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

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

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

#rockbox log for 2005-09-20

00:05:42 Quit noC|andY`fRa (Read error: 104 (Connection reset by peer))
00:07:58rasherHrm, the tables are not quite what they used to
00:08:09Zagorhow so?
00:08:20rasherThey're much darker now
00:08:35rasherHm.. maybe
00:08:45rasherMaybe it's just because of missing styles
00:08:47rashernevermind me
00:08:49Zagordarker? they are transparent now.
00:09:01Zagoror do you mean the header row?
00:09:09rasherSorry, I meant the borders
00:09:15Zagorah"> vs
00:09:57Zagorright, the borders are slightly thicker now.
00:12:07***Saving seen data "./dancer.seen"
00:28:25preglowlooks all good to me
00:29:00preglowsolved the registration problem?
00:29:59Zagornot yet
00:30:30*rasher chews on another page
00:34:22 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
00:34:36*preglow nominates rasher for Dane of the Year
00:39:59 Join ashridah [0] (
00:44:22rasherwhen's the award show?
00:45:03preglowdrammensveien 55, oslo, next weeken!
00:45:07preglowweekend, even
00:45:34preglowi'm actually going to denmark next weekend, we'll just do the awards ceremony there
00:46:51 Join edx [0] (
00:50:42 Quit markun ()
00:51:39Zagorregistration works now
00:52:54 Quit paugh ("Leaving")
00:55:57 Quit ansivirus_ (Read error: 110 (Connection timed out))
00:56:09 Join bagawk [0] (
00:56:17 Join ansivirus_ [0] (
00:58:03preglowyou didn't lose any of the mailing list info?
01:00:36rasherZagor: how about the attachment weirdness?
01:01:41Zagori'm looking at it but can't see what it is. looks like some incompatibility with the new twiki version. do we have any page where it works?
01:02:02rasherErp.. that's not good
01:02:15rasherI could upload something to a test-page?
01:02:24Zagorplease do
01:03:43rasherso now the ReleaseTodo page has an attachment
01:03:57rasherwhich.. doens't show up either
01:04:56rashernot releasetodo
01:04:57rasherhang on
01:05:14rasher still nothing though
01:06:03Zagorvery strange. i'll see what the code does.
01:10:14bagawkZagor, Just curious, why did the mail list go down if rm -rf / was run, and the only permissions where for the wiki?
01:10:44rasherIt was for the apache-user. Still curious though
01:11:00rasherwhy the mailing list would be deletable by www-data is weird
01:11:01Zagorlots of stuff was group-writable by this user, including the mailing list configs
01:11:21bagawkI see
01:11:55Zagorthey have to be for the web interface to work, iirc
01:12:56 Nick ansivirus_ is now known as ansivirus (
01:17:33preglowisn't database stuff usually owned by dbuser:dbuser, btw?
01:17:48preglowor doesn't twiki store stuff in a db exclusively?
01:18:03Zagorno, each topic is an rcs file
01:18:31Zagorlookie, attachments!
01:18:41preglowand there was much rejoicing
01:18:54Zagorthere was indeed a change in the handling in this version
01:21:40 Quit DangerousDan ("Miranda IM! Smaller, Faster, Easier.")
01:23:57 Quit ashridah ("Leaving")
01:25:54linuxstbIs the following line of C correct?
01:25:55linuxstbif (inl(0x2000) == "gfCS") { ...
01:27:04linuxstbDidn't think so.
01:27:52rasher57 pages left..
01:28:11*linuxstb needs to visit #ipodlinux
01:28:26 Join Febs [0] (
01:30:40 Quit ender` (No route to host)
01:31:12Zagorlinuxstb: was that from ipodlinux code?
01:31:31rasherI'm guessing yes..
01:31:39linuxstbYes - the most recent CVS commit:
01:32:04preglowyou'd need some pretty nifty magic for that code to work
01:32:12preglowsince it compares two pointers...
01:33:34ZagorI assumed inl() returned a long
01:33:49Zagor...which only makes it more weird
01:33:49linuxstbMe too.
01:35:53preglowwell, you still compare something to an arbitrary pointer
01:35:58preglowwhich is pretty weird
01:37:18linuxstbChecking the IRC logs, someone on #ipodlinux spotted it as well. But it hasn't been fixed yet.
01:37:30preglowi wonder what the hell that line is supposed to do
01:37:47preglowwhoever wrote it had to be seriously fscked up on something
01:37:52rasherIt looks like completely black magic to me
01:38:06linuxstbrasher: That's ipodlinux :)
01:38:31Zagorlikely address 0x2000 contains the longword equivalent of ascii "gfCS" on that particular target
01:38:46linuxstbZagor: Yes, I think so.
01:39:01rasherIt just seems so weird
01:39:07linuxstbinl is just a macro that reads from memory.
01:39:08preglowand... stupid...
01:39:10rasherand not even a comment..
01:39:32Zagorcommitting code without even trying to compile it is kinda... icky too :-)
01:39:39preglowit will compile
01:39:39rasher /* DANGER, HERE BE MAGIC! */
01:39:41 Quit Moos (Read error: 110 (Connection timed out))
01:39:42preglowno problem, perhaps a type error
01:39:50preglowwarning, at best
01:39:59 Quit xen` (Read error: 110 (Connection timed out))
01:40:13linuxstbYes, gcc gives the expected "tools.c:39: warning: comparison between pointer and integer"
01:40:34preglowbut hell
01:40:41preglowpeople are so intent on ignoring warnings, it's no surprise
01:40:59preglowfixing a warning with a #pragma is accepted practice with a lot of people :/
01:46:46rasherZagor: Now I'm locked out :-X
01:47:05Zagorhang on..
