2.5 menu
00:10:33krazykitAppleOSHelpplz, loader2 is not supported here. ask the ipodlinux channel for that.
00:10:55 Quit roolku ()
00:11:05AppleOSHelpplzok thank you, does the rockbox loader support Apple OS loading?
00:11:15krazykitit does. it also supports booting ipodlinux
00:13:14 Join sup_ [0] (i=1000@
00:13:41courtcAppleOSHelpplz: are you using a loader2 config file, or is it defaulting?
00:13:59 Quit sup (Nick collision from services.)
00:14:10courtcDo you guys get a lot of loader2 questions?
00:14:27krazykitnot a ton, but loader2 is not supported by rockbox
00:14:38 Nick sup_ is now known as sup (i=1000@
00:15:00bluebrotherthe manual tells how to load AppleOS using the Rockbox bootloader ...
00:15:04safetydananyone ever looked at using llvm instead of gcc?
00:15:09courtcIt's not a rockbox project, I wouldn't expect it to be.
00:15:17preglowsafetydan: why would we?
00:15:55safetydanpreglow, it's supposed to be substantially better at optimising
00:16:46amiconnAnd it's substantially more limited in target support...
00:16:51amiconnJust a quick glance
00:17:05preglowsafetydan: that would surprise me a great deal
00:17:47safetydanamiconn, true that
00:17:50 Quit lee-qid ("aufwiederbyebientotsayonara")
00:17:56preglowof its targets, only arm is covered
00:18:01preglowwe need sh and coldfire too :/
00:19:40markunpreglow: we could at least try llvm for arm with the gcc frontend
00:19:50 Quit Rob2222 ()
00:19:54preglowmarkun: sure, would be interesting
00:20:31safetydanI see it was proposed as a gsoc project even
00:20:55markunit looks like the llvm guy actually listen to complaint and look at patches, perhaps we could persuade amiconn to write a coldfire or sh1 backend ;)
00:21:25amiconnNo, thanks
00:22:01 Quit BigMac ("Leaving")
00:22:16markunamiconn: you probably said the same about mpegplayer optimisations a few months ago :)
00:22:39preglowyeah, i didn't see that coming :P
00:23:12courtcWell... I'll look into llvm... We only have on arch ;)
00:23:38amiconnmarkun: That's not true - I wanted this cf asm idct going in, but first didn't understand at all how it worked
00:24:23amiconnAnd nobody else did it in the meantime, so I started to look into it thoroughly
00:24:30AppleOSHelpplzok so i want to use the rockbox loader but when i browse to where my .ipod file is it doesn't show it whats up?
00:25:04amiconnWhile I only use a small subset of all rockbox features, I hate it when there are half-working ones
00:25:29 Join sarixe [0] (n=root@
00:25:32 Quit Zagor ("Client exiting")
00:25:44amiconnThat really annoys me. Some new feature gets added, and nobody seems to care to finish it :(
00:26:11markunwell, thanks for working on it anyway
00:26:11bluebrotherAppleOSHelpplz: what exactly do you want to achieve?
00:26:38 Quit elinenbe (Read error: 110 (Connection timed out))
00:26:41amiconnIt's still not optimal, but at least it's usable on X5 now
00:26:45AppleOSHelpplzI want to be able to load the defaut apple firmware from rockbox (or before rockbox loads) so that i can switch between the two
00:27:03amiconnAnd if we get this codec stack seize in, there is potential for an extra 10% speedup
00:27:10bluebrotherwell, read the manual. It explains how to do this.
00:27:19sarixe2 bugs to report: one, when i start up rockbox and resume playback from when i turned it off, the song doesn't start at the beginning, but rather a fraction of a second ahead.
00:27:27bluebrother... and we expect you to search the manual before asking for support.
00:27:28 Join Zagor [0] (
00:27:30AppleOSHelpplzRockbox is able to load and start another firmware file without rebooting. You just “play” a file with the extension .ipod
00:27:36AppleOSHelpplzright from the manual
00:27:44amiconnI wonder how much it would help to convert the other motion compensation functions to asm
00:28:02sarixetwo, recently, when i shut off rockbox, instead of shutting down promptly, it says "Shutting Down..." and starts to play the song i was listening to.
00:28:07amiconnIt would help to have statistics of a few videos
00:28:08bluebrotherAppleOSHelpplz: well, that tells about "another firmware". It also tells you how to start the original firmware.
00:28:17sarixethen after a few seconds, it stops and shuts down.
00:28:54 Quit billytwowilly (No route to host)
00:28:57amiconnThe simple block move is obviously used most, and half-pixel motion isn't used much (is it used at all in elephants dream?)
00:29:10bluebrotherthere is even a section about it.
00:29:11amiconnBut how about the destination averaging variants?
00:30:04 Quit ompaul (Client Quit)
00:32:36preglowamiconn: well, some statistics should be quickly gathered by just inserting some printfs and massaging the output
00:32:46preglowamiconn: just make sure to get videos produced by different encoders
00:33:53 Quit sarixe ("leaving")
00:34:56 Part toffe82
00:34:58 Join JdGordon [0] (
00:35:27 Quit bb (Nick collision from services.)
00:35:32 Join bb_ [0] (n=bb@unaffiliated/bb)
00:35:48 Quit barrywardell ()
00:36:33 Quit PaulJam (".")
00:38:27AppleOSHelpplzThank you for your patience and your method worked - i just had to re-install the ROLO rockbox bootloader
00:38:43 Join midgey [0] (
00:38:48 Quit sup ("Lost terminal")
00:38:56bluebrotherRoLo is not the bootloader ...
00:39:19AppleOSHelpplzyes im sorry i relised that as i hit enter
00:39:38AppleOSHelpplzbye and thanks again for your help
00:39:41 Quit AppleOSHelpplz ("CGI:IRC")
00:41:25iambenanyone ever have trouble w/ rockbox double-triple-quadrupling entries in your database? i think its happening because of auto-update db, but im not sure
00:41:39 Quit MethoS- ("Konversation terminated!")
00:41:45iambeni go to any album i have tagged, and track 1 is listed 4 times, track 2 4 times, etc
00:42:06 Quit moos ("Glory to Rockbox")
00:46:16 Quit scorche|w ("CGI:IRC")
00:49:06 Quit Soap (Read error: 110 (Connection timed out))
00:50:00preglowwhy, oh why did freescale remove the reg update functionality of movem.l
00:51:41 Quit petur ("Zzzzz")
00:52:30|Rain|iamben: what platform? I have had that happen, though
00:52:59iambenmy battery is dead now so i gotta charge a bit before i can investigate further =)
00:53:04|Rain|iamben: ditto. I have noticed that when it does happen, rockbox goes through the "commiting database" step again on boot, even though nothing has been changed
00:53:20iambenyeah ive been seeing that a LOT
00:53:27iambeni guess thats probably at the root of the problem
00:53:35iambencommiting DB more than one time
00:54:32 Quit jhulst ("Konversation terminated!")
00:54:52|Rain|I haven't had a chance to debug it yet, myself (always on the go and in a hurry when it happens), but if you reinitialize the database it'll temporarily fix the problem
01:00:24 Join donutman25 [0] (n=dagni@
01:01:49 Quit bluebrother ("leaving")
01:01:59 Quit rotator ()
01:02:18 Quit Nico_P (Remote closed the connection)
01:02:44iamben|Rain|: yeah thats what ive been doing, but w/ powerups & powerdowns it quickly gets back to broken
01:02:44 Quit kubiix (Read error: 104 (Connection reset by peer))
01:03:23iambenthis is the first time since i noticed the problem that ive had the time to work on it
01:04:09preglowohh, think i just took wma.c from 175.54 to 200% realtime
01:05:17 Quit ender` (" You know you live in Saskatchewan ... if you know all 4 seasons: almost winter, winter, still winter and road construction.")
01:05:28 Quit kkurbjun ("leaving")
01:07:36 Join Mouser_X [0] (n=Mouser_X@
01:09:40safetydanpreglow, you almost sound like you did that by accident :)
01:09:43*amiconn is still impressed by the asm averaging of 4 byte-pairs packed into 2 longwords using just 9 instructions (cf) resp. 6 instructions (arm)
01:10:09amiconnHow does one figure out that kind of algorithm?
01:10:44preglowi have no idea
01:10:51preglowsafetydan: didn't expect that much
01:11:13preglowi just coded a routine that multiplies two signals and adds it back to an array in coldfire asm
01:11:16preglowdoing four samples at a time with emac
01:12:20***Saving seen data "./dancer.seen"
01:14:04 Join Soap [0] (n=Soap@rockbox/staff/soap)
01:14:21iambenso it really should only have to commit the db the FIRST boot after initializing right?
01:14:40 Quit mirak (Remote closed the connection)
01:15:46preglowi wonder if doing only two and two samples at a time will be as fast
01:16:11preglowi just went four, since emac has four accumulators
01:16:47iambensometimes i wish there was a #rockbox-dev & #rockbox-users =)
01:18:45maxkelley#rockbox, and #rockbox-community
01:19:29maxkelleyvi-ola :)
01:19:43preglowiamben: i used to think that too, but changed my mind
01:19:46 Quit HellDragon (Client Quit)
01:21:19preglowamiconn: seems processing two and two samples is about as fast as processing four and four samples, is this surprising?
01:22:10iambeni like reading all the dev talk, i just feel out of place at times when asking a user question
01:22:56preglowiamben: understandable, but we like the benefits of using the same communications channels as users when possible
01:23:06amiconnpreglow: How would I know? I don't know what you're doing...
01:23:36preglowmentioned it ten minutes ago: i just coded a routine that multiplies two signals and adds it back to an array in coldfire asm
01:23:58preglowstandard emac fare
01:24:23amiconnYes, and that still tells me next to nothing about timing behaviour
01:24:49amiconnInstruction order is sometimes important, especially regarding emac
01:25:35preglowit's just movem.l, movem.l, macx2, movclr.lx2, add-with-memx2, rinse and repeat
01:25:37iambeni used to be a comp eng major, but i quit... so i understand SOME things
01:25:55preglowthere is some stall, no stall, since i need to increment addresses for movem.l
01:26:07preglows/some stall//
01:26:08preglowi suck
01:27:13 Join HellDragon [0] (n=Nocebo@unaffiliated/helldragon)
01:31:56 Quit Zagor ("Client exiting")
01:33:58amiconnpreglow: If you put movclr directly after mac, you'll have a 2-cycle stall
01:34:31preglowamiconn: i have two addq there, exactly enough to eat up the stall
01:34:32amiconnTry to put 2 instructions inbetween to avoid that
01:34:44preglowamiconn: i was the first person to know this, you don't need to tell me :)
01:35:24amiconnNext candidate is the movem. Line fetches are faster, unless the data is in iram
01:35:49preglowsince the two sample loop is as fast as the four sample one, i belive it is
01:36:22preglowahh, i can optimize further, the block length is always power of two
01:38:19lostlogichas anyone made any noise about porting to the iAudio 7? looks like a cool device
01:38:32preglownone i've heard of
01:40:16 Quit jhMikeS (Read error: 104 (Connection reset by peer))
01:40:55 Quit spiorf (Remote closed the connection)
01:41:45 Join jhMikeS [0] (n=jethead7@rockbox/developer/jhMikeS)
01:46:36 Join pill [0] (
01:46:50 Quit _pill ("changing servers")
01:53:46preglownano does 220% realtime, h120 207%
01:54:40 Quit Thundercloud (Remote closed the connection)
01:55:08krazykitlostlogic, someone was looking at it, but i don't think anything happened more than what you just said.
01:56:13 Join keanu|afk [0] (i=4a5ce989@gateway/web/cgi-irc/
01:57:13 Join webguest67 [0] (i=c10bd05e@gateway/web/cgi-irc/
01:57:22 Quit webguest67 (Client Quit)
02:02:14preglowamiconn: yet another test, getting rid of one movem.l and instead fetching values in parallel with mac.l proved a teeny bit faster as well
02:03:27amiconnYou save one cycle per loop with iram data
02:04:26preglowbeh, it's not consistent
02:04:40preglowthink i'll just go with the extra movem.l since that makes for the smaller code
02:04:43amiconnmovem.l needs n+1 cycles for n longwords
02:05:16amiconnmac.l with parallel load needs one extra cycle for the load
02:05:34preglowand that's if the memory is iram
02:06:04preglowand i get one cycle extra in stall too
02:06:21 Join TheDarkOne [0] (n=marc@
02:06:23amiconnIn dram, movem.l is better if it can use line bursts
02:06:36preglowtwo dwords isn't enough for bursting
02:07:39 Quit keanu|afk ("CGI:IRC")
02:07:47TheDarkOneok, so i just installed rockbox on an ipod, worked fine, copied some files over, initialized the db, and restarted it. it flashed something about battery, and shut off, now it appears to not be charging and it won't turn back on, any ideas?
02:07:52amiconnOr rather, the longword reads are 2-bursts, as dram is 16 bit
02:09:17 Join kkurbjun [0] (
02:09:22preglowone more test with four
02:09:51*preglow gets paperclip
02:10:37preglowbah, idiot typo gas should have complained about
02:10:38safetydanTheDarkOne, try leaving it charging for a few hours before trying again
02:11:22TheDarkOneshouldn't it turn on while plugged in though?
02:11:57safetydanTheDarkOne, if it's really low on battery, no. Eventually they come back though
02:12:11TheDarkOneit was half charged before i installed rockbox though
02:12:30preglowamiconn: addq.l #8, %a0 twice and lea.l (%a0, 16), %a0 is about equally fast, yes?
02:12:43safetydanTheDarkOne, can't trust the battery meter :)
02:12:46TheDarkOnetrue :P
02:12:48amiconnThe lea is faster
02:12:59TheDarkOnealright, i'll give it a few hours and if it doesn't work i'll be back :P
02:13:03amiconnIt's single cycle
02:13:12safetydanTheDarkOne, check the forums, you'll see other ipod users with the same problem solved through charging it for a while
02:13:52preglowamiconn: ok, new test, fetching four and four samples is faster by another four percent
02:13:58preglowmust have messed up during the first test
02:14:02TheDarkOnealright. now for music i can put it under Music/Artist/Album/file.ext correct?
02:14:36krazykitanything you like, really
02:14:36safetydanTheDarkOne, you can put it where you want, but most people end up with a structure like that. Just don't put it under .rockbox
02:14:48jhMikeSpreglow: what are you up to? AGC?
02:15:02preglowjhMikeS: wma :P
02:15:12TheDarkOnei'm not even too familiar with the normal ipod way of doing it, my friend loaned this to me today because he never uses it lol
02:15:18preglowjhMikeS: agc will need to do too much per sample to benefit much from this
02:15:25preglowjhMikeS: including a divide per sample :/
02:15:39jhMikeSpreglow: didn't know you worked on WMA at all
02:15:51TheDarkOneand all my music is flac, so first thing was to put rockbox on it
02:15:53preglowjhMikeS: i sometimes do random codec work, especially optimizing
02:17:29TheDarkOneso with rockbox when it's plugged in there is no charging screen?
02:17:29 Join jac0b [0] (
02:17:35TheDarkOneplugged in and turned off that is
02:17:41jhMikeSah. I got lazy and didn't do every possible thing in all the arm asm for mpegplayer. Too tired to bother after writing it :p
02:17:43 Nick fxb__ is now known as fxb (
02:17:44preglowjhMikeS: i have this one wma file that is also very troublesome, i fixed wma up somewhat and found some optimization potential while i was at it
02:17:51 Join sdoyon [0] (
02:18:03preglowjhMikeS: i had a look at that, the idct uses no muls...
02:18:04safetydanTheDarkOne, no, the ipods can sometimes get themselves in to a state where they have too low battery levels to do anything.
02:18:14jhMikeSpreglow: ;D
02:18:44TheDarkOnenot too low to charge though hopefully :P
02:19:02preglowjhMikeS: that's pretty sick, you know :D
02:19:06jhMikeSpreglow: if you think of shorter shift+add sequences, feel free
02:19:40preglowjhMikeS: hell, are you mad, if i had written that, it would be shit full of mul and mla
02:19:42jhMikeSpreglow: hehe, doesn't really look like most IDCT routines does it? There's problem some instruction cuts in the memory writing though.
02:20:46jhMikeSit's all constants though, so test with mul/mla, then convert to shifts
02:22:00preglownow, that's what i call a commit message
02:22:59jac0bsoap: hello
02:26:19safetydanthat's a huge commit message for a single line change
02:26:56jhMikeSok, I got a processor wakeup routines that doesn't depend on timing so back out of IRAM for switch_thread on PP :)
02:28:34 Join shnee [0] (
02:29:14jhMikeSwow, that is a heck of a commit message - nicely formatted too
02:29:15sdoyonsafetydan: OK, maybe I'm a little too verbose... but it's my first commit to playback.c without going through someone else :-).
02:29:35preglowthis coldfire asm uses so many regs newer gccs are bound to puke
02:30:39jhMikeSI'd think older one's would do that. odd.
02:31:07preglownope, newer ones are the only ones to exhibit that behaviour
02:31:47jhMikeShaha...though they did catch that lil' swp thing
02:32:02safetydansdoyon, no it's great to see a proper explanation
02:32:22preglowjhMikeS: running a realtime kernel on my linux now, get 1.13 ms latency doing heavy realtime dsp. i hope rockbox ends up the same :>
02:32:54jhMikeS1.13ms ? that sucks actually
02:33:07jhMikeSoh, wait, you mean audio delay?
02:33:28jhMikeSah, though you meant thread turnaround
02:33:30preglowoh, i mean 1.33
02:33:34preglowno, that would suck
02:33:42preglow1.33 ms is 64 samples
02:33:57jhMikeSnot much more than a HW FIFO
02:34:38preglowi did a test setup running ambisonics encoding on three sources, ninth order (19 channels), decoding to 16 speakers with metering on all channels, all the while tweaking source positions
02:34:45preglownot a glitchj
02:34:50preglownever thought i'd see linux do that
02:35:22jhMikeSheh, but I'm sure there a little more computing power. unless iPL Realtime did this. :P
02:35:31preglowhaha, indeed
02:35:45preglowbut that's quite a shitload of computation anyway
02:36:10preglowi wonder how the hell they manage being able to preempt important stuff
02:36:14TMMyeah, I hate that :P
02:36:23preglowapparently they even preempt critical sections and stuff
02:36:34TMMhow can you preemt critical sections?
02:36:38preglowi don't know...
02:36:39preglowi just read it
02:36:45TMMisn't that the point of 'critical'?
02:37:02jhMikeSyou can preempt them, just not reenter them
02:37:06TMMcritical sections can't be interrupted can they?
02:37:17TMMaren't they supposed to be atomic?
02:37:22preglowjhMikeS: but that'll leave them in pretty inconsistent state, no?
02:38:03preglowoh well, i'll leave all this stuff to you :P
02:38:14TMMthat won't work... unless they aren't very critical :)
02:38:19jhMikeSshouldn't matter as long as they're not altered. I think they must mean that quantums are disabled as short a time as possible. Don't know for sure though.
02:38:47preglowlinux always used to have trouble with long critical sections
02:38:47TMMwho does this anyway?
02:39:06preglowTMM: ingo molnar's realtime patch for linux
02:39:27TMMyou can't interrupt linux's critical section
02:39:35preglowdon't ask me
02:39:44amiconnpreglow: You sure this cf asm is correct?
02:39:50preglowamiconn: it works
02:39:52TMMI thought some outlandish OF or something did this :)
02:39:54preglowamiconn: anything special?
02:40:03amiconnmovem load order is fixed, first d0..d7 then a0..a7
02:40:06jhMikeSpreglow: ?? Is there some other object besides a mutex? preeption of course can happen inside the guarded core or else you wouldn't need the guarded code.
02:40:21preglowamiconn: of course
02:40:30preglowamiconn: then no, that isn't correct at all
02:40:34amiconnmovem.l (%[w]), %%a0-%%a1/%%d4-%%d5
02:40:38preglowi'll fix
02:40:52preglowsounds fine, though :/
02:41:02jhMikeSthe linux scheduler must be really weird then
02:41:49preglowgetting tired now, didn't think of that at all
02:41:51TMMjhMikeS: I'm fairly confident that (on x86) interrupts are disabled during those sections
02:42:41*amiconn wished several times that the movem load order would be switchable :/
02:43:12preglowamiconn: i want the update base reg functionality more
02:43:21preglowjhMikeS: there's some talks of spinlocks being preemptive
02:43:33preglowdon't really know what that signifies
02:44:19preglowamiconn: changing order of movem.l address range and changing order in mac.l should make it right, yes?
02:44:57 Quit kugel ("ChatZilla [Firefox]")
02:44:58amiconnTechnically you wouldn't need to change the order in the movem, but that would be confusing...
02:45:13preglowi wish gas complained
02:45:16TMMmovem? what arch is that? if I dare ask?
02:45:18 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
02:45:19preglowgas doesn't like to complain, though
02:45:23preglowTMM: m68k
02:45:32 Quit Addi2 ("Leaving")
02:45:39TMMah yes, no nothing about that
02:45:56TMMARM's weird enough
02:45:59jhMikeSpreglow: seems odd they wouldn't allow spinlock preemtion in general.
02:46:30*jhMikeS will go look at linux kernel source at some point
02:46:49TMMjhMikeS: linux has had problems with locking on the past, the whole preemtive patchset should illustrate that nicely
02:47:02preglowTHIRD bloody time i have to paperclip my h12
02:47:14preglowsometimes if you skip too fast, it just locks
02:47:17jhMikeSh12-0? hehe
02:47:34preglowannoyed and writing :/
02:47:42TMMnow, THAT would be an interesting molecule :P
02:49:21preglowman, i like coldfire asm
02:49:25preglowso much cooler than arm asm :P
02:49:27TMMdoes anyone know of an 'ease yourself into arm assembly' manual? I'm getting very confused by stuff like eor r0, r1, lsl r3 :)
02:49:57TMMmakes me wish for x86... well, minus mmx
02:49:57preglowjhMikeS: The real-time preemption patch set seeks to provide deterministic response times with a stock Linux kernel. It works by making everything preemptable, including code (spinlock-protected critical sections, interrupt handlers) which cannot be preempted in current kernels.