01:47:39rasherI'm in!
01:47:50Zagorslight mistake :-)
01:54:19*linuxstb finally has some iPod success.
01:55:00preglowelaborate :>
01:55:41linuxstbI simply got the ipodlinux bootloader to run the Apple firmware.
01:56:39linuxstbI'm just trying to work out exactly how much of my ipod ipodlinux supports.
02:11:30rasherJust start porting rockbox already
02:11:52rasherWe're WAITING
02:12:09***Saving seen data "./dancer.seen"
02:14:08linuxstbPatience. Maybe I should have bought myself an ipod that ipodlinux actually works on.
02:15:04preglowshould we put all the audio modifier stuff in the dsp layer or the pcmbuf layer of rockbox?
02:15:26preglowthe dsp layer is the most logical place, but then there'll be a latency whenever you change a parameter, like mono processing to karaoke mode
02:20:26 Quit bagawk ("Leaving")
02:20:32Zagori'm off to bed. good night guys.
02:22:06 Quit tvelocity ("Leaving")
02:24:13preglowperhaps a realtime mode, where we run at full cpu_boost all the time and with as little latency as possible, for use with such things as eq adjustment and pitch adjustment is possbiel
02:24:26preglowSlasheri: please come back to me on this if you read ;)
02:25:10rasherlinuxstb: there's your cue
02:25:59preglowit'll be a pain for crossfading mode, of course, where you might have to risk dealing with 10 seconds of buffered pcm when you want to adjust the audio in some way
02:27:43rasherI just saw why running Linux on an ipod is silly.
02:27:52rasher"video playing has part of it in the kernel for various reasons"
02:28:02rasherthat's so wrong
02:28:45preglowwell, gotta bed
02:28:53 Quit preglow ("feaches")
02:43:35 Join Strath [0] (
02:45:25 Quit dpassen ()
03:09:17Febsrasher, are you around?
03:09:34rasherbarely, but yeah
03:10:12FebsWhat needs to be done to restore the wiki pages that you've cached?
03:10:32FebsDoes all formatting etc. need to be reapplied?
03:10:35rasherCompare the current version to the cached html version, update what's changed in between
03:10:46rasherthis pretty much means "write any changes in wiki syntax"
03:10:52rasherit's not a fun job..
03:11:02FebsNo, I imagine not.
03:11:18FebsI was afraid that's what you would say.
03:11:25rasherSome are easier than others of course
03:11:31rashersome have just had a line added/removed
03:12:46FebsI'm specifically looking right now at this page:">
03:13:12FebsI take it that I have to reapply all of the wiki syntax for hyperlinks, tables, headings, etc.?
03:13:30rasherAh yes, you do indeed
03:13:59rasherRewrite from scratch :-\
03:14:13FebsWell, it could be worse. At least the text is preserved.
03:15:24rasherYes. It'll just take some time
03:16:55FebsOK. I'll work on restoring some of the documentation pages that I wrote.
03:16:57rasherJust prod me if you fix any pages, and I'll remove them from the list
03:17:01FebsI will.
03:42:11 Nick Sucka`away is now known as Sucka (
03:47:11 Join pike [0] (
04:00:55 Join DarkShadow [0] (
04:05:46 Join QT [0] (i=as@madwifi/users/area51)
04:07:29Febsrasher, ManualRockboxInstall has been restored.
04:12:12***Saving seen data "./dancer.seen"
04:12:34DarkShadowWhere do you get rockboy files? Or are they some emulator file things that you have to own the game to have?
04:17:01 Quit QT_ (Read error: 113 (No route to host))
04:25:28 Quit Sucka (Read error: 110 (Connection timed out))
04:26:01 Join Sucka [0] (n=NNSCRIPT@
04:30:36 Quit cYmen ("zZz")
04:35:35 Quit DarkShadow ("CGI:IRC")
05:03:17 Quit Sucka ("a bird in the bush is worth two in your house")
05:18:22 Quit JoeBorn (Nick collision from services.)
05:19:07 Join jborn_ [0] (
06:12:13***Saving seen data "./dancer.seen"
06:34:13 Join adiamas [0] (
06:35:26 Quit adiamas (Client Quit)
06:37:04 Join adiamas [0] (
06:41:40 Quit adiamas ("Chatzilla 0.9.68a [Firefox 1.0.6/20050716]")
06:52:13 Join LinusN [0] (
06:56:07 Quit Maxime (Read error: 104 (Connection reset by peer))
06:56:31 Join Maxime [0] (
07:25:17Bgermorning :)
07:27:58Bgerrasher r u here ?
08:05:17 Join ender` [0] (i=ychat@
08:08:08 Join B4gder [0] (
08:12:14***Saving seen data "./dancer.seen"
08:12:22Bgermorning, B4gder
08:12:40Bgerwiki is up, afaics
08:13:57Bgeri'm waiting for rasher to tell me which pages are already changed with last ones
08:14:29B4gdersounds like a fair approach
08:15:29*B4gder works on manually unsubscribing people
08:22:01LinusNpeople really can't read...
08:23:59BgerLinusN what do u mean ?
08:24:38LinusNthey can't follow directions, they try to send the "unsubscribe" messages to the list instead of the list-admin
08:28:28B4gder410 subscribers now
08:40:24 Join einhirn [0] (
08:41:39 Join linuxstb_ [0] (
08:42:52 Quit linuxstb (Read error: 104 (Connection reset by peer))
08:42:57BgerLinusN about the forensic analysis tool ... what did u use ?