02:49:57jhMikeSpreglow: I don't know. If you took the best of ARM and CF in one instruction set...
02:50:02preglowjhMikeS: got it from there
02:50:08preglowjhMikeS: that would rock indeed...
02:50:14preglowjhMikeS: but i love the fact it's not load-store
02:50:41preglowTMM: man, x86 asm sucks
02:50:43TMMpreglow: wouldn't it be hugely expensive to preserve state across those sections?
02:50:49preglowTMM: shrug
02:50:51TMMpreglow: x86 asm is EASY
02:51:04jhMikeSif CF had a shifter like ARM, and condition execution, conditional status, that would be most of what's best about ARM
02:51:06preglowTMM: and sucks, few operands, few registers, retarded instruction behaviour in places
02:51:06TMMpreglow: ugly as fuck, but EASY :)
02:51:29preglowlike BOUND, now there's an instruction for you
02:51:43TMMpreglow: do you know of a nice guide with examples? :)
02:51:43jhMikeSBOUND? A memory-barrier instruction?
02:52:22preglowjhMikeS: afaik, it checks if a certain pointer address is within certain bounds, and throws an interrupt if not
02:52:30preglowjhMikeS: the same interrupt used for print screen under dos....
02:52:33*jhMikeS got sick of looking at x86 crap a long time ago
02:52:49TMMbound's just used to do some sanity checking on pointers
02:52:56preglowjhMikeS: best of all, return address would be set to the bound instruction, so it would just trigger and trigger and trigger
02:53:15preglowresult would be that you'd hear BEEP BEEP BEEP BEEP and the printer warming up if you executed some bad code
02:53:30preglowlol indeed
02:53:40preglowi learned to associate the printer warming up with me fucking up
02:53:44jhMikeSsome time after that PortalPlayer was founded, right? :P
02:53:46preglowsince i had long sinced cut the pc speaker wire
02:54:11TMMO come on, you can use it for cheap segfaults
02:54:41jhMikeSat least lock xchg works there, or maybe?
02:54:51preglowlock works with plenty of instrs
02:55:02preglowx86 has tons of memory barrier instructions
02:55:05preglowit's good for that
02:55:18preglowthat is, lock is the memory barrier
02:55:20jhMikeSI'm just wondering if the guys who implemented BOUND implemented SWP on PP :P
02:55:22preglowbut it works with lots of instructions
02:55:55preglowthey had probably drunk the same amount of whisky anyway
02:56:08TMMI don't disagree that x86 sucks, I'd like to transcend into arm, does any of you know of a good way to learn it? :)
02:56:11TheDarkOnehow does rockbox estimate battery time remaining?
02:56:20TheDarkOnebecause it seems to be completely off lol
02:56:25TMMTheDarkOne: depends on the target
02:56:33preglowTMM: i found it really easy to get into, and i just got the quick reference sheet
02:56:45jhMikeSI wouldn't mind ARM with a lock prefix
02:56:50TMMpreglow: I keep getting confused with all the shifts
02:57:30 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
02:58:54*preglow finds the shifts easy
02:58:56jhMikeSa chainable one that locks until the next instruction without lock. so ldrl r0, [r1], bicl r0, r0, #1, strl r0, [r1], bx lr or some such. They'd never
02:59:35preglowwould be cool, yeah
02:59:50preglowthis is why i hate load-store
03:00:58TMMthat's ONE instruction?
03:01:07jhMikeSyou'd see thing like: stmneial <= :P
03:01:48preglowx86 has crazy-ass shit like that
03:02:17hcscmpxchg16b.otm (on the moon)
03:03:06 Quit Gnu47 (Nick collision from services.)
03:03:19 Join Gnu47 [0] (
03:03:28jhMikeSTMM: about what I wrote? No, it's three instruction that perform locked until the next instruction without a (l)
03:03:47TMMjhMikeS: ah, good :P
03:04:25preglowand then there will be cake!
03:05:30TheDarkOnei can't copy UTF-8 to an ipod, can i
03:05:36TheDarkOnefilenames that is
03:05:50preglowdepends what you mean
03:05:54preglowfat is unicode
03:05:58preglowbut utf-16
03:05:59TMMyou are just prejudiced :)
03:05:59TheDarkOneis it?
03:06:07|Rain|vfat/fat32 is, yeah
03:06:09TheDarkOnewell it's from an NTFS filesystem
03:06:14TheDarkOneto an ipod
03:06:16preglowntfs is utf-16 too
03:06:16jhMikeSunicode can be fat at well
03:06:18preglownot utf-8
03:06:18TMMit's called CISC, the 'C' has meaning you know :P
03:06:18|Rain|NTFS is also utf16
03:06:27TheDarkOneok then
03:06:36TheDarkOnewill rockbox read Unicode filenames and tags?
03:06:41preglowi want an instruction set that has more acronyms
03:06:44preglowso far i only have asl
03:06:59|Rain|TheDarkOne: yep
03:06:59preglowi'd like lol, rofl, omfg and also ffs
03:07:07preglowftw and bbq would be bonuses
03:07:21TheDarkOnethat is good then :)
03:07:30preglowjhMikeS: nice "clear interrupt flag" instruction :P
03:08:16jhMikeSand SAS too.
03:08:32jhMikeSSit And Spin :P
03:09:46preglowright, and NOW i see vector_fmul_reverse
03:10:02TheDarkOneit's vorbis tags though, not id3 :P
03:10:07TMMunicode? all this hippy crap, EBDIC was good enough for everybody!
03:10:09TheDarkOneor ogg tags whichever it is
03:10:15preglowi just need to do this before bed
03:10:23markunTheDarkOne: vorbis tags are always UTF-8
03:11:45sdoyonmount /usb1#
03:12:23***Saving seen data "./dancer.seen"
03:12:55 Quit sdoyon ("ircII EPIC4-2.4 -- Are we there yet?")
03:13:43jhMikeSTMM: hehe
03:13:49TMMin MY days, we had 7 bit ASCII and we were HAPPY... we had to use use codepages to read our french texts, uphill both ways through the snow
03:14:33TMMon the good days, when it wasn't raining glass shards
03:14:41 Quit Gnu47 (Nick collision from services.)
03:14:51 Join Gnu47 [0] (
03:15:15TheDarkOnedoes the default font for 4gen ipod have UTF-8 characters?
03:16:47 Part jac0b
03:17:14preglowgah, what's the syntax for the full indexed addressing mode again
03:17:54jhMikeSon CF? move.l rx, (%an)+ ??
03:18:02TMMwhat's the syntax for 'shift every bloody register by each other register' again? ;)
03:18:32jhMikeSwhere's that? could be useful.
03:18:49preglow(offs, %base, %index*[1, 2, 4])
03:18:49preglowit apperas
03:19:03jhMikeSI forget but it's in the resampler
03:20:23preglownow, let's see if that helped
03:20:25TMMa yes, now where's the 'shift every uneven register by every even register except pc?'
03:20:47hcs'...on alternate weekdays'
03:20:54preglowTMM: you can shift pc, you know :P
03:20:56preglowon arm, that is
03:20:57TMMno, I'm moving up to that
03:20:57preglowwhich rocks
03:21:11TMMpreglow: I just didn't want it on this case :)
03:21:31preglowpc is the same as any reg, heh
03:21:44preglowholy feck, coldfire just got faster than arm
03:21:47TMMpreglow: very well, I'll just call it r15 then, ok? :)
03:22:01jhMikeSyeah, that's another cisc gripe I have
03:22:03TMMpreglow: happy now?
03:22:26preglownot bad for about 30 lines of code
03:22:48TMMpreglow: is that a challenge? ;)
03:23:13preglowsure, why not :P
03:23:50TMMpreglow: well, I've got an ARM target to worry about, and I can't have you gloat too much, I'll get that down to just 250% faster :P
03:25:38 Join Rob2222 [0] (
03:26:07TMMI read up on the whole harvard vs neumann (??) design, sounded very interesting
03:27:20jhMikeSquit talking and do the work then :)
03:27:56preglowneumann 4 ever
03:28:04 Quit keanu ("Leaving")
03:28:17TMMI'm presently deciphering some rather confusing LCD routines in my OF
03:28:32 Join keanu [0] (n=keanu@unaffiliated/keanu)
03:29:30 Quit jhulst ("Konversation terminated!")
03:30:13TMMplus, it's 3:30am I doubt I'll be of much use
03:30:27preglowlet's see what nano gained
03:31:19 Quit Toxicity999 (Read error: 104 (Connection reset by peer))
03:31:35jhMikeShopefully it lost something
03:32:17TMMlike... encryption
03:32:28 Join webguest00 [0] (i=0cb5b607@gateway/web/cgi-irc/
03:32:35preglowcurrently at 223%, now clocking in at...
03:32:48TMMsee, it's going down without me even doing anything
03:33:41preglowwtf, 225%
03:33:53jhMikeSyou know, the red deltas that happened with one change make no sense at all since pretty much a perfect reversal of them didn't bring it green by a compareable amount.
03:34:13preglowwell, then figure out a better metric, i don't know how the current one works
03:34:35TMMwell, I give up :) 4 functions that seem to do the same thing, probably dead code... however when I flashed a way to old version of the OF on my new player the lcd malfunctioned
03:34:40jhMikeSIt's supposed to be (bin+mem)/2
03:34:42TMMso, perhaps it's not dead
03:35:32TMMand there's a lot of shifting
03:35:41webguest00Does anyone have a couple of minutes to help guide me through removing Rockbox from my ipod nano? I'm upgrading to a Sansa and want to make my nano ebay'able. I've tried using the instructions in the wiki with no success...
03:36:08*jhMikeS looks at the red commit and compares its reversal
03:36:39preglowwebguest00: wiping .rockbox directory then unisntalling with ipodpatcher should do the trick
03:37:00preglowwebguest00: better yet is just making itunes restore it
03:37:23webguest00I never installed itunes :) I've tried ipodpatcher...let me try wiping the .rockbox directory.
03:37:30preglowwebguest00: windows? linux?
03:37:46webguest00Old rockbox install, probably 1 1/2 old...
03:37:55 Quit pill (Read error: 104 (Connection reset by peer))
03:37:55preglowbefore the days of ipodpatcher
03:38:01preglowthen that won't help you much, i think
03:38:21webguest00It doesn't seem to do anything
03:38:33 Join pill [0] (
03:38:38preglowif you were in linux, you could download the firmware image straight from apple and "dd" it over to the nano for something pretty much like an itunes restore
03:38:42preglowbut i don't know how to do that in windows
03:38:51TMMpreglow: there's rawrite...
03:38:54 Quit shnee ("Konversation terminated!")
03:38:59preglowTMM: well, you take over from here, then :P
03:39:01krazykitthere's also dd.exe, i think
03:39:06jhMikeSso taking a bunch of queues out of IRAM that aren't used by IRAM functions was most of the increase? wtf. I give up making sense of GCC.
03:39:13webguest00Will removing the .rockbox folder brick my nano?
03:39:14preglowthat's where you get the firmware images
03:39:18preglowwebguest00: no
03:39:20safetydanwebguest00, you're best bet is probably installing itunes then doing a factory restore
03:39:59TMMpreglow: that's the extend of my knowledge from writing linux boot disks :)
03:41:07preglowthink i'll leave these babies for commit tomorrow
03:41:11preglowsorely need sleep now
03:41:20webguest00thanks for your help preglow
03:43:57 Quit Rob2222 ()
03:44:09 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
03:45:21 Quit webguest00 ("CGI:IRC (EOF)")
03:45:32 Quit HellDragon (Client Quit)
03:45:51TMMso, anyone here know anything about 24 year old cats? :)
03:46:13|Rain|they have a hard time using rockbox's menus
03:47:58TMMwell, since noone is doing any on-topic talking
03:48:19|Rain|my cat may be 24 cat years old
03:48:25jhMikeSthen usually we remain silent :)
03:48:47TMMit's sister died last week, and now it's pissing everywhere, very inconvenient the vet suggests putting her down, but I find that a bit harsh
03:49:57TMMshe's a happy little cat, but a bit senile, I guess
03:50:29*jhMikeS suggest such things go to #rockbox-community or PM 0:)
03:50:34krazykitplease take off-topic stuff to #rockbox-community
03:50:40 Quit ToHellWithGA ("You know you'll miss me a lot.")
03:50:54TMMoff-topic, this? :)
03:52:44TMM|Rain|: #rockbox-community ?
03:53:22jhMikeSthere is a #rockbox-jerks and I found myself to be the only one there ... makes sense :)
03:53:58*jhMikeS should start #rockbox-pets
03:54:01TMMjhMikeS: well, I'm there now, turns out, I'm the only one there
03:54:29|Rain|TMM: I don't really have much to offer, I've never had to snap my cat out of it. I've heard keeping the box extra clean can help, and I've heard that there are spray deterrents, but that's about it
03:56:05 Join Rob2222 [0] (
03:56:37scorche|Rain|: 3 people said #rockbox-community...
03:57:29|Rain|and it still would've taken one line to say "that's all I've got" without an extra sentence of information
03:58:00Llorean|Rain|: Frankly, that doesn't make your one sentence any less off-topic.
03:58:29|Rain|no, but joining a channel to utter a single sentence isn't any better
03:58:30 Join penguinator [0] (
03:58:48|Rain|we've already cluttered the logs far more by not dropping this
03:59:23 Quit Rob2222 (Read error: 104 (Connection reset by peer))
03:59:30Llorean|Rain|: No, the proper response was "I'm sorry, and will remember not to even start off-topic conversations in here in the future. I've been around enough to know better."
03:59:42 Part |Rain|
04:00:00scorchethat doesnt really matter, as the purpose of this conversation is to prevent it in the, it uses up a bit of lines in the logs in order to prevent future ones being entered
04:01:01penguinatorIs there a way to keep rockbox from entering disk mode when plugging it into a computer? i just want to charge it, not have use the disk. Or alternativly, is there a way to disconnect the drive mode without disconnecting the cable?
04:01:04 Join billytwowilly [0] (
04:01:28scorchepenguinator: hold menu as you insert the cable
04:01:59penguinatorscorche: that didn't work
04:03:48Soappenguinator, you using a stock build on what I assume is an iPod?
04:04:05penguinatoryes, it does work actually, you just have to do it quickly
04:05:04 Part penguinator ("Thank you, sorry for not trying it more...")
04:06:54 Join shnee [0] (
04:14:48 Join HellDragon [0] (n=Nocebo@unaffiliated/helldragon)
04:16:18 Quit Mouser_X (Read error: 104 (Connection reset by peer))
04:16:32 Join Mouser_X [0] (n=Mouser_X@
04:23:09TMMif I might drop a suggestion, as one of the worst offenders of both off topic conversations here, as well as on-topic conversations in #rockbox-community, shouldn't someone set up a #rockbox-dev?
04:24:08Soapthat is what this channel is.
04:24:10safetydanTMM, it's been talked about before but in general the devs prefer to have one channel to monitor
04:25:03TMMjust a thought...
04:25:16TMM-dev is more clear perhaps, that's just it
04:26:08TMM-community should be, #rockbox, and -dev should be,well, what this is now :) but with more banning
04:26:33safetydanTMM, it's not all about dev stuff here though and it's handy to see what users are complaining about
04:26:40scorchewell, that isnt going to happen any time soon
04:26:49TMMlike I said, just a thought
04:26:59TMMsuggestion, that's it :)
04:27:13scorchesearch the mailing list next time ;)
04:27:15 Join homielowe [0] (
04:27:36TMMscorche: I just subscribed :P much, much too much trouble to search through all of it
04:27:44TMMscorche: and I'm not the only one ;)
04:27:52 Join Reckase [0] (n=erik@
04:28:02Reckasehi folks
04:28:16scorcheTMM: google search <query>
04:28:31TMMscorche: ok, put that in the topic ;)
04:28:35ReckaseI've run into some problems today and I was hoping to see if others have had the same problems...
04:28:44scorcheTMM: why would i do that?
04:29:12ReckaseI'm getting the "Building Database... 438 found (Prev to return)" message
04:29:14safetydanReckase, ask away
04:29:29Reckasethis is a sansa e250
04:29:32TMMscorche: because I'm actually trying to make a contribution here, and I'd probably go through that trouble should I have known about it... might save some time
04:29:53ReckaseI had it working a few days ago, but in the process of trying to get it to work with one of my other machines, it's gone cracker
04:29:58 Join _pill [0] (
04:30:09safetydanReckase, that seems like a normal message during the database build. Is it just stuck on that screen?
04:30:09 Quit pill ("changing servers")
04:30:11 Join Rob2222 [0] (
04:30:16Reckaseyes, it just stays stuck
04:30:26Reckaseand there's no files on it
04:30:31 Nick _pill is now known as pill (
04:30:32Reckaseno music, that is
04:31:00safetydanReckase, sounds like your filesystem might be failing. Try a chkdsk /f
04:31:08Reckaseunder windows?
04:31:18Reckasethis is a new (well, refurb) i got a week ago
04:31:45safetydanReckase, windows is probably easier, I forget what the linux version of that is
04:32:00Reckasei reformatted the 1.8 gb partition fat32, and it's still happening. I can try the chkdsk
04:32:10TMMscorche: bad reason? :)
04:32:25 Part pixelma
04:32:58ReckaseI'll be back after chkdsk.
04:33:19Reckasedo I do that with the rockbox firmware loaded or the sansa original?
04:33:32scorcheTMM: teaching people how to use google is does not relevant for the #rockbox topic
04:33:54safetydanReckase it shouldn't matter
04:34:00Reckasethx - back in a bit
04:34:01 Part Reckase
04:35:42TMMscorche: I suppose
04:39:04 Join eigma [0] (n=cat@
04:42:42 Join Reckase [0] (n=Erik@
04:42:49Reckaseok, so I did a chkdsk
04:43:00Reckasesaid that some allocation units weren't valid
04:43:09Reckaseand said that it fixed them
04:43:23Reckasei did it a second time and it didn't find any problems.
04:43:51Reckasehowever, if I try to do a 'Complete Installation' from windows, it says that it can't find a sansa at that path
04:44:19 Quit Rob2222 (Read error: 104 (Connection reset by peer))
04:46:03 Nick fxb is now known as fxb__ (
04:46:27 Quit donutman25 (Remote closed the connection)
04:46:38 Part Reckase
04:49:27 Join sdoyon [0] (
04:50:04 Join Reckase [0] (n=erik@
04:50:58 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
04:51:26 Join miepchen^schlaf [0] (
04:52:23 Part Reckase
04:52:54 Join Rob2222 [0] (
04:53:44sdoyonI seem to have an issue with a commit about PCM interface unification: r15006 | jethead71 | 2007-10-06.
04:54:37sdoyonIt's causing the end of some .talk clips to be cut off. It happens on my x5 and never on my e200.
04:54:38jhMikeSyeah, what'd I bust up
04:54:59sdoyonI believe I have narrowed it down to one particular change. Anyone knows anything about this stuff?
04:55:18jhMikeSI made the commit, so can look into it
04:56:20jhMikeSsdoyon: _just_ .talk clips?
04:56:49 Join saratoga [0] (i=9803c6dd@gateway/web/cgi-irc/
04:57:25 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
04:57:48sdoyonOn a hunch, I reverted (put back) the part that says /* There may be more data waiting to flush, try to use it */ ... goto process_new_buffer;. Hopefully that's a big hint? ;-)
04:58:00sdoyon... I mean that seemed to fix it.
04:59:15jhMikeSit's a dangerous thing to do. I think some other method needs to be employed.
04:59:24sdoyonI'd like to be more specific, but the clipping of my filenames wasn't always exactly on the same names as I tried various revs (bisecting to find the one where this started).
04:59:45 Quit eigma ()
05:00:00saratogapreglow gives me crap for doing a 64 bit div in the wma decoder
05:00:09sdoyonjhMikeS: I'm not saying it's necessarily a proper fix. I won't claim to have even tried to understand what that patch does, but that first bit jumped at me and I tried it.Should be a good hint hopefully?
05:00:14saratogaI do ONE 64 bit divide per block
05:00:38saratogathats what, maybe 20 of them a second
05:00:40saratogabig deal
05:01:19saratogalast i checked i was still faster then tremor and faac
05:01:50 Quit Gnu47 ("You know you'll miss me :P")
05:01:55jhMikeSI think it means the interrupt fired but no data was yet ready. Ususally the pcm buffer would forcefully grab the data from the interrupt but that function contains stuff that isn't safe in that context.
05:02:07 Join Reckase [0] (n=erik@
05:02:59 Join midgey_ [0] (
05:03:01Reckaseok, I'm back. I've reinstalled after running chkdsk, and I'm still getting the database error.
05:03:29Reckaseit's getting a little further now (at least the theme is showing up)
05:03:46sdoyonIt seems to always be at the end.
05:04:00Reckasenow it says '497 found'
05:04:23Reckaseperhaps I should put something on it?
05:04:32Reckasecould it be the lack of files that confuses it?
05:05:20Reckaseit's the fact that it worked before that's getting me so frustrated
05:07:03jhMikeSsdoyon: And if playback is goind, then no cutoff?
05:08:08sdoyonjhMikeS: err that'll take a recompile... hold on.
05:08:50jhMikeSThe PCM buffer is always running there, so I'd suspect all data would be played out in that case.
05:10:02sdoyonjhMikeS: yes, correct.
05:11:04 Quit saratoga ("CGI:IRC")
05:11:12Reckaseaha! it must have music on it for the db to work properly
05:11:31 Part Reckase
05:12:26***Saving seen data "./dancer.seen"
05:14:14sdoyonjhMikeS: It's pretty noticible: I have "boo" instead of "books", "musi" instead of "music", and on files the end sounds like "dot aw" instead of "dot o g g"... some are fine though. and so long as I don't change to another rev, it seems a name that gets clipped always gets clipped at the same point. So those that are clipped are always clipped.