08:43:27 Quit linuxstb_ (Client Quit)
08:44:46LinusNBger: sleuthkit and foremost
08:45:08Bgerhm, i'll take a look at these, it's a nice to know thing ...
08:48:16LinusNi'll put up some links
08:50:33Bgeri suppose there are such tools for reiserfs ?
08:51:47LinusNi dunno
08:53:31 Join linuxstb [0] (
09:05:48B4gderI hate stupid mailers
09:06:12B4gder"Your mail to the following recipients could not be delivered because they are not accepting mail from"
09:06:33B4gderfrom a user on the mailing list
09:06:47B4gderdo I need to say ? ;-/
09:12:36 Join linuxstb_ [0] (
09:12:36BgerB4gder : u've just added more spam to your mail ..:)
09:12:36B4gderI don't care
09:12:36Bgeri see
09:12:36B4gderI get sooooooo much already
09:12:36BgerLinusN what about pdf files ?
09:12:46Bgergood :)
09:22:33 Quit linuxstb (Read error: 110 (Connection timed out))
09:27:41linuxstb_Mmm. It appears that the latest color ipods (like mine) have a PP5022 in them, not a PP5020.
09:28:50B4gderprobably to support the lcd
09:32:37linuxstb_It's reported as "register compatible" with the PP5020 and there doesn't seem to be any ipodlinux specific code for the 5022 (just 5002 and 5020), so I don't think it's much of a problem.
09:34:11Bgerbut ipodlinux doesn't support ipod photo, or i'm wrong ?
09:34:29linuxstb_Bger: It's "unsupported", but it works.
09:34:50B4gdermost models seem "unsupported" but working
09:35:01linuxstb_Current status of ipodlinux is here:
09:35:01Bgerso, u've got your ipod
09:35:19linuxstb_Bger: Yes, but unfortunately it's one of the ones least supported by ipodlinux.
09:37:23linuxstb_The only models which are unusable are the shuffle, Nano and the very latest revision of the color iPods. But I think it's just an LCD problem for the color iPods - at least, no-one has mentioned any other issues.
09:42:28 Quit rasher (Remote closed the connection)
09:47:38B4gdernow, what is the release waiting for really?
09:49:30B4gderits been 5 days since the latest archos related fix
09:57:19LinusNB4gder: amiconn has worked his ass off to find the recording bit-shift problem
09:57:49B4gderI know
09:58:04B4gderbut how long will that take and do we really need it fixed before the release?
09:58:24B4gderisn't this a bug we've had like forever?
10:03:03HClsounds like extra important to fix it
10:03:10linuxstb_I say lets give amiconn the decision - release now or wait for him.
10:03:25LinusNrecovered pdf files online as well
10:06:12Bgerhm, most of them must have been in the backup
10:06:36linuxstb_LinusN: What about other filetypes - e.g. .zip or .exe (assuming there were any).
10:07:49B4gdertalk about doing rubbish, or is it making something out of rubbish? ;-)
10:12:06LinusNbmp images up
10:12:17***Saving seen data "./dancer.seen"
10:12:31LinusNthe recover tool must have problems with monocrome bitmaps
10:12:43B4gderI must go look for my cube gifs
10:13:17LinusNthey are there
10:13:23B4gderthey are?
10:13:39B4gderthey're not on the gif page
10:13:40LinusNhmm, no :-(
10:14:53B4gderI think I still have them around
10:24:46 Quit ze (No route to host)
10:36:49 Join Dma-Sc [0] (
10:39:14 Join amiconn_ [0] (
10:43:00 Quit linuxstb_ ("Leaving")
10:57:23LinusNzip files now online
10:57:38 Quit amiconn (Read error: 110 (Connection timed out))
10:57:39 Nick amiconn_ is now known as amiconn (
11:01:35Zagori went through the zips yesterday, renaming some to more useful names
11:01:59Zagorgifs and pngs too
11:02:31LinusNsaw that
11:05:07Zagoryeah, found it
11:05:21LinusNdid you move the png's?
11:05:31LinusNi didn't see any renamed png's
11:06:01Zagormaybe I only looked at them
11:07:41Zagornice work on the pdfs. which tool did you use to snapshot the first page?
11:10:17LinusNimagemagick, of course
11:10:32LinusNwonderful tool
11:10:53Zagoraha. I didn't know it reads pdf too.
11:11:08LinusNconvert xxx.pdf[0] -thumbnail 256x256 thumb_xxx.png
11:11:28LinusNi dodn't know it either, i just tried and it worked :-)
11:11:29SlasheriLinusN: btw, if you would like to have some remote rsync backup space (few gigabytes for example), i could provide that with 100
11:11:37Slasheri+ 100 Mbit/s connection
11:11:40LinusNit uses ghostscript for the pdf parsing
11:11:43Slasheriargh, weird enter..
11:11:56LinusNSlasheri: thanks, but no thanks
11:12:09LinusNwe will make our own backup work
11:12:21LinusNand we only have a 2mbit/s connection
11:12:46LinusNshared with a lot of other sites
11:28:18HCli don't trust the output of my perl converter script at all.
11:28:25HClwho's good at perl?
11:30:22Zagorwhat's the problem?
11:31:53HCli wish i knew, heh.
11:32:17HClits supposed to upgrade an runtime database v1 to runtime database v2
11:33:17HClmaybe its just cause i forgot to close...
11:33:54HClor not
11:35:17HClforgetting to close was just one bug
11:36:52B4gderwhy add the "type" in the runtimedb?