05:15:05sdoyonWhat I don't understand is how I did not notice that last weekend... :-)
05:16:13 Join bb [0] (n=bb@unaffiliated/bb)
05:17:24jhMikeSThe difference for e200 vs. x5 might be explainable too. x5 uses genuine HW DMA with long transfers where as e200 does small bits with SW DMA. hmm.
05:18:00jhMikeSThe callbacks should happen in the same place but timing will vary.
05:19:26 Quit midgey (Read error: 110 (Connection timed out))
05:23:01jhMikeSI imagine getting at least a couple frames in there before starting a clip may help. I think it starts playing the first frame right off. The delay would be a few ms more at most.
05:24:49sdoyonjhMikeS: even though the missing part is at the end?
05:26:17 Quit pill (Read error: 104 (Connection reset by peer))
05:26:45sdoyonjhMikeS: Why does calling pcmbuf_flush_fillpos() appear to help?
05:27:36 Quit bb_ (Read error: 110 (Connection timed out))
05:30:11jhMikeSperhaps because it forces a pcm start if it's not playing
05:31:38jhMikeSThere were some other buffer changes I had that didn't get included with that commit. I thought it was a bit out of scope to include there.
05:32:15sdoyonjhMikeS: need me to try?
05:33:11 Part hcs
05:33:12jhMikeSI don't think they'll apply now.
05:35:21jhMikeSI'll give it a look and try to resync. basically the pcm buffer gets its own playing state and some other changes for interrupt safety
05:38:51sdoyonjhMikeS: OK. Feel free to send it to me so I can try it on my x5.
05:39:02sdoyonjhMikeS: The bt I reverted, is it more dangerous now than it was before?
05:41:54 Join Jsunu [0] (
05:41:59 Part Jsunu
05:42:23jhMikeSyou put the call to that back?
05:43:01 Join Jsunu [0] (
05:43:09sdoyonjhMikeS: in my test build, not in svn obviously.
05:43:39sdoyonThe part that does goto process_new_buffer;
05:44:02 Quit Jsunu (Client Quit)
05:44:07jhMikeSit is much more if so because on the first call to the callback is called with pcm_playing false via pcm_play_data. it will always then make that call and double-enter the pcm driver.
05:44:46 Join SirFunk [0] (
05:45:18sdoyonjhMikeS: hmm... funny how doing the right thing isn't always obvious. seemed to work so much better :-). OK then.
05:45:31jhMikeSI can't imagine a talk clip shouldn't be fully decoded and buffered before DMA runs out
05:47:11sdoyonjhMikeS: I have one directory named "baba" (for tests obviously). I hear just "ba".
05:47:56 Quit midgey_ ()
05:48:30 Join ramon8 [0] (
05:49:23jhMikeSMyself, I refuse to do things myself I know are not safe measures even if I have to think harder about it and the unsafe way seems to work fine. It won't always. The safe way I can be sure always will be safe.
05:54:33jhMikeSsdoyon: do some .talk clips have some silence at the end which other are trimmed well?
05:54:57sdoyonjhMikeS Sure. As I said, I'm not considering this as a long term fix, and I'll test anything you need. Just trying to see how best to get by in the meantime.
05:56:05sdoyonjhMikeS: they're all generated the same way... but perhaps depending on the phoneme at the end...
05:56:57jhMikeSI get failed hunks with the diff I made. Definitely need some resyc time.
05:57:42 Quit iamben (Read error: 104 (Connection reset by peer))
05:57:45sdoyonjhMikeS: I believe there's a bit of silence enqueued at the end in all cases in talk.c though.
05:59:15sdoyonjhMikeS: mp3_callback adds this p_silence clip, which is from tools/VOICE_PAUSE.wav.
06:00:08sdoyonjhMikeS: lasts 0.3s.
06:02:58sdoyonjhMikeS: I'll need to be going soon. I had another question on a different subject, if I may bug you further for a minute... Could by any chance the e200 be made to wakeup / turn on from something other than the POWER button, such as perhaps the hold button instead?
06:03:00 Join iamben [0] (
06:04:20jhMikeSsdoyon: No. That's hard-connected to startup the power supplies. could rewire the PCB :)
06:06:29sdoyonjhMikeS: Yeah OK :-).
06:07:35 Join Chronon [0] (
06:13:07ChrononI just stumbled on the Disktidy bug today. I'm not clear on whether I can simply bump the value of MAX_OPEN_DIRS to work around this.
06:14:09safetydanChronon, you can but you may run out of stack.
06:14:44jhMikeSis that because of recursion?
06:14:47 Quit sdoyon ("ircII EPIC4-2.4 -- Are we there yet?")
06:15:08ChrononHmm.. well running Disktidy isn't _that_ important.
06:15:29safetydanjhMikeS, yes
06:16:28safetydanshould be possible to convert it to iteration but no one's bothered
06:21:02ChrononThe offending code is in the Disktidy plugin itself?
06:21:20safetydanChronon, yup.
06:21:38Chrononok. Thanks, safetydan.
06:24:47 Join psycho_maniac [0] (
06:29:15 Quit TheDarkOne (Read error: 104 (Connection reset by peer))
06:39:11 Quit jhulst ("Konversation terminated!")
06:53:27 Quit psycho_maniac (Read error: 113 (No route to host))
07:02:18 Quit XavierGr (Nick collision from services.)
07:02:21 Join XavierGr [0] (
07:06:29 Join jenky [0] (
07:12:29***Saving seen data "./dancer.seen"
07:13:12 Join BigBambi_ [0] (n=Alex@rockbox/staff/BigBambi)
07:17:25 Part jenky
07:30:51 Join psycho_maniac [0] (
07:31:57 Quit BigBambi (Read error: 110 (Connection timed out))
08:00:41 Join leftright [0] (i=c414a3a0@gateway/web/cgi-irc/
08:01:26leftrightSlasheri, you around
08:05:02leftrightH140, ROM flashed, dircache enabled, file tree browsing, I notice that there are 11 dircache.dat files listed in .rockbox, just wondering if this is a problem
08:06:32 Quit homielowe (Remote closed the connection)
08:07:18 Part leftright
08:10:29 Join Rob222241 [0] (
08:18:34 Join kubiix [0] (
08:26:12 Join bertrik [0] (
08:27:38 Quit shnee ("Konversation terminated!")
08:28:45 Quit Rob2222 (Read error: 110 (Connection timed out))
08:29:19 Part safetydan
08:30:45 Join sin613 [0] (n=single@
08:37:02 Quit BigBambi_ (Read error: 113 (No route to host))
08:40:32 Quit Calcipher (Read error: 110 (Connection timed out))
08:46:01amiconnpreglow: around?
08:47:14sin613amiconn: am i correct in remembering your name attached to some idct coldfire optimizations?
08:47:30amiconnYes, based on mirak's work
08:47:50sin613nice. that's one hell of a speed-up
08:48:23linuxstbsin613: Which device are you using?
08:48:45amiconnYeah, video on X5 is now usable with only occasional skipping
08:48:47sin613i posted the framerates to the wiki
08:49:07amiconnThere's still ~10% more potential I know of...
08:49:11linuxstbI'm just reading the wiki - it's making nice reading ;)
08:49:19amiconnlinuxstb: Will you add the codec stack seizing?
08:50:14linuxstbamiconn: Sure, I'll do it now.
08:50:48sin613amiconn: i'm not sure if i asked you before, but is the transform really idct, or is it a properly quantized fft?
08:51:13 Join Ebert [0] (
08:51:49amiconnIt's an idct, I'm no expert on that though
08:52:46Ebertjust want to chime in that I Love the gapless-fading playback and new color choices for the gui ;D
08:53:02amiconnThe motion compensation stuff is more the kind of stuff I like fiddling with (shuffling bits all around)
08:53:14Eberthope things stay in this direction as far as customization
08:53:44sin613is the ~10% potential gain within motion comp?
08:54:18amiconnNo, with putting a bunch of const arrays into iram that don't fit atm
08:54:47linuxstbamiconn: Can you think of a reason for test_codec needing to restore the contents of the codec stack? I can't remember why I bother doing that...
08:54:51sin613if the transform is straight up idct and not a quantized fft, there's room for additional gain there as well
08:54:54amiconnlinuxstb: Btw, the optimised C yuv blitting that got committed for ipods is buggy. I get twinkling pixels all over the video on H10 now :|
08:54:57sin613from what i understand
08:55:20amiconnlinuxstb: Umm, to not crash the codec thread?
08:55:22linuxstbamiconn: Hmm, I never noticed that on my Color (assuming it's used there.
08:55:38linuxstbamiconn: But playback is stopped. So I assume the codec is reloaded...
08:55:51sin613i'm having a difficult time wrapping my mind around either, though...
08:55:51linuxstbAh, it's more than just the codec... Of course.
08:56:18amiconnlinuxstb: It only got committed in r15147
08:56:34amiconnMaybe you have an older build on your color
08:56:58linuxstbamiconn: OK, I didn't notice that commit... I remember committing some optimisations myself in the past, and thought you were referring to that.
08:57:41amiconnNo, preglow committed this, on Tuesday
08:57:57 Join aliask [0] (
08:59:55 Join ddalton [0] (
09:01:21 Join ender` [0] (
09:10:33 Quit Mouser_X (Read error: 110 (Connection timed out))
09:12:32***Saving seen data "./dancer.seen"
09:16:24 Quit bertrik (Read error: 104 (Connection reset by peer))
09:22:52linuxstbamiconn: - it seems to work for me...
09:24:00 Join petur [0] (n=petur@rockbox/developer/petur)
09:24:47 Quit Ebert ()
09:30:58 Join Hukt [0] (
09:31:37GodEater_linuxstb: do you know if your gcc multilib suggestions have been put into svn yet ?
09:31:50 Nick fxb__ is now known as fxb (
09:31:58linuxstbGodEater_: No they haven't. I guess I'll need to do it.
09:32:22linuxstbIt seems to be working fine - I've been using it for a while (and so has my build server)
09:32:46GodEater_will it mean buildservers will need to re-run ?
09:32:49linuxstbThe mrobe (DM320) port is also using the wrong -mcpu parameter in configure...
09:33:12JdGordonit is?
09:33:17linuxstbGodEater_: Yes, but only before ARMv5/ARMv6 ports make it to the stage of having automatic builds.
09:33:48GodEater_what armv5 targets do we have ?
09:34:00GodEater_(in the works)
09:35:26JdGordonlinuxstb: im playing with the mr500 now, can you tell me what the configure should be changed to?
09:37:22linuxstbJdGordon: If google is correct, you have a ARM926EJ-S core. You'll need to check the gcc docs for the correct -mcpu option
09:37:48linuxstbBut you'll also need to rebuild gcc with the multilib changes I emailed the dev mailing list about - to enable building of an armv5 libgcc.a
09:38:21linuxstbGodEater_: mrobe and the Telechips devices. Gigabeat S is armv6
09:38:35linuxstb(mrobe as in DM320)
09:38:39 Quit qweru ("moo")
09:38:54GodEater_I knew the S was armv6... just didn't know what v5 stuff we had
09:39:09 Nick ddalton is now known as ddalton_ (
09:39:20 Join davina [0] (
09:39:54amiconnlinuxstb: The twinkling pixels bug doesn't exist on the G5. Maybe it also doesn't in the Color/Nano driver
09:39:55linuxstbJdGordon: If you can wait another day, I'll try and put a patch for gcc on the wiki, and update
09:40:03amiconnDid you check?
09:40:21JdGordonlinuxstb: im happy to wait :)
09:40:22linuxstbamiconn: I didn't notice anything. Should it be obvious?
09:40:24 Quit ddalton_ ("leaving")
09:42:07linuxstbIs it just me, or does mpegplayer also take longer for the initial buffering (after pressing play in the start menu) than it used to?
09:42:33amiconnIt takes quite long on PP. I blame jhMikeS's threading stuff :>
09:43:13linuxstbI remember making the initial buffer fill (before playback started) quite short, it then continues buffering in the background. Maybe that was changed.
09:43:20amiconnOn single core it's fast..
09:44:42jhMikeSwhen did that change?
09:45:04jhMikeSI know the seeking stuff made it take longer
09:45:24linuxstbI noticed it as soon as the start menu/seeking patch was committed.
09:45:58jhMikeSI think a fix to that was removing some buffer zeroing in mpeg2_alloc
09:47:33amiconnjhMikeS: Completely different thing: As I know you made the mpegplayer arm idct completely without using mul/mla, could this method also be applied to e.g. libmad's imdct?
09:47:39jhMikeSthat's what I was told anyway but I don't see any memset there
09:47:49amiconnNote that I don't know how that one looks, so...
09:48:00linuxstbNo, it's changed - the old code read MPEG_LOW_WATERMARK bytes in the initial read, the current code fills the whole buffer in the initial read...
09:48:59jhMikeSamiconn: with 32-bit constants? It's not really good for fractional fixed point and I'd say impossible or at least for anything with 64-bit intermediate values.
09:49:32amiconnhmm. What about jpeg?
09:50:12jhMikeShmmm...some values aren't large constants
09:50:13amiconnDo mpeg and jpeg use the same idct?
09:50:30jhMikeSmpeg audio uses imdct
09:50:44*amiconn could even try to write an SH1 idct using mac ...
09:50:45jhMikeSjpeg should look like mpeg video I'd imaging
09:50:56amiconnWould lower the excessive decoding time of large lpegs
09:51:40jhMikeSeven large values might be decomposeable into a few shift+adds but just think of how large a set of possibilities that is to find it.
09:51:42amiconnlpeg == large jpeg ;)
09:52:23 Quit Hukt0nWax (Read error: 110 (Connection timed out))
09:52:23 Nick Hukt is now known as Hukt0nWax (
09:52:45JdGordonanyonw know where the pressed button is checked to turn off the dap?
09:52:47jhMikeSThe K and L constants are big numbers but the X and x are just asking for that conversion perhaps
09:53:00linuxstbJdGordon: I think it's in the button driver IIRC
09:53:43jhMikeSand different for x5 which uses a key interrupt direct to the queue
09:53:58 Join safetydan [0] (n=safetyda@rockbox/developer/safetydan)
09:55:35amiconnlinuxstb: Do you know whether libmpeg2 can be told to only decode luma?
09:55:51linuxstbNo idea, I've never looked inside libmpeg2.
09:56:07 Quit sin613 (Read error: 110 (Connection timed out))
09:56:32 Join pixelma [0] (i=pixelma@rockbox/staff/pixelma)
09:57:49*jhMikeS hopes the stream parser will become error tolerant
09:59:24linuxstbamiconn: Have you tried my iram patch? I want a second opinion before committing...
10:00:57 Join pondlife [0] (n=Steve@rockbox/developer/pondlife)
10:01:46jhMikeSshouldn't there be reserve for the codec thread to possibly run somewhat?
10:03:05jhMikeSnew stackend = current sp - some_slack_space?
10:05:26 Join bagawk_ [0] (n=lee@unaffiliated/bagawk)
10:05:30*jhMikeS though about a counterpart to thread_thaw() - thread_freeze() as well - then it won't run till thawed again
10:05:41 Join Zagor [0] (n=bjorn@rockbox/developer/Zagor)
10:07:17amiconnlinuxstb: Seems to work.
10:07:29amiconnI started a video while music was playing, and voice enabled
10:07:50amiconnVideo started, no crash. Exiting and restarting music works as well
10:08:26linuxstbjhMikeS: You mean that mpegplayer should give the codec thread a new stack?
10:08:51linuxstbThat's very possible - it could just use the copy that's been made...
10:10:06 Join ddalton [0] (
10:10:56jhMikeSyou could do that do, but I'm thinking of a way that doesn't involve tweaking the structures and is "safe" (better know what it's up to first) for any configuration.
10:10:56*amiconn goofed
10:11:37jhMikeSof course if something on the stack refers to the old stack, it's doa
10:12:14linuxstbjhMikeS: That's true... Probably a bad idea then. But is there any reason the codec thread should become active?
10:12:32psycho_maniaclinuxstb: did you commit the iram patch you were talking about?
10:12:51amiconnDoesn't change the result though
10:12:53jhMikeSit could be active at the time a yield occurs but not blocked on the queue
10:12:58linuxstbpsycho_maniac: No.
10:13:41psycho_maniacalright just curious because i saw something in svn recently about iram in ipod. i guess your name is not mark. ;)
10:13:42jhMikeSthread_freeze would keep it inactive whether or not it's on the COP or same core
10:15:48linuxstbjhMikeS: Is there anything my patch should be doing, or do you think it's OK to commit?
10:15:52 Quit bagawk (Read error: 110 (Connection timed out))
10:16:44linuxstbpsycho_maniac: No, I'm not Mark.
10:17:20linuxstbpsycho_maniac: All my patch does is move something from one part of iram to another, freeing up more space for mpegplayer to use in the future.
10:17:45psycho_maniacahhh i see. thanks for the info
10:17:53jhMikeSjust leaving slack space in case it has to run a bit before blocking on the queue I think
10:18:45linuxstbSo don't steal the whole stack, just the currently unused part, minus a little bit?
10:19:26amiconnThat might not be enough...
10:19:53jhMikeS...or devise a way to be sure it is waiting on the queue.
10:20:01linuxstbYes, I would prefer the latter...
10:20:41linuxstbCould I check the current state of the codec thread, and yield until it changes to something good?
10:21:58linuxstbSo something like "while (codec_thread.state != STATE_BLOCKED) rb->yield();" ?
10:22:46safetydananyone mind a minor patch to the mpeg player to display the fps on the lcd remote if it's there?
10:22:52jhMikeSit can block other places though. hmmm.
10:23:33amiconnsafetydan: As long as it only does this when fps display is enabled...
10:23:49amiconnNotice that doing so will slow down mpegplayer slightly
10:23:58jhMikeSI guess the playback stop will wait until it sets audio_codec_loaded to false, and then it just blocks after that
10:24:01linuxstbIt could be nice to display it on the remote instead of the main LCD, if the remote is present.
10:24:02amiconn...especially if ticking reduction is enabled
10:24:59*amiconn would like that "Test mode" option instead of the 3 options that all need changing for performance tests...
10:25:18jhMikeSlinuxstb: that should work as long as it's not a COP thread. codec_thread.state could be seen as STATE_BUSY there.
10:25:18 Join obo [0] (n=obo@rockbox/developer/obo)
10:25:48linuxstbsafetydan: I noticed in the logs that preglow asked you to sync the hw eq removal patch - have you had chance yet?
10:27:17*amiconn thinks that we need a function that tells whether an *LCD* remote is connected - and that function should also be available to plugins
10:27:47amiconnRight now we only have a function that tells whether a remote is connected - whether LCD or simple
10:28:02amiconnOnly the remote LCD driver knows whether its LCD is there
10:28:51pondlifeI'd hope that individual plugins shouldn't need to care about such detail, it should be handled lower down...
10:29:03amiconn(rather, it assumes based on the remote type - there is no way to tell whether it's actually there)
10:29:42jhMikeSlinuxstb: I suppose that's ok for now. I guess any COP stuff can be dealt with at the time.
10:30:00pondlifeWhat if the user unplugs or plugs in the remote whilst something is running? Should the plugin check before each update?
10:30:04amiconnpondlife: Plugins don't necessarily _need_ to care - but they might _want_ to act different when an LCD remote is connected
10:30:26jhMikeStest_fps? as a small example.
10:30:37amiconnYou can call lcd_remote_update() anytime - that's not a problem
10:31:14jhMikeSI guess it shouldn't run remote tests is the remote wasn't plugged
10:31:19amiconnBut e.g. a game could offer a true dual player mode. It could hide that option when there is no remote lcd
10:31:21 Join Diamond_iceman20 [0] (i=c35d1562@gateway/web/cgi-irc/
10:31:37amiconnWhy bother the user with a mode he can't use because he doesn't have a remote?
10:31:44linuxstbjhMikeS: So you're saying my patch is good enough to commit?
10:32:05jhMikeSwith the state check I don't see a problem atm
10:32:35Diamond_iceman20I think the Rockbox f/w sounds better than the Apple f/w on the iPod
10:32:41jhMikeSgood thing is that you won't see STATE_BLOCKED if it still has messages to process
10:33:22amiconntest_codec should probably do the same then?
10:33:44jhMikeSI think so
10:34:22jhMikeSsometimes it does mess up playback afterwards, I suspect that's what would cause it
10:35:02Diamond_iceman20when you go into record mode on the ipod, the ipod crasheds
10:37:09linuxstbDiamond_iceman20: It shouldn't crash, but playback is known to not work after recording.
10:37:47 Quit Diamond_iceman20 ("CGI:IRC")
10:38:43 Join homielowe [0] (
10:39:06 Join Diamond_iceman20 [0] (i=c35d150a@gateway/web/cgi-irc/
10:39:14Diamond_iceman20yea thats what i mean
10:39:27Diamond_iceman20can i plug a microphone into the headphone socket for working?
10:39:49 Join spiorf [0] (
10:40:02linuxstbDiamond_iceman20: Only on the 4G Greyscale and Color ipods
10:40:07Diamond_iceman20oh right
10:40:15Diamond_iceman20i have an ipod 5g 60g
10:40:46jhMikeSiPods still do that stuff?
10:40:58Diamond_iceman20so how could i record on a 5g Video 60g?
10:41:48linuxstbjhMikeS: Yes, at least when I last tested (about a week ago)
10:43:31linuxstbamiconn, jhMikeS: Patch updated with the state check -
10:43:56jhMikeSOn H10, not using r12 fixed it. I wonder what gcc is up to.
10:44:20linuxstbWhen was that committed?