11:37:05HClto support empty hashes, and filehash upgrades
11:37:36B4gderwhy is that related to perl/java?
11:38:04HCli figured it'd make sense to assign the perl algorithm a different number than the java algorithm since iirc they do not produce the same values
11:38:18LinusNthey should
11:38:42B4gdercrc32 on 32K data should be the same
11:39:45B4gderwhat happens when we support the C coded version?
11:39:50HClthe java version is attempting to be smart by skipping the id3 tag in mp3 files
11:39:54B4gderor if someone else decides to write one?
11:40:07B4gderHCl: and so does the perl version
11:40:18HClthey do not produce the same results though.
11:40:32HClbecause the backends differ in quality of figuring out the id3 tag.
11:40:57HCli don't really see why they should be the same anyways.
11:41:11B4gderbecause you lock the user into a single db
11:41:16B4gderdb tool
11:41:57B4gderwhat's the hash for?
11:42:03HCli don't see how we're doing that, at least, not with the hashtype upgrade.
11:42:06B4gderif the hash can differ, then what's the point?
11:42:21B4gderright, you need to run a conversion tool if you decide to build a new db
11:42:24HClthats one of the reasons why the hashtype field is getting added.
11:42:36B4gderthen what type do I use when I write my new db tool?
11:42:47HCldepends on how compatible it is with the rest.
11:43:05HClif its 100% compatible with say, perl, then the perl type.
11:43:08B4gderI say it is the wrong way out of this problem
11:43:48HClthen propose a better one :)
11:43:57HClwhile at the same time solving the 0 crc problem
11:43:59B4gdersure - make them produce the same crc
11:44:06LinusNmake the two tools produce the same hash
11:44:10B4gderthe 0 crc is not depending on what version you use
11:44:14HCland exactly how do you propose to do that :)
11:44:19B4gderHCl: debugging?
11:44:21HCland exactly how do you propose future hash upgrades? :)
11:44:37B4gderyou mean when you change hash algo?
11:44:38HClin the future, we'll want to hash on the music data
11:44:56LinusNwhat do we hash now?
11:45:03B4gderthen you change db version
11:45:04HClthe first 32kb, including id3 tags.
11:45:16HCland then destroy everyone's runtime database?
11:45:18HCli'd think not.
11:45:20 Join ashridah [0] (
11:45:26LinusNthen how can the two tools produce different hashes?
11:45:29B4gderno, and convert it to the new
11:46:04B4gderthe perl version already checksums the audio data
11:46:06HClwell, be my guest in either altering the perl or the java version in producing the equivalent of the other
11:46:12HClin the meanwhile
11:46:19HCli'll just use my own approach on my local tree.
11:46:29LinusNHCl: how cooperative
11:46:40HClsorry, but i'm not going to fix the java or perl version
11:46:43HClinto generating the same hash.
11:46:50HClbecause i hate perl
11:46:56HCland the java one relies on a backend
11:47:03HCland though its completely open source with the source available
11:47:05B4gderso you want to bury the bugs by creating an additional tool layer?
11:47:12HCli am not going to figure out how it works, black box principle and such.
11:47:40B4gdersounds insane in my ears
11:47:42HCli personally think its better to support different type of hashes in the database itself
11:48:10B4gderpossibly, but these should not be different types
11:48:14LinusNHCl: because you don't want to fix bugs?
11:48:18B4gderwe have a defined way
11:48:21B4gdera defined format
11:48:29HClLinusN: because i want to support upgrading hash algorithms
11:48:29B4gdereither we follow the format, or we don't
11:48:45HClyou can change the java and perl version back to hashing the first 32k if you want
11:48:50HClthey'd be equivalent then.
11:49:01HClbut worse, obviously.
11:49:17B4gderthen I say we dump the java way and we won't have this problem
11:49:34HClheh, good luck
11:49:47HCli'm not going to write a perl version to generate the database
11:49:59B4gderI already did
11:50:00HCland the java version is already superior to the perl version
11:50:15HClLinusN: its id3 reading backend is much better.
11:50:24HClthe perl version generates crap pYi things
11:50:27HClin the database
11:50:33LinusNoh, a bug
11:50:34B4gderbut we fix bugs in the perl version
11:50:39B4gderyou refuse to do so in the java one
11:50:53B4gderyou want to hide them instead
11:51:05HClthe java version doesn't do that 
11:51:05HCli 'm not going to fix bugs in code that other people wrote and isn't transparaent, no.
11:51:08HCli don't want to hide them.
11:51:24HCli want to upgrade the hashing algorithm to rely on music data
11:51:28HClthat, yes. i'm lagging tons
11:51:40B4gderwe already checksum music data
11:51:49HClbut not consistently.
11:51:55B4gderthen that's a bug
11:52:10HClthe difference between perl and java
11:52:15HClis *not* the reason for the database update
11:52:31HClthe reason is to support database hash upgrades
11:52:47LinusNand to support "no hash" instead of crc=0
11:52:50HClwithout having to bump the runtime database version all time
11:53:44B4gderdo we really need to do hash upgrades?
11:54:11B4gderI can see how we should support CRC == 0, but not why we'd need to change hash
11:54:27HClunless we have a perfect hashing algorithm that hashes on music content, i can see how we'll have to upgrade it a few times
11:54:30HClon top of that.
11:54:39HClsince we need to support crc == 0 in the first place
11:54:47HClwe have an extra field in the database anyways
11:54:50HClthat we can use for hashtypes
11:55:07HCland thats simply 0 for no hash.
11:55:47 Join tvelocity [0] (
11:55:56B4gderand when used, it is "crc32 on music"
11:56:06B4gderon 32K of music even
11:56:19HClwe don't have that reliably implement.
11:56:22LinusNok, so the perl code has trouble finding the music content
11:56:27HClboth the perl and the java versions
11:56:32HClhave trouble finding the music content
11:56:36HClin certain files
11:56:44B4gderthey may have bugs
11:56:52HClin some situations the perl version does better finding the music content
11:56:54B4gderbut the perl version does as described
11:57:01HCland in some situations the java version does better at finding the music content
11:57:12HCland the perl version has id3 tag bugs
11:57:15HClthat the java version doesn't have
11:57:35Zagorwouldn't the simplest way be to always read the 32KB from the middle of the file? the beginning is hardly the most unique part.
11:57:42LinusNthen i suggest we either 1) fix the content-finding bugs or 2) read the hash data from 128k into the file
11:57:44Zagorand there's not many tags in the middle
11:57:49B4gderthink sounds like a reason to reconsider the crc idea
11:57:55B4gderthis sounds
11:58:03HClwe already discussed that zagor, the middle changes because the id3v2 tag data is at the front
11:58:05B4gderI like the 128K idea
11:58:10HClthe same goes for 128k
11:58:12HClnot reliable
11:58:17HClthe amount of data in the front changes
11:58:22HClwhen id3v2 tags are changed
11:58:25LinusNah yes
11:58:32Bgermiddle 128k ?
11:58:38Zagorok, we want to be "tag agnostic"
11:58:44B4gderand tool agnostic
11:59:03Bgermiddle of frame data
11:59:05LinusNso we should fix the bugs then
11:59:20 Quit B4gder ("Lämnar")
12:00:16LinusNso both tools find the same music content
12:00:31HCli agree
12:00:55HClfor the java tool, this probably means writing some code to properly find the location of music data in an mp3 / ogg
12:01:05HCli've been wanting to get rid of the icky backend that does it for it now
12:01:20HCli also need samplelength and playlength though
12:01:36HClwhich is annoyingly tricky to get
12:02:16HCli gotta go
12:12:19***Saving seen data "./dancer.seen"
12:13:30LinusNgotta go too
12:13:33 Part LinusN
12:17:07 Join linuxstb [0] (
12:18:29linuxstbTalking about the tag database, we now have a reasonably good implementation of file/tag parsers in apps/metadata.c Why don't we just use that and convert the generator to C?
12:19:59linuxstbAlso, FLAC and (I think) wavpack already have an md5sum of the entire uncompressed PCM data stored in the metadata - could that be used as a hash?
12:20:52Zagorif reliable, then sure
12:24:15linuxstbIf you're talking about the md5sum, then AFAIK it's part of the FLAC spec, so must be correct in a valid FLAC file. I would expect wavpack to be the same.
12:24:47linuxstbBut it's 128-bit - I'm guessing that will be an issue.
12:24:53 Quit tvelocity ("Leaving")
12:58:21 Join amiconn_ [0] (
12:59:17Bgerhi, amiconn
13:04:15linuxstbHow are your recording tests going?
13:05:25amiconn_Not too well :/
13:05:58 Quit amiconn_ ("CGI:IRC (EOF)")
13:06:48 Join webguest83 [0] (
13:10:10 Join preglow [0] (
13:23:24 Join rasher [0] (
13:23:57Bgerrasher ?
13:24:32Bgerwhat pages are already updated?
13:24:37Bgerin the wiki
13:24:50rasherAnything that isn't on my page
13:25:17rasherWell, and a few more, I've not updated it since last night
13:29:55rasherCheck the page now">
13:34:07preglowanyone got any clever ideas on how to handle audio modifications without heaps of latency on rockbox?
13:37:31Bgerthe changelog is BIG...
13:38:04ZagorI think we need a "digested" version for the average user
13:38:38linuxstbpreglow: What kinds of audio modifications are you talking about?
13:38:55rasherZagor: That's what I intended the releasenotes "what's new" section to be
13:38:59rasherI should update these.. hang on
13:39:35preglowlinuxstb: eq, replaygain, stereo width, etc
13:39:57linuxstbJust normal playback processing then?
13:40:34preglowlinuxstb: when changing a parameter and having the processing in dsp.c, you might have to wait the entire length of the audio buffer (which can be VERY long for crossfading modes) before you can hear the result of your change
13:41:02linuxstbOK, I understand the problem.
13:41:16webguest83hi - i have a problem with recordings - the same as in bug [ 1152291 ] - has there been a bugfix for that?
13:41:36preglowthe other option is applying the processing right before the audio is sent to the dac, but then it's already been reduced to sixteen bits, and a lot of precision will vanish in subsequent processing
13:42:04linuxstbpreglow: And the audio in the buffer could be from different tracks - so different resampling and replaygain would be needed.
13:42:29preglownever thought about that
13:43:27Bgerrasher : how is the "checkmart" in wiki lang ?
13:45:27preglowmight actually need to buffer the data in 32 bit format and apply dsp processing at the other end of the buffer :/
13:46:27linuxstbIs it common to change the DSP parameters during playback? Can't we just flush the buffer?
13:47:38preglowwell, that's what i thought about as well
13:47:55preglowand that might fail for a slow codec
13:48:07preglowespecially if other stuff is going on, like loading
13:48:36preglowchanging a parameter sometimes requires a bit of extra prccessing as well, like for eqs
13:48:58linuxstbI don't know if this affects what you are thinking about now, but I would like Rockbox to be able to output PCM data at the native samplerate on targets that allow it (such as ipod).