10:44:36jhMikeSand re-fixed for e200 since changing pcm causes an unrestored r12 use
10:44:46jhMikeSwith the pcm changes
10:45:27linuxstbI've just tried again, and playback still freezes after recording...
10:45:34 Join barrywardell [0] (
10:45:35jhMikeShard lock?
10:45:52Diamond_iceman20Im pleased that the EQ now works without skipping playback no backlight timeout seen
10:46:02linuxstbDiamond_iceman20: You need to build yourself a cable - some instructions are in the ipodlinux wiki.
10:46:09Diamond_iceman20oh right ok
10:46:21linuxstbjhMikeS: Yes, the backlight thread is still running...
10:46:40linuxstbjhMikeS: In fact, it's not a freeze at all, just that playback doesn't work...
10:47:04jhMikeSit never resets the headphone socket from what I saw
10:47:33linuxstbThe WPS is just stuck - the audio debug screen shows the PCM buffer is full and never emptied.
10:47:52 Quit Diamond_iceman20 ("CGI:IRC")
10:47:57linuxstbjhMikeS: OK, let me test line-in
10:48:59jhMikeSwhich wm codec does it use?
10:49:37linuxstbjhMikeS: Ah, that could be it... Playback is working after recording from line-in... I wonder why it's freezing Diamond_iceman20's 5g though.
10:50:09ddaltonWhat action is activated when down is pressed in the file browser?
10:50:39linuxstbNo, ignore that, playback is broken - the wps is moving but I'm just getting silence..
10:50:56jhMikeSyou restored the socket state?
10:51:10linuxstbjhMikeS: My Color is wm8975 IIRC. firmware/SOURCES should indicate
10:51:25linuxstbjhMikeS: No, I used line-in recording, which shouldn't change the socket state.
10:51:29jhMikeSThought I'd get a quicker answer just asking :)
10:51:31ddaltonjhMikeS: do you know what action is activated when down is pressed in the file tree?
10:51:47jhMikeSddalton: depends on the keymap.
10:52:53jhMikeSmaybe it's just ACTION_STD_NEXT
10:53:45ddaltonjhMikeS: in tree.c?
10:54:22*amiconn put some instrumentation in the C montion compensation
10:54:46amiconnNow it should print call stats to all 16 motion compensation functions when exiting mpegplayer in teh sim
10:56:11jhMikeSddalton: it calls gui_synclist_get_sel_pos
10:56:43ddaltonjhMikeS: ok
10:56:49linuxstbamiconn: mpegplayer patch committed - have fun ;)
10:57:10ddaltonso how would I call a function that only executes after a .talk clip has played or a file has finished been spoken out.
10:57:22ddaltonlike spelt or numbered as well
10:57:35amiconnYou can't - voice is asynchronous by design
10:57:42safetydanlinuxstb, just finished syncing it then
10:58:14jhMikeSit could send events I imagine if that were added
10:58:22amiconnlinuxstb: I just got an mpegplayer crash on X5 (with your first patch) ...
10:58:31amiconnIllInstr at 0x00000002
10:58:46amiconnI guess it's not directly related to your patch
10:59:03linuxstbamiconn: Hmm...
10:59:11*jhMikeS tries to track down where it handles directional keys
10:59:15amiconnI did something I never tried before - let mpegplayer play until the end with frame skip and frame limiting disabled
10:59:39amiconnIt's a bit funny if it reaches the end of the video, but audio is still playing...
11:00:12linuxstbAh, then I would be surprised if we can blame my patch for that.
11:01:15jhMikeSif one stream overtakes another, it will not properly block them but basically abort the stream
11:01:24amiconnMaybe mpegplayer doesn't handle some end-of-stream condition properly?
11:01:41linuxstbsafetydan: Thanks - I'll try and test it this evening (EU time).
11:02:04jhMikeSit runs past all available data right back up to the head of the buffer
11:02:27safetydanlinuxstb, haven't actually got it to compile yet... something's gone screwy
11:02:37amiconnN6w 3et the stats r4n,,,
11:03:06amiconn*Now let the stats run...
11:03:11amiconnSorry for numlock ;)
11:03:12jhMikeS0# 5617
11:05:44amiconnHmm, the sim threading seems to act up after a while in mpegplayer when frame limiting is disabled
11:05:46jhMikeSddalton: I'm sure JdGordon would know where the heck it's handling that stuff. I can't seem to track it down. :\
11:05:59jhMikeSact up?
11:06:19amiconnAfter a while, the video begins to stop periodically, then continue after a few seconds
11:06:29amiconnThat's on windows
11:07:03amiconnCPU load stays at 100%
11:07:34amiconnI guess the background disk reading can't keep up
11:07:45ddaltonneither can I
11:08:05 Join pill [0] (
11:08:20amiconnCompile a windows sim. Load a video, disable frame skip and framerate limiting, and let it play...
11:08:54amiconnMaybe it also happens on linux
11:09:07ddaltonis there a function I can call to start recording to a file and one to stop recording?
11:09:21ddaltonand exit the rec screen?
11:09:23jhMikeSIf it were hitting the end of data, the parser would just abort it. It might be the opposite - disk reading is dominating the video.
11:09:58amiconnIt doesn't happen if framerate limiting is on
11:10:48amiconnOr maybe it's just due to audio & video being too far off?
11:12:35***Saving seen data "./dancer.seen"
11:13:22jhMikeSit could be along with the far apart mem access, sure. could cause pagefile hits.
11:13:54amiconnelephants dream uses halfpixel motion, but no destination averaging at all
11:14:49jhMikeSit has no B-frames encoded either which help keep quality up and size down...besides the dropping advantage
11:15:30amiconnI thought using B frames makes for smaller files...
11:16:12jhMikeSor same-size files with higher quality :)
11:17:39jhMikeSlow-power devices get smoother dropping with them too since dropping P means you can't draw anything again until another I. B frames are unreferenced.
11:17:50Zagorhmm, how do I find out if I'm being called in interrupt or thread context?
11:18:24jhMikeScheck cpsr?
11:18:41Zagorthat doesn't sound terribly cross-platform?
11:19:12Zagorlet me rephrase: do we have some method for finding out the current context?
11:19:18jhMikeSI guess implement an is_interrupt_context or something. That's really the only way to know.
11:21:11jhMikeSWhy do you need to know?
11:21:52Zagorwell I'm thinking about trying some delays before sending, but I assume I'm not supposed to call yield() from interrupt code?
11:22:23jhMikeSdefinitely not
11:22:43Zagorso I need to know when it's safe to use yield and when I have to loop
11:23:03amiconnImo you could just use udelay for experimentation
11:23:30 Quit idnar (Nick collision from services.)
11:23:32 Join idnar_ [0] (i=mithrand@unaffiliated/idnar)
11:23:56Zagoramiconn: yeah, I guess I'll do that. but I was curious :)
11:24:00jhMikeSAny threading object could be wrapped interrupt-safe on the waking end. wouldn't take much if object troubles have anything to do with it.
11:25:14 Join moos [0] (
11:28:22 Join PaulJam [0] (
11:31:25amiconnHere's the elephants dream stats (160x128):
11:32:34amiconnI would expect that if the video uses B frames, we would see calls to the avg variants...
11:34:44*jhMikeS would now use semaphores and not generic message queues for transferring data in
11:37:21jhMikeSamiconn: I tried a recommended program to encode ED with B-frames and the sync is way off. Not sure why it would mess that up.
11:37:47jhMikeSIt plays though
11:39:41preglowamiconn: am now
11:40:00pixelmalinuxstb: your last commit caused one additional warning in the sims
11:40:38amiconnIn case someone is interested in the patch to gather those stats:">
11:40:42amiconnIt's fairly simple
11:40:49linuxstbpixelma: Thanks, I'll fix.
11:40:52 Join sup [0] (i=1000@
11:41:04amiconnIt just tacks on a counter array to those functions
11:41:17ddaltonwhat function should I use to go to the rec screen?
11:41:21preglowamiconn: interesting
11:41:22ddaltonIn recording.c
11:41:36amiconn(only the C version - mpegplayer won't compile for target with that patch)
11:41:45preglowwhat encoder is used for elephant's dream?
11:42:35linuxstbmpeg2enc - from the mjpegtools project
11:42:49preglowit doesn't use the averaging ones at all
11:42:56preglowi wonder if that's a trend we'll see in other encoders as well
11:42:57jhMikeSthat's what gave me a bad transcoding with B-frames :)
11:43:57jhMikeSjust because they don't want to take time to implement certain things? eck.
11:44:23preglowit's usually for speed
11:44:36ddaltonjhMikeS: do you know how to get to the rec screen? what function?
11:44:40amiconnpreglow: Perhaps the averaging would be used for B frames?
11:44:42preglowoptimizing motion compensation not to be shit slow is a hard task
11:44:57linuxstbmpeg2enc is generally optimised for quality rather than speed though...
11:45:06jhMikeSddalton: anywhere rec_screen is called
11:45:18preglowamiconn: you said the avg routines did the same as the other ones, but just averaged the block with the already existing data?
11:45:38preglowamiconn: then i don't think so, b frames interpolate motion vectors, not data, afaik
11:46:09amiconnThey need to modify the output framebuffer(s) at some point...
11:46:21preglowsure, using interpolated motion vectors
11:46:38preglowi really don't know, i only reliably remember stuff about i and p frames
11:46:39jhMikeSddalton: excuse me, it's recording_screen
11:46:46preglowfor some reason i consistently forget what i'v read about b frames
11:46:52ddaltonso I just write recording_screen ();
11:47:10ddaltonNow ignoring pixelma -BOFHNet IRC Ignore-
11:47:16preglowbut ok, time to commit the other asm stuff so wma can be fasted on coldfire than arm :>
11:47:25jhMikeSddalton: it has a paramter to allow source changes as well: bool recording_screen(bool no_source)
11:47:46pixelmaddalton: ?
11:47:51ddaltonok thanks for that
11:47:52safetydanpreglow, linuxstb, just posted a basic resync of the hw eq patch
11:48:32jhMikeSddalton: It's used from the FM screen to hide the other sources
11:49:11ddaltonSo I give either true or false?
11:49:15preglowsafetydan: most excellent
11:49:34ddaltonso what does true do?
11:49:51jhMikeSdisables and hides the audio source selection
11:50:33jhMikeSit then needs to be set in advance before calling the screen
11:51:21ddaltonso I just can't call recording_screen (false); from my function?
11:52:09 Quit psycho_maniac (" HydraIRC -> <- The alternative IRC client")
11:53:43 Join Thundercloud [0] (
11:55:00jhMikeSddalton: if it's just to take it there using the global setting, then you should use false
11:55:13 Join lee-qid [0] (
11:56:11*jhMikeS thinks bool recording_screen(int source) would be better where SOURCE_DEFAULT uses the global setting and allows selection. maybe a change for later.
11:56:30safetydanhrm, now should cutoff be a choice setting or an int setting?
11:57:45ddaltonok I will try this.
11:57:48ddaltonthanks for your help
11:59:01preglowsafetydan: choice as in strings?
11:59:32preglowsafetydan: if we use pure ints, it'll be hard to say which is the higher cutoff setting for some users, perhaps
12:01:52safetydanpreglow, well strings for the user anyway, something like "lowest low high highest"
12:01:54jhMikeSddalton: I was just speculating on a nicer was to do something. The last thing I said won't work.
12:02:02preglowsafetydan: i think we decided against that
12:02:24preglowbut it's not going to be much different anyway
12:02:39safetydanpreglow, so just straight 0 to 3 (1..4) then?
12:02:56preglowsafetydan: do that for now, and i'll ask around to see what people think
12:03:01jhMikeS"good better best bestest" :P
12:03:06preglowsomething is weird on nano now
12:03:45preglowsometimes i get data abort/illinstr and pink spots in the wps
12:03:53jhMikeSpreglow: what did I do now? :P
12:04:29jhMikeSwasn't nano misbehaving when warm?
12:04:30preglowmight be me, but i think these asm snips are correct
12:04:40preglowjhMikeS: yeah, but mine isn't, i think
12:04:43preglowperhaps that's it, though
12:05:01jhMikeSchill it in the fridge and see
12:05:08preglowwell, it doesn't do it reliably
12:05:48jhMikeSpink spots doesn't sound like anything I've worked on would cause. You'd either crash or not.
12:05:55LloreanWell, pink spots in the WPS have never been reported as part of the warm-Nano problem
12:06:33LloreanIt pretty much restricts itself to "garbled playback" and "data aborts and illegal instructions at completely random addresses"
12:06:52*linuxstb points committers to their mailing list and asks for help
12:07:20ddaltonjhMikeS: It compiled. now the question is: "Will it work?"
12:07:20preglowthe pink spots are transparent sections, i'm quite sure
12:07:26preglowbut god knows why the colour key is ignored
12:07:44preglowlinuxstb: have you got any wma files?
12:07:46pondlifeamiconn: Did you see the Nano offer at ?
12:07:55supis there any statistics on which mp3 player is most popular with rockbox?
12:08:08jhMikeSddalton: I guess that's why one tests things. :)
12:08:29linuxstbpreglow: A few, why?
12:08:38jhMikeSvideo, right?
12:08:55preglowlinuxstb: would be cool if you could check out if they sounds same afterwards
12:09:20linuxstbYou mean if they still sound like a bag of shit?
12:09:28preglowyeah, mainly :D
12:09:55Lloreansup: Most popular as in "The most people have bought this player" or Most popular as in "the people who already use Rockbox mostly pick this player when they need a new one"?
12:09:59preglowlinuxstb: most helpful would be checking coldfire, i guess
12:10:19preglowlinuxstb: boost ratio goes from 175% to 255% on the bangles track on coldfire since yesterday
12:10:25linuxstbpreglow: OK, I'll add it to my to-do list for this evening.
12:10:31preglowlinuxstb: sweet
12:10:37ddaltonwell. Doesn't somone have a download log on his website.
12:10:37preglowlinuxstb: realtime ratio, i mean
12:10:56ddaltonwich says how many downloads each player has had.
12:11:27linuxstbddalton: Bagder mentioned some stats a while ago - the ipod video was by far the most downloaded IIRC.
12:11:42supLlorean: well, besically, what version is the most downloaded..
12:12:20ddaltonyes and I would not reccomend it.
12:12:22linuxstbsup: The ipod video is one of the worse Rockbox targets though...
12:12:30Lloreansup: iPod Video then. Apple sold so many thousands it really couldn't *not* be #1 very easily
12:12:56supnot the nano?
12:13:26preglowargh, oops
12:14:10pixelmaand download stats don't tell everything, people might compile their own builds or update in different intervals
12:16:01pondlifeThe stats on rockbox-themes indicate a large iPod Video presence, although this probably indicates more that iPod users want eye candy more
12:16:33ddaltonjhMikeS: I think it is working but it doesn't work yet. There is no code to write to the file.
12:16:48preglowi believe we do have lots of downloads of video on too
12:16:59 Join mrkiko [0] (
12:17:05ddaltondo you think this is a good idea? A quick memo type feature. Hold rec down and record. Once you are done release it and it saves your recording?
12:17:05mrkikoHi all!
12:17:07amiconnpreglow: loadsa RED...
12:17:13 Quit spiorf (Remote closed the connection)
12:17:16ddaltonso it is only recording when you are holding rec down
12:17:42pondlifeddalton: Anything that uses buttons is probably not a good idea ;)
12:17:43mrkikoIs someone using an iRiver340 with rockbox?
12:17:50mrkikoDoes it work well?
12:17:51pondlifeOr rather, changes their use.
12:17:51linuxstbAlso, the ipod video is (was?) the only hard-disk DAP easily available to run Rockbox on.
12:17:55amiconnAh, new build started...
12:18:03pondlifemrkiko: H340 works great
12:18:14jhMikeSddalton: with recording, the disk spinup time to load an encoder would need to be waited on.
12:19:02ddaltonyes but I would still probably use this feature. so not to worried if it is committed or not. But I don't see much use for the long rec button at the minute. It just goes to the rec screen. And I have pressed it by mistake and ended up there.
12:19:09ddaltonWhen resetting my settings.
12:19:28mrkikopondlife: I live in Italy but my iRiver has an US firmware evidently. If I want, can I reinstall the US firmware and have the DRM support back? I hate DRM but I was curious to know if DRM is an hardware or software chip
12:19:41ddaltonVery annoying when speakers are attached to the player. I think this should be removed from rb. But that is just my opinion.
12:19:46pondlifeNo, once removed, you have lost DRM forever AFAIK
12:19:57supso when is rockbox fr g2 nano coming out.. : ]
12:20:14pondlifeWhen you crack the encryption and port it :)
12:20:17ddaltonits not until it is done
12:20:26ddaltonand a port first needs to be started.
12:20:31ddalton(is one?)
12:20:38pondlifeNever, is the best guess
12:20:54jhMikeSthe thing that makes it hard to read something
12:21:02pondlifeOr run something
12:21:07supwhat is encrypted? : ]
12:21:20jhMikeSrunning requires it be read initially :)
12:21:52mrkikoI'm also blind - is in your opinion reasonable to thrust in rockbox speaking?
12:22:11pondlifesup: The later iPods don't have any way to run proprietary code. So unless someone finds a weakness there, or Apple tell us, there is no chance of a port even starting.
12:22:16mrkikoAnd hence - I red on the manual that plugins doesn't support speech. So I will not be able to read txt files, righ?
12:22:16*sup thrusts
12:22:29jhMikeSmrkiko: I think you can offer your opinion here
12:22:45markunmrkiko: there is a text to speech plugin in development, but there are some license problems
12:22:49linuxstbmrkiko: At the moment, yes. But realtime text-to-speech is a very wanted feature.
12:22:56 Quit FOAD ("I'll be back")
12:22:57supwhat processor does the 2g nano use?
12:23:11linuxstbAn unknown Samsung chip with an ARM core.
12:23:37pondlifesup: Most of the 2g hardware is different from the 1g, and even more mysterious.
12:23:40mrkikothe problem is that I can't let my iRiver crash - it's a present also :)
12:23:58markunmrkiko: if it crashes you can just reset
12:24:34mrkiko... "reset" ..
12:24:55pondlifemrkiko: The only risk is when flashing the bootloader. I've done that 100s of times with no brickage; you'd only need to do it once.
12:24:55markunwhich iriver do you have?
12:25:24markunah, h340
12:26:04mrkikoOk ... when I find a sighted user I will ask for help and installa rockbox
12:27:13supmrkiko: are you really blind? : D
12:27:54*pondlife hopes to avoid a subscription to
12:27:59markunsup: you thought it was a joke?
12:28:35mrkikosup: yes...
12:28:58mrkikofrom when I was born
12:29:03*jhMikeS wonders how a PP chip was ever made run 3rd party FW since "PortalPlayer" should be not much more descriptive than an Apple logo.
12:29:05mrkikobut it's not so bad
12:29:09supwell, i just think it sounds hard setting up irssi to read the text in linux..
12:29:31supoh, so youre not Totally blind?
12:29:36linuxstbjhMikeS: You mean how did IPL do that initial work?
12:29:44jhMikeSlunuxstb: yes
12:29:46 Quit barrywardell (Connection timed out)
12:30:27linuxstbjhMikeS: There is a very high-level product brief available that describes what the chip does. And the Apple original firmware isn't encrypted... I don't think it's impossible...
12:30:31mrkikothe fact is that I'm using a braille device
12:30:37ddaltonHow do I use the get_action function to detect ACTION_STD_REC?
12:30:38mrkikoso I don't need irssi to use the festival plug-in
12:31:25safetydanall this effort to make wma faster is just going to tempt people to use that format :)
12:31:39markunmrkiko: do you know espeak?
12:32:01mrkikoI used it some time ago
12:32:06mrkikoAniway: I'm at school now
12:32:10mrkikoso I should just escape
12:32:22markunok, take care
12:32:26mrkikoOk, I'll look at that URL
12:33:00safetydanpreglow, the patch now has numeric settings and the settings should be usable for any target with configurable bass/treble cutoff
12:33:13ddaltonmrkiko: Are you blind like me? :-)
12:33:17mrkikoEhi boys: you are the first really helpful and nice irc channel i visit
12:33:27mrkikoddalton yes, and I saw your site
12:33:35mrkikoYou're a partecipant to the rltty mailing list
12:33:42 Join FOAD [0] (
12:33:50ddaltonmrkiko: What site? my web one? or a task I opened
12:34:17ddaltonoh right. I see...
12:34:34 Quit Soap ()
12:34:35mrkikoyour web site
12:34:44mrkikothe one your append to your messages on the brltty mailing list
12:34:48ddaltonmrkiko: Have you seen some of sdoyon's patches?
12:35:10jhMikeSlinuxstb: encryption was a good start
12:35:10mrkikosdoyon's patches? I don't know of them.
12:35:29ddaltondo you build your own rockbox from source?
12:36:52ddaltontry this link. Hopefully it works.[]=
12:37:34 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
12:42:12mrkikoOh... I don't know how to proceed: what advantages will I have building my own rockbox?
12:42:46 Join barrywardell [0] (
12:42:53 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
12:48:02linuxstbmrkiko: If you can program, you can change Rockbox's behaviour to suit your own personal needs. Or you can apply changes (called patches) made by other people, but not incorporated in the official version.
12:49:16ddaltonsome good patches for blind people are at the link I said before. so if you can build rb from source you can apply those patches and build.
12:49:18mrkikoOh, I know. I'm able to program in C
12:49:21mrkikodiscretely well...
12:49:37mrkikoAnd I know the patches concept... I don't know those mentioned by ddalton
12:49:42ddaltonSo for example if you applied his talking playlist viewer patch you could then use the playlist viewer with speech.
12:49:52ddaltonthis isn't possible in the normal version of rb
12:50:03mrkikogood! :)
12:50:30mrkikoWhat parts of rockbox are accessible without patches? (i.e.: the fm radio tuner is accessible? )
12:50:43ddaltonwell sort of...
12:51:06ddaltonit doesn't voice the freq, but it voices menus and stuff.
12:52:02ddaltonmrkiko: well these just give you improvements lots is accessible. But lots isn't. Also you can other patches. Like maybe one that changes the playback or something.
12:52:08ddaltonuse lots
12:52:30ddaltonAlso I have written a couple but not sure if you will be interested.
12:53:01preglowsafetydan: excellent
12:54:01pondlifeNico_P: Morning
12:54:22pondlifeNico_P: Did you rebase your git recently?
12:54:35Nico_Pto include jhMikeS' changes
12:55:04pondlifeOK, so still in-sync-ish..
12:55:08Nico_Pit was 37 hours ago ;)
12:55:12Nico_Pyes I think so
12:55:43Nico_Pgenerally mob will just be on top of master so you can see the last svn commit that is included
12:56:04Nico_Pr15154 currently
12:56:12ddaltonjhMikeS: Do you know how I use the get_action funtion?
12:56:22ddaltonwhat arguments do I give it?
12:58:31pondlifeNico_P: I've been running MoB on the sim lots recently, seems good there. My H340 isn't with me at the moment though.
12:58:31mrkikoThe fact the iRiver boot loader is so closed-source is sad...
12:58:37Nico_Phaha is getting updated every day with better figures :)
12:59:03jhMikeSddalton: it takes a context and a timeout and returns the action
12:59:10Nico_Ppondlife: cool :) have any of the bugs from MetadataOnBufferTesting disappeared ?
12:59:19ddaltonok will " button = button_get(true);"?
12:59:23pondlifeNot sure, need to retest on target
12:59:40Nico_PI haven't done much progress lately due to my being ill
12:59:55Nico_Pmaybe today I'll have courage to work oni int
12:59:56pondlifeNo problem, concentrate on getting well
13:00:07jhMikeSddalton: that just returns raw buttons
13:00:24Nico_Pdon't worry it's just a small flu :)
13:00:52pondlifeNot playback.c-itis ?
13:01:01Nico_Pmaybe that too
13:01:08ddaltonjhMikeS: ok.
13:01:10jhMikeSflus are relaxing since I don't care about anything but getting better :)
13:02:41JdGordonlol pondlife
13:02:58Nico_Ppondlife: can you still repro the segfault on UP + SELECT in the sim ? and the others ?
13:03:17jhMikeSsince "-itis" refers to "swelling-of"...I'd hope it's the opposite
13:04:32jhMikeSNico_P: Was it really nasty to resync?
13:05:09Nico_PjhMikeS: a little harder than usual, especially as I had to edit commits back in history so that they would all compile
13:05:31Nico_P(to make bisecting possible)
13:06:10pondlifeNico_P: The sim segfault was fixed by your callback-with-no-ID3 mod.
13:06:14Nico_PjhMikeS: apps/playback.c: 390 insertions(+), 407 deletions(-)
13:06:21pondlifeI'm not convinced that was the right fix still though
13:06:42pondlifeI still get stalled playback if I repeat UP+SELECT enough though
13:06:44pondlife(In the sim)
13:06:45Nico_Pyeah I need to really understand what that loop does
13:06:52 Quit barrywardell (Read error: 110 (Connection timed out))
13:06:54jhMikeSsorry for that. I wanted to get the disrupting over with.
13:07:09Nico_PI've noticed that too, but at least you can recover by simply restarting playback
13:07:21pondlifeBasically, why do an unbuffer callback for a track that was never buffered?
13:07:41pondlifeIt shouldn't even have been considering that..
13:07:55*ddalton Wrote his first patch that chrashed rockbox
13:08:12*pondlife has lots of those in his cupboard
13:08:18Nico_Ppondlife: I agree but I think the difference is mainly in how track_widx and track_ridx are used
13:08:48*ddalton was to do with recording :-)
13:09:18pondlifeNico_P: I was hoping the circular use of those could be scrapped some day... hint.
13:09:20Nico_PjhMikeS: also you have to know that some additions in playback.c are #if 0 and #endif around big blocks of code ;)
13:09:32Nico_Ppondlife: how ?
13:09:33pondlifei.e. let the callbacks be dealt with by buffering.c
13:10:17Nico_Ppondlife: callbacks in buffering.c is a good idea but I don't see how it changes anything for track_?idx
13:10:33pondlifeProbably doesn't. I'm being random
13:10:37mrkikoI should escape miseducately
13:10:40 Quit mrkiko ("leaving")
13:10:46pondlifeProbably getting the flu
13:12:39***Saving seen data "./dancer.seen"
13:13:33ddaltonjhMikeS: Do you know what variable to speak to find out the elapsed time in the wps? (In the stop seek case)
13:15:03jhMikeSddalton: I've hardly ever opened any WPS c files. UI in general isn't something I've done much work on.
13:15:19 Quit safetydan ("Leaving")
13:15:23ddaltonok then. Thanks for all your help though
13:15:55Nico_Pddalton: you want to know where the elapsed time is ? id3->elapsed
13:16:31ddaltonNico_P: thanks is that in seconds?
13:17:06ddaltonoh ok that's why it wasn't working last time. ms is hz/2?
13:17:28Nico_PI don't know, what's HZ ?
13:18:07ddaltondoesn't matter so how do I format it in to hours/minutes/seconds?
13:18:27ddaltonI think you use "HZ" to get one second for example would be HZ*1
13:18:49peturthe tick is 10ms, HZ = 100 so HZ ticks gives 1 second
13:19:08Nico_Pddalton: then 1 ms is HZ/1000
13:20:32 Join midgey [0] (
13:20:52Nico_Pddalton: format_time in apps/misc.c
13:22:06ddaltonNico_P: but for voice?
13:22:06*JdGordon has a cursor following the stylus across the lcd :D
13:22:18ddaltonso how do I get it in seconds?
13:23:50Nico_Pddalton: to get it in seconds, simply multiply the value by 1000... 1ms is a thousandth of a second
13:24:08ddaltonoh that's what google told me as well. :-)
13:24:16Nico_PJdGordon: a cursor in rockbox ??
13:24:25JdGordonit wont last long
13:24:26ddaltonso hz* id3->elapsed_time or what ever it is called?
13:24:39JdGordonbut i needed to make sure it was getting the right pixel values from the touchpad
13:25:17*Nico_P is havins some freezes in mpegplayer
13:26:37Nico_Pddalton: safest is 1000 * id3->elapsed
13:26:47Nico_Punless you need HZ to be in there
13:26:56ddaltonok thanks I will do that
13:26:57amiconnDivide, not multiply...
13:27:04Nico_Perr yeah
13:27:15amiconnunless you want it in microseconds :P
13:29:28JdGordonjhMikeS: does button.c know anything about the sansa wheel delta?
13:29:58JdGordonI need to get some data from button_read_device() to button.c and not sure how to do it
13:30:20jhMikeSJdGordon: Nope, it just passes it along. it has the function to use it but there's no real integration
13:30:39amiconnThere must be some debug code in the H10 OF...
13:30:47jhMikeSpost it to the queue data
13:30:49*amiconn just ran 'strings' on it
13:31:16jhMikeSamiconn: I believe I mentioned that it's full of them :)
13:31:28amiconnThere are several groups of strings which look like some kind of debug screens
13:31:29jhMikeSI got some reg names from it
13:31:33JdGordonjhMikeS: post it where? when the button happens or from inside button.c?
13:31:53jhMikeSpost it to the queue
13:32:28jhMikeSthe wheel driver just posts the data along with the BUTTON_SCROLL_UP/DOWN message
13:32:52JdGordonoh, righto
13:33:05amiconnhehe, and a couple typos...
13:33:40jhMikeSJdGordon: though button read device doesn't seem to have any facility for extra data :\
13:33:51 Quit BigBambi (Read error: 113 (No route to host))
13:33:54Nico_Panyone know why mpegplayer freezes towards the end of elephant's dream ?
13:34:19JdGordonwe may have to think of a better way to do buttons... but for now ill just copy the wheel driver and implement button_repeat manually
13:34:20jhMikeSNico_P: no matter if you start at the beginning or not?
13:34:45Nico_Pah, haven't tried not starting from the beginning
13:35:33jhMikeSI just get "missing packet start code prefix: nn"
13:35:58amiconnRTXC3.2 RTOS
13:36:29amiconnAnd it looks like the OF might have USB host
13:36:32jhMikeSlots of UTF-16 in there too
13:36:32JdGordonjhMikeS: its fine to do the post in an IRQ right?
13:36:46jhMikeSqueue_post is completely IRQ safe
13:36:57JdGordoncool, just chcking :)
13:37:04preglowcan't queue_post block?
13:37:34jhMikeSqueue_send therefore it's forbidden by IRQ handlers
13:37:54ddaltonhow do we convert seconds to minutes? for example 3700 secons. 1 hours and how many minutes?
13:38:00 Quit midgey ()
13:38:12amiconnHaha, numerous filenames including path.
13:38:14ddaltonI know it is 1 hour and one min but how do we do this in c what operation?
13:38:20amiconnThat thing was compiled on a windows box
13:40:24preglowddalton: minutes = totalsecs/60, seconds = totalsecs%60
13:40:55jhMikeSamiconn: just wait till you get to realizing there COM objects in there too :)
13:41:22jhMikeSof course that needs a little disasm as well
13:41:37amiconnJanus drm... I don't think MS has a devkit to build this on something else than windows
13:41:37ddaltonok I will try that thanks.
13:41:43*ddalton Is no good at maths
13:42:38 Join MethoS- [0] (
13:44:30jhMikeSdon't we have utility functions to just decompose thing into time units? /me thinks we should.
13:45:11JdGordonjhMikeS: putting the quque_post() in the irq instead of button_tick seems to have made it stop working (BOOTLOADER is defined), any ideas why?
13:45:47jhMikeSJdGordon: what are you doing? there's no interrupt in the bootloader for e200. the buttons are polled there.
13:45:57JdGordonmrobe touchpad
13:46:18JdGordonsorry, should have said that
13:46:29JdGordoni was only interested in the e200 because of the delta stuff
13:46:39 Quit SirFunk (Remote closed the connection)
13:46:39JdGordonwhich is sort of similar but not quite :p;
13:46:50jhMikeSAll the wheel support is disabled in the bl
13:47:48JdGordonnot the wheel stuff... button_get_data() should work in the bootloader right?
13:48:49 Quit EnterUserName (Read error: 104 (Connection reset by peer))
13:49:13 Join EnterUserName [0] (
13:50:10JdGordonbah, button_status was being called instead of button_get for some reason
13:51:30 Quit iamben (Read error: 104 (Connection reset by peer))
13:52:49JdGordonyay :) its working again
13:53:40 Join cendres [0] (n=ashes@2001:5c0:8fff:ffff:0:0:0:79)
13:55:18 Quit Zagor ("Client exiting")
13:55:29jhMikeSbutton_get_data will work as long as the queue used, sure
13:55:34jhMikeS*is used
13:57:55jhMikeSpersonally I'd prefer button_get be able to return the data right during the call instead of stashing it in a global static
13:58:54 Join Zagor [0] (n=bjorn@rockbox/developer/Zagor)
14:01:51JdGordonI agree
14:03:46JdGordonhmm... we dont have a queue_peek or similar?
14:05:26 Quit ashes (Read error: 113 (No route to host))
14:05:38jhMikeSto peek for what?
14:06:13 Join Arathis [0] (
14:06:40jhMikeSqueue_empty is true if there's no messages at the time, queue_count gives the count at the time
14:08:28 Join mf0102 [0] (n=michi@
14:09:00 Quit ddalton ("leaving")
14:09:09 Join ddalton [0] (
14:09:55 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
14:10:10 Quit ddalton (Client Quit)
14:15:54*linuxstb wonders why bluebrother dismissed so quickly
14:16:19JdGordon"The device can be mounted manually as root."
14:16:28JdGordonits a linux problem.. not rockbox
14:16:53linuxstbIIUC, the poster is describing different behaviour under Rockbox compared to the original firmware.
14:17:17Lloreanlinuxstb: On a device where that really shouldn't be possible though. Hardware USB.
14:17:44linuxstbRockbox still does _something_ though.
14:18:01preglowseems ffmpeg survives some errors in wma streams
14:18:08pregloweven if decode_block itself errors out
14:21:58JdGordonSo, im going to start getting the touchpad working in rockbox, we dont want that compiled in for all sims so people can use the mouse right?
14:22:43Nico_PI'd like to be able to use the mouse in all sims
14:22:49Nico_Pmabe make it a command line arg
14:22:55 Join n1s [0] (
14:23:15JdGordoncommand line is a waste i tihnk.. its not like the backdrop where it will annoy people who dont use it
14:23:32JdGordonI tihnk there was some objection to it last tim I brought it up though
14:23:45Lloreanlinuxstb: My H120 has no problem automounting in Ubuntu in Rockbox.
14:23:49pondlifeThe objection is that the sim should be as close to the target as posible.
14:24:17pondlifeI too would like to have the option of using the mouse though!
14:25:16pondlifeJdGordon: How will the touchscreen/mouse support handle going back/cancelling etc. An on-screen button I suppose?
14:25:42JdGordontouching different regions gives different acitons
14:25:55JdGordoniirc top left is always the "menu" button
14:26:59JdGordoncompiled and on the dap.. here goes nothing! :D
14:27:16linuxstbLlorean: OK. But it's odd that the poster is reporting different behaviour in the OF and Rockbox - assuming he/she really is.
14:27:29JdGordonoh poo!
14:28:07Lloreanlinuxstb: Yeah, I would've asked about Bootloader USB, as well as rebooting the computer.
14:28:27LloreanIt could be that the automount function decided to give up after his OF test and before his Rockbox test
14:32:32amiconnAfaik some people have trouble with bootloader USB on iriver
14:32:40*preglow gives up finding out how ffmpeg handles this
14:33:37Lloreanamiconn: I thought all the bootloader USB problems were on H300
14:33:46amiconnI'm not sure
14:34:03LloreanI pretty much use bootloader USB exclusively on my H120, except when I need to test something, and I've not had a problem yet.
14:34:03 Quit Nico_P (Read error: 104 (Connection reset by peer))
14:34:09LloreanThough I'll admit I don't use it nearly as much any more as I used to
14:34:19amiconnIt doesn't affect all H300s either. Or perhaps it depends on the host
14:34:41amiconnBootloader USB works fine for me, both on H1x0 and H300
14:34:41 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
14:34:59amiconnThat is with latest official bootloaders
14:36:00pondlifeLatest official H300 bootloader never works for me
14:36:04pondlifeSVN is fine though
14:36:35amiconnI never tried plain SVN; too dangerous on irivers
14:36:41pondlifeI'll bug LinusN when he gets back from China about a new release... :)
14:37:06preglowreturn AVPROBE_SCORE_MAX/2+1; // this must be less than mpeg-ps because some retards put id3v2 tags before mpeg-ps files
14:37:18preglowsounds like they have experiences like ours wrt. tagging :P
14:37:23amiconnBut iirc LinusN once gave me an updated one which should fix the bootloader USB for the affected H300s. That one actually broke bootloader USB for me...
14:37:45 Join RoC_MasterMind [0] (
14:38:02pondlifeI believe he found out why and the current SVN should be fine
14:39:40linuxstbpreglow: ;) They seem more tolerant than us though - they like to support any combination that's physically possible...
14:40:00linuxstbWhereas we just say "fix your broken files"...
14:40:01 Join spiorf [0] (
14:40:13 Quit RoC_MasterMind (Client Quit)
14:40:40amiconnlinuxstb: Imho rockbox shouldn't fail on wrongly tagged files, it should just skip them, but not evaluate them
14:42:24amiconnIt should definitely not crash on bad files
14:44:18*preglow should optimize speex some
14:44:18linuxstbI agree. Most codecs will skip an id3v2 tag at the moment, but I think that's about as far as Rockbox goes.
14:44:19 Join roolku [0] (
14:44:50amiconnI mean it should never crash, even if you e.g. pass an mp3 file to the wma decoder or something like this
14:45:04Lloreanpreglow: Speex still doesn't work with stereo files, does it?
14:45:32preglowLlorean: haven't the foggiest
14:45:41GodEater_no it doesn't
14:45:46GodEater_you just get silence
14:47:08preglowwell, that sucks
14:47:34GodEater_safetydan took a look at it but I'm not sure if he ever worked out what was wrong
14:48:43*GodEater_ wonders why OSS podcasts don't make more use of speex
14:49:37JdGordoncoz they are hypocrites!
14:49:46roolkuBagder: the "annoying 'Saving settings' splash" has its purpose: It is the only feedback that the button press has registered (delay while disk is spinning up). If you are standing in the freezing rain and want to quickly move on to the next textfile this is quite important
14:49:48markunGodEater_: I saw that project gutenberg is now starting to use speex for audiobooks:
14:50:01roolkuBagder: could you tell me what is annoying about it?
14:50:23JdGordonBagder is in china atm iirc
14:50:32Nico_Proolku: I think you mean Zagor
14:50:48roolkuwhoops, yes, easy to confuse
14:50:59Zagorcan you tell me what its' purpose is?
14:51:07roolkuI did?
14:51:27roolkugive feedback that the button has registered and you don't need to keep pressing
14:51:39Zagorsorry, didn't see the above line.
14:51:53ZagorI think such feedback would be better achieved by simply clearing the screen or some such.
14:52:13ZagorI found the splash exteremely annoying
14:52:17amiconnlinuxstb: Do you have plans to do optimisation work on libdemac?
14:52:54*jhMikeS threw a "Saving Settings" splash in jpeg.c but only when they change. It help distinguish it from a lockup.
14:53:06amiconnThere is quite some potential for speedup especially on coldfire
14:53:08roolkuwell, looking at a blank screen strikes me as less usefull than seeing the text with a splash on top
14:53:22linuxstbamiconn: Not for Coldfire - I don't know it. But yes, I would like to spend some more time on ARM at some point.
14:53:44Zagorroolku: sure, but the point is normally that splash i shown for a mere fraction of a second and just becomes "UI noise"
14:53:45amiconnThe scalarproduct() would fly when using emac
14:53:47markunamiconn: would be great if you could at least get some compression level realtime
14:54:08amiconnAnd the additions and subractions could at least profit from burst reads & writes
14:54:25amiconnAre these buffers in iram or dram?
14:54:51linuxstbAs much is in IRAM as I could put there - I spent some time optimising iram usage for Coldfire.
14:54:52Nico_Proolku: the splash should probably be shown only when the settings actually need saving
14:55:09linuxstbamiconn: Which buffers specifically?
14:55:36amiconnThose on which scalarprodcut() and vector_add() / vector_sub() are used upon
14:55:41roolkuNico_P: they are always saved on exit, unless I overlooked something
14:55:48Nico_PGodEater_: I've just used SVN on my gigabeat for a bit and I don't find it any faster to skip to an unbuffered file when the disk isn't spinning
14:55:48amiconn(in filter.c afaics)
14:55:58 Quit MethoS- (Remote closed the connection)
14:56:03Nico_Proolku: then maybe it could be nice to not save them each time
14:56:41roolkuNico_P: could you name a situation when this doesn't happen (we are talking text position here)
14:57:09Nico_Proolku: ah, right. I didn't realise this was part of the settings too
14:57:27linuxstbamiconn: Yes, they are.
14:57:58amiconnEven for the largest filer size?
14:58:22roolkuthere is a significant delay on exiting the text viewer dues to disk spin up (on targets with disks)
14:58:28 Join thegeek_ [0] (
14:58:35linuxstbNot for the "insane" 1280 order filter, no. But I don't think that's realistic on anything other than the Gigabeat S
14:58:37Zagorroolku: ...which in itself is annoying.
14:59:08 Join MethoS- [0] (
14:59:13Zagorroolku: I think you/we need to consider there are both casual and "power" users of the viewer. the constant setting saving is good for power users, but annoying for casual users.
14:59:13roolkuZagor: I would aggree, but unless we implement delayed write, I don't see how this can change?
14:59:13linuxstbamiconn: But the buffers for compression level up to and including -c4000 are in IRAM. Only -c5000 (aka insane) uses that 1280 order buffer, which isn't in IRAM.
14:59:18 Join low_light [0] (i=c730190a@gateway/web/cgi-irc/
14:59:22Nico_Proolku: is it really necessary to save text position ? IMHO only saving the settings (if necessary) would be enough
14:59:25amiconnI agree with roolku that plugins should indicate that they're saving settings on exit
14:59:58amiconnBut also, plugins should keep track of their settings, and *only* save them if they really changed
15:00:01Nico_PI agree too on the splash, but I remember being annoyed with having to wait for the disk to spin up on exit
15:00:36 Quit thegeek (Read error: 110 (Connection timed out))
15:00:36amiconnSaving text position should either be optional, or when automatic, only apply to larger text files
15:00:48roolkuthe best solution would be to be able to queue the settings write somehow and do it later
15:00:52amiconnlinuxstb: Do we have test files?
15:01:06amiconnAnd I remember having seen a speed table...
15:01:30linuxstbamiconn: I can give you my test file
15:01:33 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
15:01:36Lloreanamiconn: Optional IMHO, otherwise you'll have to set an arbitrary decision as to what "larger" is.
15:02:25roolkuthere is the same issue with saving resume position in the video players
15:02:27 Quit hcs (Client Quit)
15:02:28low_lightjhMikeS: according to the mrobe100 disassembly (pp5020), there's some register misnaming...
15:02:31amiconnLlorean: A natural limit would be "doesn't fit in the buffer as a whole".
15:02:45low_lightspecifically, XMB_NOR_CFG & XMB_RAM_CFG
15:03:00Nico_Proolku: btw, it would be nice if the wide view could scroll column by column and display a scrollbar at the bottom
15:03:25amiconnScroll by column? Why that?
15:03:35jhMikeSlow_light: I thought that was pp5002 - amiconn said
15:03:43roolkuwhat constitutes a column?