13:49:41linuxstbObviously if cross-fade was enabled, we would probably have to resample everything.
13:50:48preglowi think rockbox should be able to do this now as well, for those sample rates where it's possible for spdif out, or for all frequencies when spdif out is disabled
13:51:06 Join muesli- [0] (
13:51:34linuxstbSo does the iriver spdif out support different samplerates?
13:52:49preglowyeah, but not certain which of them are actually possible
13:53:32BgerChangeLog->General Section ok :)
13:53:40preglowi know of 48000, 44100 and 32000
13:53:41preglowmight be more
13:55:11rasherBger: don't bother with it - I have a completely up to date version of it lying arount
13:55:24Bgereh :(
13:55:39rashersorry I missed that
13:56:09Bgerrasher : so, where to start from ?
13:56:28BgerHowtoUpdateLangfile ?
13:56:56rasherthat's fine.. don't bother fiddling with the list at the bottom
13:57:03rasherif the rest is correct, don't touch it
13:57:11rasherI have a script that outputs the list at the bottom
13:57:45Bgeri didn't understand what do you want to say ...
13:57:47rasheryou're picking all the wrong ones!
13:58:14Bgerok, give me topics to work on
13:58:34rasherWell.. anything but ChangeLog and HowtoUpdateLangfile really
13:58:41Bgerhahaha :)
13:59:26Bgerok, i must join in the lottery today for sure ...
14:02:33preglowoh well, i should talk to slasheri about this
14:03:56Bgerrasher IaudioX5Info ... i created nearly empty page, see your wiki rescue and u'll understand why...
14:05:09rasherIt *was* a nearly empty page except for a few image.. not much to do about that
14:05:18rasherthere's a typo though
14:05:48Bgercorrect me :)
14:05:57rasherNo, in my rescue page I meant
14:06:10rasherit said the cache I had was r1.2, but it was in fact 1.12
14:08:14 Join noC|andY`fRa [0] (
14:10:54Bgerrasher : what to do with the nonexisting links in text ?
14:11:09Bgerlike %ATTACHURL%/decoded.gig
14:11:58rasherI've just put the filename in red
14:12:20***Saving seen data "./dancer.seen"
14:12:21rashershould make it stand out for people going over the pages
14:12:46 Join cYmen [0] (
14:14:27Bgerok, clear
14:15:02 Part webguest83
14:15:17rasherWhich page?
14:16:43Bgerlast edited :)
14:16:46preglowwhere does one adjust the custom stereo configuration?
14:22:00 Join DeepB [0] (n=joe@
14:24:22 Quit pike ()
14:26:55Bgerrasher InsideMPIOHD200 done
14:27:23 Join Moos [0] (
14:28:21BgerIpodLinux done also
14:28:31MoosGood day all!
14:29:33Mooscongratulations for the speed recovering Wiki
14:29:41FebsUgh, I am not looking forward to recreating the FileMenu and MainMenu wiki menu pages!
14:31:13FebsUsing many tables and cross-links seemed like a good idea when I first created it ... not so much fun to re-do, however.
14:32:11FebsAt least it will give me the opportunity to do something that I've been meaning to do for a while, and add some screen shots.
14:32:38Mooshehe :)
14:38:17BgerIriverBDM done
14:43:10rasher44 pages to go.
14:43:30*Bger will talk with his gf for a while
14:45:45*Bger notices the bullsh*ts in IriverH3XXXComp..., he wrote 2-3 months ago :)
14:47:18rasherErp.. the "view raw" links are now displayed in a text-field.. will searchengines cache that?
14:47:27*rasher is thinking if this happened again
14:47:56muesli-Bger its a never ending story, never look back what you have done before ;-)
14:56:09Zagorrasher: i think they will, yes
14:56:34Zagorbut I wouldn't say it's a great improvement
14:56:52 Quit Seed (Nick collision from services.)
14:56:53 Join Seedy [0] (
14:57:15 Quit Dma-Sc ("What?! Open source isn't good enough for you? Bersirc 2.2 [ - Open Source IRC ]")
14:57:41 Join bobTHC [0] (n=bobTHC@
14:57:48bobTHChi folks !!
14:58:07MoosBonjour bob
15:02:15 Quit noC|andY`fRa (Read error: 104 (Connection reset by peer))
15:09:56 Quit Febs (Read error: 110 (Connection timed out))
15:18:06HCllinuxstb: darn, thats pretty sweet
15:18:11HClabout flac/wavpack
15:20:20linuxstbIs a 128-bit checksum going to cause you problems though?
15:21:50linuxstbAlso, am I right in thinking you create a checksum of the first 32K of the compressed music data? What if the track starts with silence - isn't there a big chance of a collision?
15:22:46rasherlinuxstb: I investigated this.. on my entire collection, using the first 512 *bytes* was enough to avoid collisions
15:23:09linuxstbDid that include the ID3 headers though? Which filetype(s) are in your collection?
15:23:17rasherOf course, I don't have any WAV files beginning with perfect silence
15:23:38rasherI'm not sure, I *think* I skipped id3 headers, but I'm not sure
15:23:44rashermix of mp3/ogg
15:23:49rasherabout 80/20
15:24:47linuxstbI quite often edit MP2 radio recordings by splicing pre-encoded about 2 seconds of silent MP2 frames to the beginning.
15:25:05rasherWhat the.. no PluginSolitaire in the wiki?
15:25:25linuxstbThis is about 83 frames at 576 bytes each - so more than 32K.