15:03:47amiconnI agree on the scrollbar though (on bitmap targets of course, and optional like the vertical one)
15:04:04amiconnroolku: Yeah, good question as well
15:04:06low_lightjhMikeS: no, another pp5020
15:04:28amiconnjhMikeS: I never said m:robe 100 is PP5002...
15:05:14 Quit pill (Nick collision from services.)
15:05:19roolkuvertical scrollbar could be based on visible contents only I suppose, you don't want to have a huge text file to find the maximum line length
15:05:23 Join pill [0] (
15:05:36jhMikeSamiconn: either you mistyped it or I recalled incorrectly
15:05:45 Part Llorean
15:06:05roolkuaerm, horizontal scrollbar
15:06:16jhMikeSlow_light: I'm pretty sure DEV_TIMING1 isn't XMB_NOR_CFG
15:06:20 Join Llorean [0] (n=llorean@rockbox/administrator/Llorean)
15:06:25amiconnlinuxstb: Umm, and the predictor also looks like something that might love emac
15:06:30 Quit pill (Nick collision from services.)
15:06:39 Join _pill [0] (
15:06:56Nico_Proolku: yeah, looking for the max line len only on the screen would make sense. a column is like a vertical line
15:06:57jhMikeSI'll have another look at the H10 fw
15:06:58amiconnroolku: The vertical scrollbar already exists...
15:07:36roolkuamiconn: yes, I mixed them up
15:07:51Nico_Pamiconn: I'd like to scroll by column because I hate paged scrolling and currently in wide mode the scrolling behaves like pages scrolling
15:08:11Nico_P(the horizontal scrolling, that is)
15:08:33JdGordonwell this sucks!
15:08:36amiconnHehe, I prefer paged scrolling because line scrolling is too slow. Actually you can always do both
15:08:38roolkuNico_P: but your approach would only work for fixed-width fonts?
15:08:50amiconnAnd a column is undefined for proportional fonts
15:09:32amiconnYou could of course use some approximate width as "column", e.g. the width of 'A' (as used in several places)
15:09:35linuxstbamiconn: Yes, there's potential in lots of places...
15:09:47Nico_Phmm true. but maybe scroll by a smaller amount them
15:09:55Nico_Por allow some overlap at least
15:10:07jhMikeSlet's just have lists have colums that are defined :)
15:10:17amiconnlinuxstb: demac is all 16 bit arithmetics?
15:10:40PaulJamNico_P: isn't there already a buttoncombination to scroll by column? (on h300 play + left/right iirc)
15:10:50linuxstbNo, I think the predictors are 32-bit, and filters 16-bit (or vice-versa)
15:12:11roolkuI think it could in fact work, just by chopping off the same number of chars at the beginning of the every line
15:12:17amiconnHmm. The predictor uses mla on arm
15:12:37amiconnSo it might be 32 bit but not overflow 32 bits in multiply-add?
15:12:37JdGordonkkurbjun: time to wake up :D /me is stuck
15:12:42***Saving seen data "./dancer.seen"
15:14:04jhMikeSlow_light: well, how bout that :) I guess DEV_TIMING1 should be renamed and the other
15:14:39 Join nicktastic [0] (n=nick@unaffiliated/nicktastic)
15:14:45linuxstbamiconn: The C code uses 32-bits everywhere, so I think that's enough. It's been about 6 months since I looked at this though, and have forgotten the details...
15:15:18linuxstbs/6/4/ (my memory is bad...)
15:17:17jhMikeSlow_light: don't hestite to correct errors. they're bound to occur.
15:17:54 Quit MethoS- (Remote closed the connection)
15:18:04 Join MethoS- [0] (
15:18:56 Quit MethoS- (Remote closed the connection)
15:19:08 Join MethoS- [0] (
15:22:07 Join nerochiaro [0] (
15:22:26JdGordonis there any reason why an irq would only trigger once in a regular build, but work fine in a bootloader build?
15:22:33low_lightjhMikeS: ok...but I'm not in position to do any commits
15:23:52 Join ToHellWithGA [0] (
15:23:53nerochiaroquick question: anyone ever tried to see if it's possible to port rockbox to the creative zen v player ?
15:25:09Lloreannerochiaro: As long as a player is firmware upgradeable it's always "possible", it's just a question of whether there are people with the drive to do all the necessary work.
15:25:27amiconnlinuxstb: I think at least -c2000 should be manageable realtime on cf. Perhaps even -c3000
15:25:45nerochiarowell, i meant to ask if there was already an attempt in progress somewhere, but i worded it wrong
15:25:46linuxstbamiconn: -c3000 would be the most useful - I think that's the default.
15:25:54amiconnJust found the wiki page about your decoder - it's quite hidden...
15:26:21jhMikeSJdGordon: that sounds odd
15:26:27Lloreannerochiaro: Pretty much every known attempt has a thread in the New Ports forum as an attempt to call attention to other interested parties
15:26:48nerochiaroi'll check that
15:27:16JdGordonjhMikeS: well, im not actually sure if the irq isnt triggering... i only get the BUTTON_TOUCHPAD once, and i tihnk ive found everywhere it could be filtered out, so im confused
15:28:17jhMikeSput a counter in the handler and splash it somewhere or show it the debug menu somewhere?
15:28:32linuxstbnerochiaro: There are people working on porting Rockbox to the m:robe, which has the same CPU as the Zen - that will make a Zen port easier, but still needs interested and skilled people to do it.
15:29:31nerochiarolinuxstb: the zen V ? i see something about the zen vision M, but not the V
15:30:29linuxstbnerochiaro: I just assumed they all used the same CPU. I could very well be wrong though...
15:30:48 Quit MethoS- (Remote closed the connection)
15:30:59 Join MethoS- [0] (
15:31:33nerochiarolinuxstb: not sure what hardware the V has inside. i'm really asking here on behalf of a friend who has it
15:31:35JdGordonjhMikeS: yep, the irq counter isnt gong up... its only being triggered once!
15:32:08 Nick Tanuva|Zzz is now known as Tanuva (n=tanuva@
15:33:40 Quit jhulst ("Konversation terminated!")
15:34:02 Quit MethoS- (Remote closed the connection)
15:34:32 Quit roolku ()
15:35:30 Quit Jon-Kha (Remote closed the connection)
15:37:44 Join Jon-Kha [0] (
15:38:26 Quit aliask ("ChatZilla [Firefox]")
15:40:20 Join mrkiko [0] (
15:40:23mrkikoHi all!
15:40:32nerochiaromrkiko: hi there
15:44:31mrkikonerochiaro: italian?
15:47:24mrkikonerochiaro :)
15:47:36mrkikoyour nick says this
15:47:39jhMikeSJdGordon: I can see no reason for that unless something else isn't implemented properly somehow. :\
15:48:02mrkikonerochiaro: rockbox user ?
15:49:06jhMikeSJdGordon: does it get acked in the bl and not in the fw?
15:49:25JdGordonyeah, works fine in the bl
15:49:48JdGordonunfortunatly, the earliest point i can do any usefull debuggin is in action.c
15:50:04JdGordonbut ive got it splashing the irq count if it changes, and it sits on one
15:50:25jhMikeSand counts up in the bl?
15:50:55JdGordoni havnt checked, but the cursor follows the stylus, so it must be working.. thats the only way the touch gets into the button queue
15:52:33JdGordonyes, goes up about 100 every touch!
15:52:35 Join Gnu47 [0] (
15:53:14JdGordonyeah, the irq is triggered as soon as the data is ready
15:53:23JdGordonwhich is very fast :p
15:53:23jhMikeS100 times?
15:53:46JdGordonit reads its inputs at 12mhz iirc
15:53:58jhMikeSshouldn't data be ready once?
15:54:29jhMikeSthat seems weird it should act that way
15:54:30 Join MethoS- [0] (
15:54:56JdGordonyeah, but it works for now...
15:55:43jhMikeSI wonder...that doesn't seem like it's really working right
15:56:37JdGordonna, it is... we could set it to only send the data when we ask for it, but this way is how the OF set it so we havnt been bothered changing it
15:59:54 Nick parafin|away is now known as parafin (
16:00:55 Join bluebrother [0] (i=55IFMLeW@rockbox/staff/bluebrother)
16:03:20 Join scorche|w [0] (
16:22:48 Quit Jon-Kha (Remote closed the connection)
16:24:53 Quit kubiix (Read error: 104 (Connection reset by peer))
16:25:02 Join Jon-Kha [0] (
16:28:20 Quit Zagor ("Client exiting")
16:30:08Nico_Ppondlife: I've pushed some changes you might like
16:30:17pondlifeOK, will pull
16:30:47Nico_Ppondlife: btw, thanks for your test album, it's very useful for testing and funny too
16:31:50Nico_Pwith the latest commit, you can have ~30 tracks in the buffer and skip back and forth within these without having the disk spin at all
16:31:55 Quit MethoS- (Remote closed the connection)
16:32:17pondlifeCool, so you only unbuffer when you need the room?
16:32:33pondlifeThat's how it should be
16:33:02pondlifep.s. I never did see the buffering debug screen...
16:33:07pondlifeAm building now.
16:33:29pondlifemake install
16:33:30LloreanNico_P: Does this also mean that if you set put, say, 3 tracks in a playlist and they all fit in the buffer, "Repeat" won't result in rebuffers?
16:33:52LloreanAssuming, that is, the playlist isn't reshuffled?
16:34:04pondlifeLlorean: It should do... a bug if not I reckon :p
16:34:15Lloreanpondlife: Well, it's a long, long standing bug right now, I believe.
16:34:44Nico_PLlorean: good question, but I think it will rebuffer. That's probably because of the playlist code though. I'll test to make sure
16:34:46LloreanSo if it doesn't, it's not a new bug, just not a fix for the old one.
16:35:12pondlifeOr a bug in the new code if it's intended not to rebuffer in this case...
16:35:24Lloreanpondlife: Well, it's more a playback code issue, I'd imagine.
16:35:31pondlifeI think the old behaviour was more by design than a bug
16:35:41mrkikoIs there a completely hardware-opened mp3 player where rockbox can be used?
16:35:53 Join toffe82 [0] (
16:35:53peturNico_P: there's also an issue where multiple times the same song causes the same file to be buffered again and again... will your system also handle that?
16:35:54Lloreanpondlife: I think the old behaviour was more because the original buffer often wouldn't hold even one song.
16:35:56Nico_PI think when repeat is set, the playlist code makes the playback code believe it's like an infinit playlist
16:36:21pondlifeLlorean: Indeed, the design was the same since Archos Player.
16:36:33LloreanNico_P: Would it at all be possible to have it detect if a song in the playlist is already buffered, and go to that?
16:36:37Nico_Ppetur: ATM it doesn't but it could
16:36:58LloreanNico_P: For example if my playlist is song A, B, A, C A, instead of buffering A three times, somehow handle that case? (It seems like that'd also fix rebufering with repeat, even with shuffle)
16:37:15*petur discovers he asked about the same thing Llorean did :/
16:37:17Nico_PLlorean: I think so, as the filename is stored in the handles, so bufopen could return the id of an existing handles in case of duplicates
16:37:57JdGordonexcept you have to make sure you dont trash the handle untill everything using it is done
16:37:58Nico_Pthe only problem in that case is that I'd need to be very careful before bufclosing
16:38:14Nico_PJdGordon: we agree :)
16:38:14LloreanWell, just something to keep in mind at least. :)
16:38:37*JdGordon not sure its worth the overhead
16:38:50Nico_Pthe case i had in mind for that is album art... we don't want to buffer ten times the same bitmap
16:39:14LloreanJdGordon: For short playlists (or long playlists of very short files) it could drastically reduce disk spinups.
16:39:29*Llorean is curious how long his gigabeat would last on RAM-only playback.
16:40:06Nico_PLlorean: the main buffer is roughly 30 MB so you do the math according to your avg bitrate :)
16:41:25LloreanNico_P: If you take a 30 megabyte segment of my playlist the average bitrate could be as low as 100bps or as high as 990kbps or so.
16:41:41 Join MethoS- [0] (
16:42:17 Nick fxb is now known as fxb__ (
16:42:18LloreanActually it's probably closer to maybe 500bps
16:42:21pondlifeNico_P: I found your debug screen...
16:42:32pondlifeWhat's the "useful" figure mean?
16:42:33Nico_Ppondlife: was it that hard to find ?
16:42:37pondlifeNo :)
16:43:00Nico_Ppondlife: basically it's the data that hasn't been played yet
16:43:38Nico_PI know the labels can be a bit confusing if you're not familiar with the code
16:43:51pondlifeHmm, so you consider data as useful still even though it could be skipped back into?
16:43:55pondlifeOr rewound into
16:44:06Nico_Phmm I have a one track long playlist and the track is buffered 4 times
16:44:31JdGordonthat makes sense though
16:44:52mrkikoIn my opinion the solution to have rockbox on the hard drive and not on the flash is very very intelligent
16:44:52Nico_Ppondlife: no. useful data is the data between the current read position and the write position
16:44:59JdGordonNico_P: if you really want to add handle sharing, you should only do it for repeat playlists.. not for when 1 track is in the playlist multiple times
16:45:11pondlifeWhy do my 7 buffered tracks have 14 handles? Not all codecs, surely?
16:45:23Nico_PJdGordon: that distinction doesn't apply in buffering.c
16:45:35Nico_Ppondlife: each track has metadata and audio
16:45:38pondlifeAh, ok
16:45:52pondlifeI imagined lots of buffered MP3 codecs
16:45:54JdGordonNico_P: hmm, ok
16:46:06Nico_Ppondlife: data that can be seeked back to or skipped back to isn't "useful"
16:46:15 Join Rob2222 [0] (
16:46:35Nico_Pat least not until you go back to before it
16:46:46 Join Crash91 [0] (n=evil91@
16:46:58Nico_Ppondlife: to see it you can try pressing up and down in the debug screen
16:47:14nerochiaromrkiko: sorry, i had people at the door. no, i'm not really a rockbox user. i just lurk here from time to time and borrow some rockbox source to use in another other project
16:47:21Crash91is there any specific reson why rockbox cant support 32x32 icons?
16:47:56LloreanCrash91: I don't think there are any fonts 32 tall, even.
16:48:30LloreanDoes it just not load the images?
16:48:48Nico_PJdGordon: maybe to handle multiple bufopens for the same file I could add a counter in the handle struct ? so that it has to be bufclosed the same number of times before it actually disappears
16:48:53Crash91i havent tried yet, im thinking of making one, but it says its limited to 24x24
16:49:17LloreanCrash91: My guess is it was probably decided that icons that large would take up too much screen space. We have to pre-allocate memory for them, so a limit had to be set somewhere.
16:49:28Nico_Ponly problem with that is I can't detect multiple identical bufallocs with metadata
16:49:32Crash91kk =) 24px it is
16:50:07Crash91btw...on the sansa...what if the disk is full, where does rockbox store its memory?
16:50:14Crash91in the FW partition?
16:50:30LloreanCrash91: I don't understand the question, at all.
16:50:45JdGordonNico_P: I woudltn worry about it untill mob is working fully anyway
16:50:48Crash91well rockbox has some stuff on disk, like RAM right?
16:50:53LloreanCrash91: If the disk is full and Rockbox needs disk space, there will be errors. But "memory" refers to RAM, not disk space.
16:50:53pondlifeNico_P: Nice work
16:51:05Nico_PJdGordon: yeah, my current goal is "no regressions"
16:51:45LloreanJdGordon: I just brought it up so that the idea can be there, so he doesn't accidentally code it into impossibility if it isn't already. :)
16:51:50Nico_PI'm thinking I can commit when there aren't any major regressions left and improve after that
16:52:22pondlifeOnly slight problem is that skipping backwards repeatedly, results in another rebuffering. But that's fair enough, our main aim is playing forwards!
16:52:44Nico_PLlorean: *if* it's impossible, it's because of the current code. hopefully the new buffering code is flexible enough to handle it correctly
16:53:08Nico_Ppondlife: yes, but you can't predict how far the user is going to skip backwards :)
16:53:19LloreanNico_P: Hopefully. :)
16:53:21pondlifeBest to assume one track like you do I think..
16:56:01mrkikoehi guys!
16:56:20mrkikoHow can I work with fwpatcher.exe in Linux?
16:56:30Nico_Pmrkiko: it works with wine
16:56:37 Quit Pyromancer (Remote closed the connection)
16:56:40Nico_Pbut you can use rbutil or the command line tools
16:56:45 Quit Rob222241 (Read error: 113 (No route to host))
16:57:11mrkikoNico_P: ok, I'll try cbecause I can't use graphic-mode apps
17:00:48 Quit Jon-Kha ("leaving")
17:01:07 Join jgarvey [0] (
17:01:15 Join barrywardell [0] (
17:08:47 Join mickleby [0] (
17:09:09 Part mickleby
17:10:55 Part Crash91
17:12:10Nico_Ppondlife: do you find skippinf to an unbuffered track to be any slower in MoB than in SVN ?
17:12:35pondlifeNot really possible to tell on the sim
17:12:35Nico_Por rather "skipping"
17:12:44***Saving seen data "./dancer.seen"
17:12:46Nico_Ppondlife: haven't tried on H300 ?
17:13:01pondlifeI left it in the car..
17:13:08Nico_Pbut from your previous tests ?
17:13:08pondlife..which is with mrs pondlife
17:13:19Nico_Pit shouldn't have changed recently in MoB
17:13:29pondlifeEarly ones seemed slower in MoB, but I thought that improved
17:13:47pondlifeShould be possible to knock up a timer to prove it, right?
17:13:58pondlifeDo we have a system tick counter?
17:15:16Nico_PI think so
17:15:26pondlifeOne thing which I think is new to MoB... my WPS has a playlist index displayed (9 of 2300) or whatever...
17:15:58 Quit MethoS- (Remote closed the connection)
17:15:59Nico_PI see nothing wrong with that
17:16:02pondlifeI thought the number (9) used to increase immediately when I skipped. Now it's rather slow
17:16:21Nico_Pah yes it has a small delay compared to the rest of the metadata
17:16:24pondlifeI mean when skipping during buffering
17:16:43pondlifeI've just got the wrong track displayed
17:16:58pondlifePlaying Spiritualized, but displaying Elvis Costello..
17:17:22pondlifePlaying track 15, but displaying track 16
17:17:26Nico_Phow did you manage that ?
17:17:35pondlifeJust skipping forward during buffering
17:18:04pondlifeForward repeatedly fast
17:19:03pondlifeCan't repro it yet though
17:19:45pondlifeRemember I crossfade everything too, in case that's relevant.
17:21:40Nico_Ppondlife: about the (x of y) playlist indicator, I'm pretty sure MoB is an improvement over SVN. In SVN, when you skip to an unbuffered track (say, from 14 to 15), the playlist pos would stay on 14 while the rest of the WPS shows the filename of track 15 (because metadata isn't ready), and then the playlist indicator moves to 15 when the metadata arrives
17:22:12Nico_Pin MoB, metadata is displayed immediately when you skip
17:23:03pondlifeI'll need to compare again..
17:23:20pondlifeI recall that the x of y part was quick though
17:23:41pondlifeSo I could get to a track quite quickly before waiting for buffering
17:24:41 Quit JdGordon (Remote closed the connection)
17:25:04Nico_Pah, comparing I see that in mob the playlist position isn't always updated. there are some skips, like 9->12->15
17:25:14Nico_Pin SVN there seems to be no skips
17:28:19Nico_Phmm I managed to break playback by skipping forwards repeatedly. now stopping then restarting doesn't even work
17:28:29 Join dxm [0] (
17:29:55pondlifeNico_P: Yes that's what I'm seeing, SVN always moves smoothly a track at a time
17:29:59pondlifeSorry, work calls.
17:30:43Nico_Pyeah, now I see what you mean
17:33:48dxmrockbox for ipod touch works good?
17:34:49bluebrotheryes. It hasn't even ported.
17:35:17*bluebrother wonders how it should work "good" if the front page says "not Touch"
17:37:53 Join desowin [0] (
17:39:31linuxstbdxm: No, Rockbox won't work at all on the ipod touch.
17:41:28GodEater_not only won't it work, it won't even install
17:47:02low_lightwouldn't it be possible to run Rockbox as an app on the itouch/iphone?
17:47:14Nico_PGodEater_: I've been doing some tests and I don't see any cases where mob is slower to skip to unbuffered tracks than svn... maybe it's just because I'm on gigabeat though
17:47:58GodEater_I've not had a MoB build on my ipod for a couple of days - I wanted to see how the proper multi-core stuff was working out
17:48:04GodEater_I'll build another one now and try it again
17:48:34Nico_Pdoes what I say suprise you ?
17:48:41 Quit Gnu47 ("You know you'll miss me :P")
17:49:32GodEater_let's just say it doesn't match my experience :)
17:50:08Nico_Phmm then I'd be very grateful if you could use git bisect to track down the faulty commit
17:50:42*GodEater_ goes to try to replicate his experience in svn first
17:53:44GodEater_Nico_P: how are you doing your test ?
17:54:03Nico_Pdo you mean on the player or with git ?
17:54:07GodEater_on the player
17:55:05Nico_PI've tried starting playing an album with the second track (from the browser), waiting for the disk to spin down (set the spin down time to min), and then skipping back
17:55:52GodEater_ok - let me do that exact same test then
17:56:09 Quit atsea-22 (Read error: 104 (Connection reset by peer))
17:56:11Nico_PI've noticed that svn seems to spin the disk when I skip to the beginning of a track, which is strange, so I used the audio debug screen to skip back
17:56:12GodEater_what time are you getting before playback of the unbuffered track
17:56:16n1sNico_P: in svn spindown time doesn't affect spindown when buffering, it just spins down as soon as it's done
17:56:40Nico_PGodEater_: roughly 3 secs
17:56:49n1sdid you change that with mob?