15:25:56rasherI was advocating first+last X bytes, but that was ruled out because of increase disk-seeking
15:27:04Zagori still think (file_size-tag_size)/2 + tag_size is a good position
15:32:43linuxstbZagor: I agree - the middle of the compressed data is a good place. Or simply just the middle of the file.
15:32:54linuxstbIt's also possible to pad the end of a track with silence.
15:33:04rasherfilesize/2 should be fine?
15:33:26linuxstbI suppose we want to guarantee we are not including any metadata tags.
15:33:28rasherit's not like the tag_size is significant compared to the filesize
15:33:53linuxstbrasher: It shouldn't be, but I think some people do odd things with id3v2 tags
15:34:36rasherEmbedded music video!
15:43:27preglowdetermining the tag size will probably require even more seeking, though
15:43:55HCli agree with zagor, though eventually we might want a hash of the uncompressed pcm data
15:44:32HCli had collissions with the first 512 bytes
15:44:45HClas for the 128bit nature
15:44:51HClwe'll just have to somehow scale it down
15:45:10HCleither that or add an 128bit hash support to the database
15:45:54preglowscaling down a hash is badness
15:46:15preglowit very much defeats their purpose
15:46:20HCli personally don't see too much problems with increasing the hash size in the database
15:46:26HClbut i don't know about other people.
15:46:30preglowthat's what should be done
15:46:46HClokay, so 128bit hash, as well as a field to say whether there is a hash or not?
15:47:02preglowperhaps a hash/checksum type field, describing size and type :>
15:47:08preglowwhere 0 means no hash
15:47:10HClthats what i suggested
15:47:16HClbut some people were highly against it
15:47:21HClfor reasons still fairly unknown to me.
15:47:28HCli dunno.
15:47:45Bgerrasher: IriverH3XXHardwareComponents done
15:47:45HClBagder: ?
15:48:24linuxstbIf we are going to support "native" checksums like the FLAC/Wavpack md5sum, then a "hash type" field does make sense - different kinds of hashes for different kinds of files.
15:48:27rasherBger: ok
15:49:07HClwell i'm glad my idea and implementation isn't totally rejected :p
15:49:12HCljust need to upgrade the hash size
15:49:24rasher35 wikipages left.
15:52:23*Bger hopes that the backup system will be working very soon again
15:55:34preglowanyone know why Lear says the ogg proglem has only been partially fixed? assuming there's enough memory for malloc and there are no leaks, i'd think it should work very well now
15:57:52Bgerrasher i hard can see any difference between IriverHardwareComponents rev 1.51 (in the wiki) and 1.52 (rescued)
15:58:04rasherI'll see
15:58:04ZagorBger: i'm doing manual backups now until things are in order again
15:58:16 Quit ender` (Read error: 110 (Connection timed out))
15:59:09rasherah.. that's a tough one :-\
16:00:31 Join muesli__ [0] (
16:01:02Bgermaybe a typo ?:)
16:01:41rasher(F)P22AD-LVX245 is changed
16:02:25rasherand that's it
16:02:33rashertwo places
16:02:42Bgeronly one
16:02:49Bgerthe other is %TOC%
16:03:03rasherah okay
16:03:05Bgerjust renamed (F) to fairchild
16:03:11Bgerok, i'll rename it
16:03:43Bgeri was checking tables, text ...
16:04:02Bgerohm resistance ... :)
16:04:08preglowhmm, i need to call apps/dsp.c code from firmware/sound.c, is that even possible?
16:05:21preglowthen dsp.c needs to be in firmware/
16:05:37linuxstbOr sound.c needs to be in apps?
16:05:57preglowbut all the audio stuff for iriver is in apps now
16:07:13preglowand to implement the 'stereo width' and 'channels' settings, i need to do work in dsp.c
16:07:27preglowbut the interface is in firmware/sound.c :/
16:09:39 Quit muesli__ (Read error: 104 (Connection reset by peer))
16:09:53rasherZagor: erp, locking is 60 minutes now.. could we lower it to 30 minutes?
16:10:04*rasher summons Febs
16:10:44BgerIriverIfpPort done
16:11:03Bgerhah how does this sound :)
16:11:19linuxstbpreglow: That must be because some targets can do it in hardware. So it seems that dsp.c should belong in firmware - because it duplicates functionality provided by hardware on some targets..
16:11:25Bger"i just saw in the irc log that Rockbox already runs on iriver ifp" ....
16:11:57preglowlinuxstb: yes, i agree
16:12:01linuxstbSimilarly, some hardware will not need samplerate conversion.
16:12:08linuxstbetc etc.
16:12:21***Saving seen data "./dancer.seen"
16:12:42preglowi really think stuff like pcmbuf belongs in firmware as well, heh
16:12:46preglowit's pretty lowlever
16:12:49BgerIriverInstall is large :(
16:12:50preglowlowlevel, even
16:17:58preglowthe playback speed change i want to do also needs to have it moved, hrmph
16:18:56 Join ender` [0] (i=ychat@
16:20:48linuxstbpreglow: firmware/ sounds the right place to me.
16:21:49Zagorrasher: NonArchos done. found among recovered files.
16:21:58Bgerrasher : after ~ 10 min i'll go offline, so i WONT finish IriverInstall
16:22:52rasherBger: okay.. just save it locally and work on it when you get the time
16:22:55rasherZagor: noted
16:23:16Bgerrasher: i'll go home, and will continue to work on it tomorrow
16:23:40rasheronly 26 left!