17:56:50GodEater_yeah same here with svn
17:56:58Nico_Pn1s: I don't think so
17:57:03*GodEater_ goes to update his git tree
17:57:06n1sgood :-)
17:57:21Nico_Pn1s: do you happen to know what code does that ?
17:57:22 Quit ToHellWithGA ("You know you'll miss me a lot.")
17:58:05Nico_Pn1s: maybe the ata_sleep(); call ?
17:58:23Nico_Phmm I need to change that
17:58:58n1scould be that, I just remember that I asked one of the devs familiar with this stuff about it sometime
18:02:33 Quit desowin ("use linux")
18:02:37Nico_Pthe ata code desperately lacks comments :(
18:02:58mrkikoI found only rbutilqt for linux: no console tools
18:03:21Nico_Pmrkiko: you want to pacth the firmware file, right ?
18:04:03 Join Lear [0] (i=chatzill@rockbox/developer/lear)
18:09:28 Quit petur ("work->home")
18:14:02 Join midgey [0] (
18:15:12amiconnNico_P: svn always rebuffers when skipping back
18:15:36amiconnAnd yes, if you want to spin down immediately when done with buffering, you have to call ata_sleep();
18:15:39Nico_Pamiconn: is this desirable behaviour ?
18:15:47amiconnIt's KISS behaviour
18:15:56moosit doesn't rebufer in any case
18:16:11Nico_Pamiconn: I'm not sure I agree
18:16:28moosdepending what's left on the buffer, and if the track was completly buffered before
18:16:37amiconnThe ata code isn't complicated imho...
18:16:53amiconnThe playback and buffering code, however, is
18:17:12Nico_Pamiconn: ah, you were saying KISS about ata_sleep() ?
18:17:27amiconnNo, about always spinning up when skipping back
18:17:36Nico_Pwhy is it ?
18:18:02moosexemple: A,B,C fit on the buffer completly, then if I want to skip back for track B, it will don't rebuffer
18:18:21Nico_Pmoos: I thought that too but I don't think it's true
18:18:29amiconnsvn has no sophisticated code to track whether the backskip could go without rebuffering
18:18:42 Nick bagawk_ is now known as bagawk (n=lee@unaffiliated/bagawk)
18:18:49moosNico_P: that's true
18:19:01amiconnAt least that's what I know, I never skip much
18:19:41Nico_Pamiconn: but you agree it's desirable to avoid rebuffering when skipping back to something we have already in the buffer ?
18:19:44amiconnI said it's KISS behaviour, not that it's desired behaviour
18:19:55 Join atsea-22 [0] (i=atsea-@gateway/tor/x-184533c34ee9999e)
18:20:20moosthe only case you will have rebuffering for track B in our exemple, that's if you made the disc to spin for browsing for exemple, cause the buffer is refiled at nay disc activity
18:21:32 Quit midgey ()
18:22:28 Quit barrywardell ()
18:24:01moosNico_P: currently it *doesn't* rebuffer if the track is in totality in the buffer
18:24:28Nico_Pyeah that's what I just saw... I must've been confused
18:24:28 Join Jon-Kha [0] (
18:24:38moosyou will have refilling for the first or the last filed files
18:25:01Nico_Pwhat do you mean ?
18:25:20moostake my simply exemple
18:25:45moos3 tracks buffered
18:26:11moosyou will not have rebuffering for the track B only
18:26:31Nico_Pwhy ?
18:26:45moosbecause the first one isn't 100% filled and the last one too
18:27:04Nico_Phmm than I believe my code is an improvement
18:27:31 Quit obo ("bye")
18:27:36mooslet's hope then :)
18:27:49Nico_Phave you tried it or had a look at it ?
18:28:48moosbut this buffering thing for skiping back is more "problematic" for targets with less availiable memory
18:29:33moosfor exemple on my X5, 12mo and something isn't big, then I have big chance to have rebuffering for a skip back
18:30:15amiconnlinuxstb: Looks like test_codec freezes the target if it's waiting for the final button press (which btw. is illogical) and the poweroff timeout kicks in
18:31:09moosbut on my gigabeat with almost 30 mon, that's better cause betwen the first and the last buffered tracks, there are few 100% buffered tracks that will not need disc for skip back
18:32:00moosNico_P: I didn't find the time for test your new code yet
18:32:45moosNico_P: pm for french, that's better :)
18:34:20 Quit DataGhost (Read error: 110 (Connection timed out))
18:36:51 Join Ebert [0] (
18:45:44 Quit Ebert ()
18:47:33 Join midgey [0] (
18:48:28 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
18:49:20pondlifeNico_P: Just a quick note - On H340, I'm still getting the playback break-up/pause when initially starting.
18:49:29pondlifei.e. a short break in playback
18:49:34pondlifeAbout 0.5s
18:49:57Nico_Ppondlife: after playing the first few seconds ?
18:50:04pondlifeSame as before
18:50:17Nico_Phas it always been like this ?
18:50:39pondlifeOn MoB, yes
18:50:42pondlifeNot SVN though
18:50:55Nico_Pyeah, that's what I meant ;)
18:51:10pondlifeOops just crashed by skipping back
18:51:18pondlifeHard crash
18:51:44pondlifeSorry; I wasn't even trying to make it fail
18:52:06Nico_Phehe, don't apologise ;)
18:52:41pondlifeYou need to borrow a slow target, maybe
18:53:32 Join desowin [0] (
18:54:45Nico_Ppondlife: I have an H300 but I sold it to my mother
18:54:59Nico_Pand I don't live with my parents
18:55:31pondlifeSuggest you see if you can borrow it back for a bit :)
18:56:36 Join Gnu47 [0] (
18:57:07Nico_Pnext time I go back home it'll be for a week though, so I'll be able to use it
18:57:18Nico_PI'll have less time to work on MoB though
18:58:17pondlifeMaybe swap it for your Gigabeat until MoB is committed, or shortly after? ;)
18:58:24Nico_Ppondlife: would you mind testing the commit I just pushed ?
18:58:54Nico_PI do hope to commit mob before I go back home (in a bit more than a week)
18:59:14pondlifeCool and ambitious
18:59:22pondlifeI'll test shortly, hopefully
18:59:34Nico_Pwell it's probably time to make the development a bit more public
18:59:55Nico_Ppondlife: cool :) it *might* fix the dropout
19:00:22pondlifeBefore you commit, could I ask you to do another run through the code and put even more comments in.
19:00:27pondlifeEspecially playback.c
19:00:41Nico_Pthat's probably a good idea, yes
19:00:54pondlifeWhile you have the knowledge
19:00:58 Join petur [0] (n=petur@rockbox/developer/petur)
19:01:22pondlifeHmm, on another topic, I though the Database Info screen used to update automatically during a rebuild or update?
19:01:29pondlifeIt only does it if I press a button now.
19:04:01 Join ilgufo [0] (
19:06:03 Join dez`` [0] (
19:06:24dez``hey fellow geeks!
19:06:35pondlifeSpeak for yerself
19:06:43dez``ok :)
19:06:56dez``i need some help
19:07:41krazykitwell, you don't get any unless you tell us what you need help with
19:08:39 Join lazka [0] (
19:08:45dez``i got the zip with the doomf.wad and the rockdoom.wad and my doom still hasnt worked
19:09:02dez``on the 30 ipod 5.5g
19:09:13 Join bertrik [0] (
19:09:28krazykitdez``, i didn't think those were distributed in zip files. are you following the wiki exactly?
19:09:46dez``theres a wiki for the rockbox doom?
19:10:10krazykityeah. do a search.
19:12:45***Saving seen data "./dancer.seen"
19:14:56 Join Domonoky [0] (
19:17:13dez``nope still doesnt work :'(
19:17:26dez``it says "missing base .wad"
19:19:09 Quit dxm ()
19:19:10amiconnlinuxstb: Up to 93% realtime for -c3000 on coldfire :)
19:19:33amiconn(starting from 64%)
19:19:54amiconnThat's with just the filter math optimised
19:22:03bertrikdez``: did you follow the doom wiki?
19:22:53 Join ompaul [0] (n=ompaul@gnewsense/friend/ompaul)
19:26:22preglowamiconn: you're optimizing ape?
19:26:24 Part pondlife ("Gone")
19:26:31 Nick fxb__ is now known as fxb (
19:27:48preglowwhy? :P
19:28:05amiconnAnother half-baked feature...
19:28:29dez``got it working
19:28:32dez``thanks guys
19:29:52 Join haemmy [0] (n=stefan@
19:30:40*preglow bakesall his features at 225 degrees for fourty minutes
19:31:10 Quit Gnu47 ("You know you'll miss me :P")
19:35:52BigBambiDo we have a splash or anything appearing for low battery on Sansas?
19:37:09BigBambior is this chap talking about something in the OF:
19:37:55 Join Gnu47 [0] (
19:39:17bertrikThere's a Battery empty! RECHARGE! message and I think I've seen it once
19:40:03BigBambibut he can't be getting that + 40%
19:40:11BigBambiI think I shall query him
19:40:41 Join mrkiko_ [0] (
19:42:02bertrikI'm thinking of writing a texture mapped cube demo, perhaps re-using the rockbox logo that's shown at startup
19:42:27bluebrotherDomonoky: do you have anything that we should add before making a new release of rbutil?
19:43:25Domonokyno, feel free to make a release..
19:43:53Nico_Pbluebrother: please make a tag for the release as well
19:44:14bluebrothergood. I really need to do that soonish before RL occupies all my time (and I'm waiting for internet access)
19:44:18 Quit mrkiko (Read error: 110 (Connection timed out))
19:44:50Domonokyyes, please make a release, before you start your new job.. :-)
19:44:54 Join kugel [0] (i=kugel@unaffiliated/kugel)
19:45:14bluebrotherI need to move too ... will get quite rushy :(
19:45:42lazkais this known..? changing themes on my sansa sometimes resets rb settings.
19:46:11 Join kkurbjun_ [0] (n=kkurbjun@alamode.Mines.EDU)
19:46:23Domonokybluebrother: good luck..
19:46:46bluebrotherthanks ...
19:46:58 Join ToHellWithGA [0] (
19:50:09 Join Guile`` [0] (n=Guile@
19:51:53 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
19:57:21 Join obo [0] (n=obo@rockbox/developer/obo)
19:58:11 Quit lazka ("I'm off now")
20:00:59 Join kubiix [0] (
20:01:38 Join WalterEgo [0] (
20:01:42 Join FOAD_ [0] (
20:01:45hcsdoes anyone know offhand why the high score updating in rockbox isn't in show_details?
20:02:13hcsseems like that's where it ought to be, and because it isn't when switching hold on and off the high scores are erased and not redrawn
20:03:12 Quit Gnu47 ("You know you'll miss me :P")
20:05:01Nico_Pbluebrother: in rbutil's about dialog, I get $ symbols around the SVN revision
20:05:53Nico_Pbluebrother: "Version SVN $Revision 15190 $ (m1.0.2)"
20:05:58bluebrotherNico_P: yes, that's normal.
20:06:04 Quit dez`` ("Don't push the red button!")
20:06:10Nico_Pbluebrother: really, why ?
20:06:41 Nick mrkiko_ is now known as mrkiko (
20:06:54mrkikoHi boys!
20:07:03bluebrothersimply because it's something like the svn $Id$
20:07:16mrkikoI'm reading the wiki, but I still can't figure out from where descramble comes from
20:07:20bluebrotherand I haven't bothered removing the $ symbols −− I don't think it matters much.
20:07:29mrkikotools/*: what bz2 archive contains these tools?
20:07:54bluebrothermrkiko: you need to get the sources from svn
20:08:07mrkikoah ok
20:08:10mrkikocan you pass me the url?
20:08:42Nico_Pbluebrother: but in "#define VERSION "SVN $Revision$ (m1.0.2)", $Revision$ should be substituted with the svn rev, shouldn't it? then why do the dollar symbols appear ?
20:09:22bluebrotherbecause the $ symbols won't get replaced. It's the same as as $Id$ −− the $ symbols won't get stripped too.
20:09:36 Join Gnu47 [0] (
20:11:22 Join jac0b-work [0] (
20:12:20jac0b-worktoffe82: did you check to see if the F series battery fits in the S series
20:13:24toffe82jac0b-work: yes, it fits, but I don't think the one you found will fit
20:14:21toffe82jac0b-work: with the battery of the F in the s it is dificult to reinstal the small board withe the connectors, it is really tight
20:14:52jac0b-worktoffe82: I see
20:15:00WalterEgoI see JDGordon has worked a bit on the M:robe scrollpad thing.. D'you guys reckon some of that code could be used to work the H10 unidimensional (up-down) scrollpad someday?
20:15:28Nico_PWalterEgo: I doubt it, it's really not the same thing
20:15:50toffe82jac0b-work: but markun show me another ref of battery from the same provider that say for gigabeat s an f
20:16:14WalterEgoShame... If I understood anything at programming I'd have a look.. Alas, basic php includes for webpage is about all I could understand :p
20:16:15bluebrotherZagor: when reading the logs, can you copy the new rbutil 1.0.2 to the download server?
20:16:19jac0b-worktoffe82: if it is tight with the F series battery then for sure the iPod battery won't fir
20:16:51jac0b-workyeah but its 850mah
20:17:12jac0b-worki got a username to sign on to the cameron sino site
20:17:23toffe82this one say for all the gigabeat, is it the same as yours ?
20:17:57jac0b-workit looks the same
20:18:08 Quit FOAD (Connection timed out)
20:18:08jac0b-workbut who knows what you will get in the mail
20:18:08 Nick FOAD_ is now known as FOAD (
20:19:30 Quit hcs ("Leaving.")
20:21:42 Quit BigBambi (Read error: 110 (Connection timed out))
20:22:09jac0b-workthis is what cameron sino's site has
20:23:10n1sbertrik: this page has most of the dirt on the rockbox graphics api
20:23:24bertrikah thanks!
20:34:25jac0b-workhow is the metadata on buffer coming along?
20:35:11 Quit Gnu47 (Nick collision from services.)
20:35:19 Join Gnu47 [0] (
20:35:25 Quit kugel ("Benutzer ist abwesend.")
20:36:16 Join kugel [0] (i=kugel@unaffiliated/kugel)
20:39:37 Quit jac0b-work ("ChatZilla [Firefox]")
20:41:47 Quit miepchen^schlaf (Read error: 104 (Connection reset by peer))
20:46:39 Quit WalterEgo ()
20:48:08 Quit Gnu47 ("I'm off to stalk my Pah'Mach'kai")
20:48:28 Join Gnu47 [0] (
20:48:28 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
20:48:43 Quit Nico_P (Remote closed the connection)
20:48:46 Quit spiorf (Remote closed the connection)
20:49:51 Join Nico_P [0] (n=nicolas@rockbox/developer/NicoP)
20:53:10 Join rockboxnewbie [0] (i=44647301@gateway/web/cgi-irc/
20:54:31 Quit rockboxnewbie (Client Quit)
20:54:34 Join rockboxnewbie [0] (i=44647301@gateway/web/cgi-irc/
21:00:02 Quit rockboxnewbie ("CGI:IRC (Ping timeout)")
21:02:15 Join barrywardell [0] (i=c101acbc@gateway/web/cgi-irc/
21:02:39 Quit keanu (Read error: 104 (Connection reset by peer))
21:03:58 Join rockboxnewbie [0] (i=44647301@gateway/web/cgi-irc/
21:05:32 Quit barrywardell (Client Quit)
21:06:04 Quit rockboxnewbie (Client Quit)
21:06:09 Join rockboxnewbie [0] (i=44647301@gateway/web/cgi-irc/
21:07:45 Quit Toxicity999 (Read error: 104 (Connection reset by peer))
21:08:37 Join keanu [0] (n=keanu@unaffiliated/keanu)
21:10:44 Join Toxicity999 [0] (n=bryan@unaffiliated/Toxicity999)
21:11:07 Quit rockboxnewbie ("CGI:IRC (Ping timeout)")
21:12:47***Saving seen data "./dancer.seen"
21:16:40 Quit Tanuva (Read error: 104 (Connection reset by peer))
21:18:04 Quit kugel ("Benutzer ist abwesend.")
21:19:36 Join hcs [0] (n=agashlin@rockbox/contributor/hcs)
21:21:35 Join Tanuva [0] (n=tanuva@
21:30:29 Quit davina ("xchat on Ubuntu 7.04")
21:31:32bertrikanybody knows what causes the OF on the sansa to refresh its database?
21:35:59desowinit happens always when you saved anything to device in msc usb mode
21:37:49 Quit haemmy (Read error: 104 (Connection reset by peer))
21:38:08 Join Ebert [0] (
21:45:51krazykitbertrik, if you use an older version, the bootloader can prevent it from refreshing
21:46:27krazykitoh, you were answered in #rockbox-community already.
21:47:47bertriki won't mind helping out with finding the special bit for the latest version of the original firmware
21:47:57 Join przemhb [0] (
21:48:11przemhbHi all
21:48:22bertrikI mean, the special bit for updating the database that is set when something is written
21:49:30przemhbI've got a problem with linking a plugin I am writing - linker returns "undefined reference to memcpy"
21:50:05bluebrotherprzemhb: you need to access the functions through the plugin api pointer
21:50:08 Join pepie34 [0] (
21:50:34przemhbbluebrother: I am not using the memcpy at all in my code
21:50:38preglowprzemhb: gcc uses it
21:50:45preglowyou need to wrap memcpy so gcc sees it
21:50:46przemhbyes I know
21:50:55preglowi think a couple of plugs do this already
21:51:01Domonokythere is a macro for this... i dont remember the name..
21:51:06przemhbso how to do this?
21:51:20rasherprzemhb: MEM_FUNCTION_WRAPPERS(rb)
21:51:27przemhbthe only I do is coping GPIO's to a array...
21:51:50rasherrb being the plugin api pointer
21:51:51 Join DataGhost [0] (
21:52:06przemhbfor example I have: gpio[GPIOA][0] = GPIOA_OUTPUT_VAL;
21:52:16bertrikmaybe -fno-builtin stops gcc from using <string.h> functions?
21:52:35amiconnbertrik: It does not.
21:52:35przemhbI've stared the plugin basing on helloworld plugin code
21:52:51amiconnWe're alreadycompiling with -ffreestanding
21:53:11rasherprzemhb: put MEM_FUNCTION_WRAPPERS(rb) somewhere at the top of your plugin - not within any function, and the rb pointer must be global
21:53:19amiconnThe target environment must provide the 4 basic mem* functions
21:53:21Domonokyprzemhb: just add the macor rasher mentioned, to your code, and it will work.. :-)
21:53:33amiconnAh, yes
21:53:45amiconnDidn't remember we have a macro for that nowadays...
21:54:50rasherWhy don't splashes use the foreground/background colours?
21:54:57przemhbI've put this; I am waiting for make results now...
21:55:50amiconnrasher: They use the foreground colour
21:56:07rasherBut not the background? Sounds "dangerous"..
21:56:11przemhbit helped !
21:56:15amiconnBackground is different because that makes the more visible
21:56:17przemhbthank you guys
21:56:28amiconn(except on mono displays of course) ;)
21:56:41rasheramiconn: what if I'm using a light foreground? Splashes become useless..
21:57:28rasherseems not
21:57:52rasherlooks to me like splashes are always black on white (or light gray, perhaps)
21:58:03 Quit pepie34 ("Ex-Chat")
21:58:04amiconnAh, yes
21:58:11amiconnThey're black on grey
21:58:49amiconnThat must have changed at some point (the foreground, not the background)
21:59:17rasherAnd with good reason - using only one of them opens up potential black-on-black situations
21:59:22 Quit Gnu47 ("I'm off to stalk my Pah'Mach'kai")
21:59:29 Join eigma [0] (n=cat@
21:59:34amiconnNah, only grey on grey ;)
21:59:36kkurbjun_hey eigma :)
21:59:48eigmaok so this was the best idea I could come up with:
22:00:06eigmawe register remote_heartbeat_tick() which simply sends out a heartbeat packet. nothing more.. no reads.
22:00:11rasherI understand the point about making it more visible, but it also looks really strange with the gray in the middle of it all.
22:00:38kkurbjun_I did try registering that seperate
22:00:47kkurbjun_it didn't seem to work too well
22:00:48eigmabutton_read/button_read_device() non-blockingly reads the uart hardware fifo (which can be configured up to 32 bytes long) and sees if there's any heartbeat responses
22:00:56eigmaif so, it parses them and returns the key status
22:01:04 Join davina [0] (
22:01:04eigmaand then it purges the fifo
22:01:14eigmaand does not send any heartbeat
22:01:37kkurbjun_so right now that's close to what I am doing in the latest code
22:01:37eigmacan you explain what didn't work?
22:01:49kkurbjun_it's working well at the moment
22:02:21kkurbjun_but I was thinking about how to add in screen updates in a generic fashion without blocking
22:02:45kkurbjun_since we only have a 32 byte fifo, it's not nearly enough to fit the whole screen
22:03:04eigmaokay, then introduce an irq-driven TX queue
22:03:23eigmabutton_read_device() and remote_screen_draw() just queues packets to that queue
22:04:04kkurbjun_and remote_screen_draw would be a tick task also?
22:04:06eigmait could be a circular buffer large enough to fit the largest burst of screen updates we'll ever have.. my gut tells me 1K is plenty
22:04:13eigmaremote_screen_draw is just called whenever
22:05:13eigmatick -> remote_heartbeat_tick() -> uart_queue_tx_bytes()
22:05:28eigma(time of day changed) -> remote_screen_draw() -> uart_queue_tx_bytes()
22:05:37amiconnlinuxstb: ping
22:05:42eigmaUART TX Int -> uart_queue_tx_drain()
22:05:57kkurbjun_I see, what would the arguements be to the uart driver for uart_queue_tx_bytes?
22:06:09amiconnrasher: Do you have a better suggestion?
22:06:16eigmaconst char *buffer, size_t bytes??
22:06:41eigmaI get the feeling that was a trick question :)
22:06:56rasheramiconn: Not sure I do, no - perhaps a "shadow" in light-gray at the right-bottom?
22:06:56kkurbjun_:), somewhat
22:07:02eigmawhat was your point?
22:07:14amiconnrasher: That would make them look "hollow"
22:07:18 Quit Tanuva ("leaving")
22:07:30kkurbjun_well, I'm thinking for multiple commmands to be sent
22:07:36rasheramiconn: hrm... I'll think about it
22:07:42amiconnAlso, the plain grey background improves readability when a background image is loaded
22:08:06kkurbjun_and I guess we could just copy the data to the ring buffer though
22:08:12amiconnIt doesn't have to be grey, it could be configurable
22:08:30rasheramiconn: well I wouldn't want to use the background image for splash background ever
22:08:40rasherI agree that it should be a plain colour
22:08:54rasherperhaps it should be just another theme setting
22:09:17amiconnThe grey was introduced when we got greyscale display on the H1x0. The splashes looked a bit dull with the same white background as the rest of the screen
22:09:43kkurbjun_eigma: I think I am seeing how you are envisioning this, so we would effectively have two copies fo the remote screen buffer in memory, one in the ring buffer, and one for the rest of rockbox to draw to?
22:10:37 Join uki [0] (
22:10:51eigmathere would only be two copies for the (short) amount of time between the draw packet being queued and it going out on the UART
22:11:01mrkikoI'm on rockbox!
22:11:13eigmaI see your point now; we could have a periodic tick handler that updates the screen straight out of the framebuffe
22:11:19mrkikowowowo! It doesn't speak again, even if I put english.voice in .rockbox/langs
22:11:20ukihi, im here for the first time because i got a question
22:11:54kkurbjun_that could work also, but I think I like your solution better
22:11:57ukiare you going to create roxkbox for iriver clix? is it possible
22:12:05kkurbjun_a bit more memory, but simpler overall
22:12:25eigmaI think 2 KB vs 1 KB is negligible
22:12:26 Join Gnu47 [0] (
22:12:31BigBambiuki: If someone with the clix has the skills to do it it will happen. Ports are made by those who want them
22:12:43eigmaespecially since there's no potential for this to multiply 100x over
22:13:07BigBambiuki: The RockBox 'team' doesn't decide to do a port, interested owners do. interested bu my skills...
22:13:52ukibut thanks for the answer
22:13:56eigmakkurbjun: I think it's also more flexible when it comes to turning the BL on/off.. the respective event handlers will just have to queue a command somewhere and forget about it.
22:13:59desowinyou can collect information about hardware and put those on wiki
22:14:49 Quit mf0102 ("Verlassend")
22:15:05kkurbjun_Agreed, I'll work toward teh ring buffer concept, it keeps the proper priority of events well and eliminates some blocking scenerios
22:15:34 Quit XavierGr (Nick collision from services.)
22:15:35ukimaybe it will mobilize me to learn programming:)\
22:15:37 Join XavierGr [0] (
22:16:18mrkikoHow can I turn off my iRiver?
22:16:19BigBambiheh, if you learn enough to do a rockbox port then you will indeed have learnt programming :)
22:16:28BigBambimrkiko: Which iriver?
22:16:36 Quit nicktastic ("Leaving")
22:16:43ukiit will take long time...i whink ill buy sth that supports rockbox now :)
22:16:49 Quit amiconn (Nick collision from services.)
22:16:51BigBambiuki: A good idea
22:16:55 Join amiconn [0] (n=jens@rockbox/developer/amiconn)
22:17:16BigBambimrkiko: Hold stop (I think). This is in the manual
22:17:39ukiBigBambi what player u have got?
22:17:55BigBambiiriver H140 and gigabeat F60
22:18:07mrkikobut it doesn't work
22:18:46BigBambiSo you have checked which button in the manual, but it doesn't work?
22:18:59BigBambiAnd do the buttons work otherwise?
22:19:23BigBambihow long are you holding it?
22:19:54BigBambiwere you in fact, not holding it? :)
22:19:55mrkikoit seemed to crash repeating the same song tick for a little but now it's off
22:19:55eigmakkurbjun: I feel really bad that I havne't been writing any code for Rockbox lately :\ and next week isn't going to be any better, I have tons of schoolwork
22:20:06 Quit uki ()
22:20:10 Join uki [0] (
22:20:30scorche|weigma: welcome to the club :)
22:20:30BigBambiuki: If you are wondering which players etc., check out
22:20:41 Quit uki (Client Quit)
22:20:46 Join Zagor [0] (
22:20:57eigmascorche: I don't want to be in this club :(
22:21:05linuxstbamiconn: I'm around now.
22:21:07eigmascorche: undergrad? grad?
22:21:07mrkikomy player work so better
22:21:13mrkikoit' very fast and
22:21:25mrkikoi notice less disk spin ups/downs
22:21:37mrkikoNow I'll try to find out why it doesn't speak
22:21:47scorche|weigma: well, referring to the general "RL" sense
22:21:49BigBambimrkiko: Glad you like it. The manual is a fine resource.....
22:21:51mrkikoAniway surely a problem exist: when I start the player using my remote controller
22:21:58mrkikothe player turn itself off
22:22:06BigBambido you have hold on on the remote?
22:22:13mrkikopressing "play" directly on the player works
22:22:28 Join qweru [0] (
22:22:40mrkikoholds are off: the remote controller's one and the player's one
22:22:49BigBambiAh, do I remember something about the curent H300 bootloader release not supporting remote?
22:23:02BigBambiAnyone with an H300 like to comment?
22:23:04amiconnlinuxstb: pm
22:24:15mrkikoI'm very happy aniway with rockbox
22:24:48mrkikoMay be with the help of my sister I get it speaking also...
22:25:00mrkikoThe file structure is ok with the manual blessings
22:25:01BigBambimrkiko: Do you have a non-LCD remote?
22:25:18mrkikoin the sense that it doesn't have an lcd screen?
22:25:31BigBambiThe current release bootloader doesn't support start up with that type of remote
22:25:46BigBambiThe next release hopefully will
22:26:11mrkikoI may try it later...
22:26:25mrkikoaniway now I should be able to have "speech menu"
22:27:03BigBambiKeep an eye on the front page, when a new bootloader is released it will be announced. You will then need to repatch an .hex original firmware and reflash the rockbox bootloader
22:31:20mrkiko... I see
22:34:11mrkikoI don't remember where are saved settings...
22:34:17mrkikoI'll re-read the manual
22:35:29BigBambiA good idea
22:36:35 Quit Domonoky (Read error: 104 (Connection reset by peer))
22:39:02 Join NAPse [0] (
22:39:24NAPseDoes it exist any OSs to iPOD classic?
22:39:54desowinyes, the Apple one
22:40:19NAPseIs it good?
22:41:26scorche|wNAPse: what does this have to do with rockbox?
22:41:42 Quit ilgufo ("So Long, and Thanks For All the Fish -")
22:44:07 Part low_light
22:45:13NAPseWell I hoped that it might exist a rockbox witch is working at iPOD classic, but it seems to be wrong. So I hoped somone could tell me about anotherone.
22:45:40krazykitthere are no other ones
22:46:06 Quit hcs ("Leaving.")
22:47:19 Quit desowin ("use linux")
22:55:30 Nick parafin is now known as parafin|away (
23:00:25mrkikoWhat should I do to exit
23:00:35mrkikothe place where rockbox ask you what name to assign to a preset?
23:00:43mrkikoI can have spelled a table of symbols
23:00:48mrkikoabcdefgh... and so on
23:00:52mrkikoBut I can't exit there
23:02:42rashermrkiko: Read the manual..
23:04:43rasherAssuming there is a section on the virtual keyboard.. if not, take a wild guess, I suppose
23:05:29mrkikobut ehehe
23:05:36mrkikothe manual says to press play to save the changes
23:05:42mrkikobut the player simply ignores it
23:06:09mrkikoIs it possible rockbox has crashed?
23:06:16mrkikoand, if this happen, what should I do?
23:06:31BigBambido the other buttins still do anything?
23:07:31mrkikoeveneven a sighted user can't make the player react
23:07:49BigBambiIf the other buttons still do stuff, it hasn't crashed
23:07:56rasherMaybe the manual is wrong.. it's happened before
23:08:01 Quit atsea-22 (Read error: 104 (Connection reset by peer))
23:08:24BigBambiAt this point, I'm going to suggest pressing all the buttons to see what happens :)
23:08:57mrkikoactually nothing :)
23:09:37BigBambiSo none of them do anything? For instance does up down left and right move the cursor around the virtual keyboard?
23:09:37mrkikoI guess aniway the hardware engine (fm radio) would still go ahead for ours now...
23:09:53BigBambimrkiko: I don't understand
23:10:04mrkikomy sister sais the cursor is not visible
23:10:21BigBambiIf no buttons do anything at all, it may indeed have crashed. What player again?
23:10:30linuxstbA long press of the "on" key should quit the keyboard on the h300.
23:11:22mrkikono, the on key does nothing: I kept it pressed for 30 secs
23:11:55mrkikoAniway: even if it's crashed it would be nice to be able to power the playert off
23:12:02linuxstbThen it seems it's crashed. You'll need to press the reset button. I don't know where it is on the h340...
23:12:07BigBambiIf it really has crashed (although the H300 port is pretty stable), I think the H300 has a reset button on the bottom (although I don't own one myself)
23:12:23BigBambiI thin it is a little hole on the bottom where the power input is
23:12:39linuxstbOn the h140 you need something like a paper-clip to press it.
23:12:45BigBambiYou need to put something small in it like a paper clip
23:12:50***Saving seen data "./dancer.seen"
23:13:05mrkikoI'll search
23:14:52 Quit Lear ("ChatZilla [Firefox]")
23:15:24rasherThe manual should describe this
23:16:14PaulJamthe reset on h300 is on the left side below the microphone
23:16:19 Join atsea-22 [0] (i=atsea-@gateway/tor/x-badacca9aab3f736)
23:16:39rasher.. but it doesn't.
23:18:00mrkikoI resetted it and now the system works as normal...
23:18:07rasherOr at least, not in the description of the player in the Quick Start section
23:18:10mrkikoNow... I' not a developer
23:18:11BigBambiPaulJam: cheers
23:18:15n1sjust to be clear, short press on play is indeed accept, long play does nothing and stop is cancel in the keyboard on h300
23:18:31BigBambirasher: I just searched the entire pdf manual for reset and din't find anything relevent
23:18:35mrkikoI looked at the voice mpeg sequence structure. I guess the mp3 reader skips at the right offset when it needs to say something, right?
23:19:12mrkikoWen I kept pressed the right key for a while, probably I made rockbox go mad searching for the right offset of the letter to say...
23:19:20mrkikoAm I explaining myself?
23:20:04BigBambiYes, but I don't use voicing so I have no idea how well or otherwise various parts of it work
23:20:08rasherBigBambi: I think this info belongs in section 3.1.2 −− Turning the player on and off
23:20:16BigBambiYes, it would be helpful
23:20:35rasherAnd also in 3.1.1 for players with a dedicated reset button
23:21:35mrkikoYes: probably I can reproduce the crash this way...
23:22:19rashermrkiko: is this a crash in the virtual keyboard you're talking about?
23:22:28mrkikoand... another interesting thing to write on the manual is probably that
23:22:35pixelmamrkiko: at the right side there is a blank part you could use for spaces - it should wrap around once you reached the end
23:22:41mrkikorasher: yes
23:22:44BigBambimrkiko: If you can, have a look on the bug tracker to see if it already there ( and if not add it.
23:23:19pixelmamrkiko: if you meant the virtual keyboard as I thought
23:23:21mrkikooh... Ill do this tomorrow cause now I'm very tired...
23:23:24 Join MethoS- [0] (
23:23:28BigBambimrkiko: sure
23:23:36 Join webguest41 [0] (i=45f66aca@gateway/web/cgi-irc/
23:23:36mrkikoyes, I meant it
23:24:09webguest41Does anyone know how to get .WAD games for the rockbox?
23:24:19BigBambiwebguest41: The manual or the wiki page
23:24:34BigBambiBoth will tell you
23:24:36webguest41Wiki page? What's the link?
23:24:39 Quit MethoS- (Remote closed the connection)
23:24:42mrkikoAniway: I don't know if I said it, but when I use rockbox
23:24:55BigBambiwebguest41: Search the wiki for PluginDoom or such
23:25:05mrkikoI can't turn the player off with the "play" button on the remote controller.
23:25:19mrkikoThat's probably not a problem: I'll have a look at the sleep timer.
23:25:19webguest41Can you tell me whats the difference between DOOM and PACBOX or w/e?
23:25:34BigBambiwebguest41: The manual explains what they are, please read it
23:25:37mrkikoAnother important note: the Debian lenny/sid environment is not "sane" to build rockbox
23:25:54rashermrkiko: Oh? How so?
23:26:02mrkikothe ...
23:26:20rashersid works perfectly well - I can't imagine why lenny wouldn't
23:26:32krazykitlenny works just fine, i use it all the time
23:27:05*linuxstb wishes Debian would drop the silly names ;)
23:27:09mrkiko... may be it's my incompetence
23:27:11mrkikoI got:
23:27:12mrkikoNormal build selected
23:27:12mrkikoUsing source code root directory: /home/mrkiko/fw_iriver/svn/rockbox
23:27:12DBUGEnqueued KICK mrkiko
23:27:12mrkikoUsing m68k-elf-gcc 2.95.3-kgpd (295)
23:27:15mrkikoUsing m68k-elf-ld 2.1420030612
23:27:16Zagorlinuxstb: well just call it unstable or testing
23:27:17mrkikoCreated Makefile
23:27:21BigBambimrkiko: Please don't
23:27:24mrkikoand now... with make
23:27:24BigBambiUse pastebin
23:27:41rasherlinuxstb: we can use "unstable" or "testing" .. or "3.0" and "3.1" if you like?
23:27:43linuxstbZagor: I do... But then I get confused when other people call it by the name of a Toy Story character...
23:27:47rashernot for unstable then, of course
23:27:49krazykitlinuxstb, there are actually very good reasons for the silly names, but it's offtopic
23:27:50mrkikomake[1]: `rdf2binary' is up to date.
23:27:51mrkikomake[1]: `convbdf' is up to date.
23:27:51mrkikomake[1]: `codepages' is up to date.
23:27:51mrkikomake[1]: `scramble' is up to date.
23:27:51***Alert Mode level 1
23:27:51mrkikomake[1]: `descramble' is up to date.
23:27:53mrkikomake[1]: `mkboot' is up to date.
23:27:55mrkikomake[1]: `bmp2rb' is up to date.
23:27:58mrkikocpp0: config.h: No such file or directory
23:27:59BigBambimrkiko: Please stop it
23:28:00mrkikoCreate sysfont.h
23:28:03mrkikoHeader parsed
23:28:06mrkikocpp0: config.h: No such file or directory
23:28:10mrkikoCC sysfont.c
23:28:12Mode"#rockbox +o Zagor " by ChanServ (ChanServ@services.)
23:28:13mrkikoHeader parsed
23:28:14BigBambimrkiko: STOP NOW
23:28:15mrkikocc1: Invalid option `5206e'
23:28:18mrkikocc1: Invalid option `strict-align'
23:28:18Kick(#rockbox mrkiko :Zagor) by Zagor!n=bjst@rockbox/developer/Zagor
23:28:18***Alert Mode level 2
23:29:00 Quit webguest41 ("CGI:IRC (EOF)")
23:29:38n1swow his gcc is ooooold
23:30:55 Join mrkiko [0] (
23:30:59mrkikoSorry all...
23:32:08Zagormrkiko: your compiler is very old
23:33:19mrkikoInfact, that's the problem
23:33:41mrkikoThe probem is that using apt-file also I can't find packages offering that compiler
23:33:50BigBambimrkiko: did you run to build the cross compilers?
23:34:03BigBambiThat will do it all for you automatically
23:34:51n1sbut you probably want to uninstall those old m68k-elf packages first
23:35:23mrkikoI knwew about that script
23:36:10mrkikoIn general I read the manuals and do right things, but today is not going this way. I'll try to do this tomorrow...
23:36:20mrkikothank you all...
23:36:36BigBambimrkiko: Don't worry :)
23:36:41mrkikoAnd sorry for misrespecting channel regulations
23:36:57Mode"#rockbox -o Zagor " by Zagor (n=bjst@rockbox/developer/Zagor)
23:37:13mrkikoAniway I red that
23:37:33mrkikothere is a task (assigned to no-one) for adding espeak plugin to rockbox
23:37:36mrkikoVery nice idea...
23:37:44mrkikoGood night!
23:37:52 Part mrkiko ("rockbox")
23:38:19***Alert Mode OFF
23:41:18Zagorbluebrother: files uploaded
23:41:28 Quit NAPse ()
23:42:32 Quit midgey ()
23:43:12CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
23:43:12*keanu wonders how USB's coming
23:43:54 Join midgey [0] (
23:44:09rasherWow, where on earth do you even find gcc 2.95 this day and age
23:44:15Zagorkeanu: I've hit a snag. the first bulk transmit works great, but the second becomes all-zero and the rest are ... strange
23:44:41preglowmoos: here?
23:44:48Zagorrasher: I think there once was a m68k cross-compiler package in debian. I guess that's what he dug up.
23:44:50keanuZagor, ok, good luck on getting it working
23:45:27bertrikZagor, maybe first you should get the base USB transfers rock solid before trying to pile the MSC stack on top
23:45:38moospreglow: hello
23:45:45Zagorbertrik: base=setup/control ?
23:45:50preglowmoos: did you even find out what was wrong with your musepack?
23:45:52Zagorthey are already solid
23:46:11Zagormsc is simply the first test case for bulk transfers
23:46:18bertrikI mean, write a simple USB bulk echo or something like that
23:46:47bertrikusing libusb on the hostside
23:47:22Zagoryeah maybe. but msc is very simple so it's not like it's adding a lot of complexity anyway.
23:48:05moospreglow: nope, I fall to stop my musepack uses on my gigabeat. I tested again with all dsp stuff off, but nothing
23:48:08bertrikBOT is more complex than it seems, mostly because of all the different cases where host and device disagree on the direction/amount of data to be transferred
23:48:32preglowmoos: have you tried test_codec and wav writing?
23:48:48Zagorbertrik: ah, the famous 13 cases :)
23:49:00bluebrotherZagor: thanks.
23:49:11 Join Pyromancer [0] (
23:49:35Zagorbertrik: your point is very valid. I'll do that.
23:49:43moospreglow: I didn't, but I probably will, cause that's very strange that no one had never complain about. Mybe defected chip. Who knows...
23:50:04preglowmoos: i doubt it, but we already have gigabeat trouble with vorbis
23:50:12preglowmoos: a test_codec run with wav writing would be very helpful
23:50:14 Part przemhb
23:50:24moospreglow: I'll try this week-end and let you know the results
23:50:24preglowmoos: do all files sound shit, or just some?
23:50:40 Join safetydan [0] (i=dc9d468b@rockbox/developer/safetydan)
23:50:46moosall unfortunatly, all that I tested at least
23:50:58preglowany gigabeat f people here that can be bothered to do some testing?
23:51:11bertrikgood luck!
23:51:19 Quit bertrik ("bye")
23:51:26moosIIRC Nico_P tested with no troubles
23:51:27krazykitpreglow, sure
23:51:41preglowkrazykit: do you have test_codec plugin?
23:52:02Nico_PI sent some parts of the WAV output to buschel and he found them to be fine
23:52:18preglowkrazykit: you pretty much only do if you make your own builds and enabled it
23:52:27krazykitpreglow, no, but i can enable it if you tell me how ;-)
23:52:36krazykiti have build environment sitting right here
23:52:40moosthat's maybe my gigabeat that sucks ???
23:52:42preglowkrazykit: just put "test_codec.c" in plugins/SOURCES
23:52:49preglowmoos: i'm about to find out
23:52:54moostoffe82 :P
23:52:59 Quit alienbiker99 ("( :: NoNameScript 4.02 :: )")
23:53:28preglowmoos: but you have tried other codecs, right, and they sound good?
23:54:07moosyup, I encountered probles just with my mpc files, and since the changes I pointed to you
23:54:24toffe82moos: yes :)
23:54:46mooshey toffe82 :)
23:54:46markunpreglow: what kind of trouble is there with vorbis on the gigabeat?
23:54:59markunhi moos :)
23:55:13toffe82moos: long time no see
23:55:13preglowmarkun: the latest tomal optimizations seem to bug out on gigabeat, don't they?
23:55:17preglowmarkun: something in imdct
23:55:43toffe82moos: you put a new hd on your gigabeat ?
23:55:53markunmaybe my build is too old, but I pretty much use vorbis exclusively and didn't notice anything
23:56:09 Join MethoS- [0] (
23:56:23moostoffe82: not yet
23:56:26 Join lazka [0] (
23:57:01 Quit homielowe (Read error: 110 (Connection timed out))
23:57:06krazykitpreglow, hold on, i got a build error. i'm supposed to add the test_codec.c to apps/plugins/SOURCES?
23:57:08preglowthat patch was closed, then reopened because of gigabeat trouble, afaik
23:57:13preglowkrazykit: myeap
23:57:27krazykitoh bugger, i typo'd it
23:57:33preglowexplains that, then :P
23:57:39pixelmaIIRC tomal's changes were excluded from the gigabeat or something like that
23:57:52*preglow has too few targets :/
23:57:57markunpreglow: ah, it's the code I turned off instead of fixing I think :)
23:58:06preglowmarkun: i do believe so, yeah :P
23:58:12preglowmarkun: do you remember the symptoms?
23:58:23preglowi would very much like to see that resolved in a better way
23:58:33 Join Rob222241 [0] (
23:58:35preglowit'll just rot in the tracker if not
23:58:46markunthe problem was that the gigabeat doesn't have .icode so we should use .text
23:58:58preglowwas that the entire problem?

Previous day | Next day