16:23:59Bgergood :)
16:24:10Bgeru're definitely quicker than me ;)
16:24:37rasherI pick the easy targets, I think
16:25:24 Quit muesli- (Read error: 110 (Connection timed out))
16:26:58Bgerthis iriverinstall is written for idiots ...
16:28:20preglowwell, you can't rule out them when writing docs, now can you
16:28:46*preglow tries moving dsp.x and sees what happends
16:29:55Zagormost of the undeleted files are broken. i find fragments of the files mixed with each other. often the same fragment in multiple files. frustrating.
16:30:24preglowyes, that's usually the case
16:30:37preglowi recovered a whole harddrive once, most of it was intermingled rubbish
16:30:44*preglow summons Slasheri
16:31:03Bgerrasher: what's the wiki code for the "idea's lamp" ? :)
16:31:21preglowlinuxstb: if i move dsp to firmware, the entire playback subsystem has to move as well, it seems
16:31:54Zagorpreglow: might be a good time to step back and rethink the layers.
16:32:01preglownow, why the flaming hell are the DSP_ constants in _playback.h_
16:32:16Zagorthe ones we have are very old and not designed for what we do today
16:32:38preglowi still think they're ok
16:33:03Zagorin general, yes. but some parts might need to be adjusted more than just "i'll move this here and this there"
16:33:20preglowi'm just giving things a try now
16:33:55preglowif it's possible to just do a straight move, it's a solution that'll actually work for now and changes more or less nothing
16:34:41rasherBger: hrm, %T% maybe (Tip)
16:35:57rasherBger: they're defined here:
16:36:19linuxstbpreglow: It would be nice if you had to move some code out of playback.c - it's getting far too big IMO.
16:36:28Bgerok, found it
16:36:30Bgernever mined
16:36:34preglowi don't think i'm up too a big code restructuring, i'm very bad at that
16:36:35Bgergotta go
16:36:45BgerIriverInstall is ready up to "STEP 4"
16:36:51rasherhave you saved it?
16:36:58Bgerand it's unlocked
16:37:02rasherI'll finish then
16:37:14Bgeras u wish
16:37:25Bgerif you don't do it, i'll finish it tomorrow
16:37:36rasherthanks for the help
16:37:53Bgerthanks 2 u to
16:38:26Bgeri don't think we must thank each other, everyone does this for all:)
16:38:29preglowbut indeed, playback.c has grown horribly
16:40:05 Join DangerousDan [0] (
16:41:17preglowBger: still, hearing that ones work is appreciated from time to time is nice
17:01:07Slasheripreglow: hi. Hmm, why would you ever want to move dsp part to firmware?
17:01:46Slasherieven pcmbuf is not a firmware layer (which is used by dsp)
17:02:00preglowSlasheri: because i need to interface with stuff in firmare/
17:02:06preglowlike stereo width, channel settings
17:02:12preglowfirmware code can't call app code
17:02:12 Join Sucka [0] (
17:02:15SlasheriAh, hmm..
17:02:24Slasherican you write an api to interface with that?
17:02:34Slasherii think dsp shouldn't need directly to access them
17:03:09preglowwell, i could pepper the apps/ code with ifdefs to call firmware code in the case of archos, and apps code in the case of iriver
17:03:12preglowbut i think that's wrong
17:03:22Slasherihmm, yes..
17:04:49preglowthe archos/iriver stuff really needs to be unified by some brave soul :/
17:05:05 Quit ashridah ("Leaving")
17:05:52preglowand there's another thing i'd like to ask your opinion about
17:06:04Slasherii think the mpeg.c needs to split into firmware and application layers..
17:06:26Slasheribut now some food ->
17:06:43preglowtodays log 13.34
17:07:42preglowsure, lemme know when you've read it
17:07:45*preglow goes washing up
17:11:02 Nick jborn_ is now known as JoeBorn (
17:18:10Slasheripreglow: hehe, there is no 1334 on my log :) now i have to guess the time zone.. searching :D
17:19:02Slasherioh well, it was 1434 ;)
17:20:07preglow13.34.07 # <preglow> anyone got any clever ideas on how to handle audio modifications without heaps of latency on rockbox?
17:20:17Slasheripreglow: yes, that is problematic..
17:20:19preglowahh, i meant the rockbox log, heh
17:20:59Slasherieven if we could access directly the pcm buffer, we cannot "undo" the crossfading without re-decoding
17:21:06preglowSlasheri: we _need_ to fix that, imho, in no way can i (or most users, i guess) accept a five second latency in their eq adjustments
17:21:31Slasherior undo any other changes..
17:21:43Slasherithat needs decoding the affected part again
17:21:47preglowoh yes
17:21:49 Join kurzhaarrocker [0] (
17:22:39pregloweasiest way of doing it is storing the buffer in full precision and doing the dsp part closer to actually sending the data to the dac
17:22:41kurzhaarrockerStrange. I can't find out to which of my email adresses the rockbox mailing list sends to. I don't find any of my personal mail adresses in the header.
17:22:44preglowbut that wastes a lot of memory, of course
17:23:40Slasheripreglow: hmm, that's true.. but yes, it doubles the memory consumption
17:27:08kurzhaarrockerfound it
17:27:10 Part kurzhaarrocker
17:27:10 Join IanPM [0] (
17:29:46*rasher stares at ManualMainMenu
17:30:03rasherThat's so harsh
17:30:06IanPMI don't suppose anyone knows any projects similar to Rockbox for the SonyHD5 do they
17:30:52linuxstbI don't know any projects similar to Rockbox.
17:31:35linuxstbAll I know is Rockbox, ipodlinux, and then various projects which hack linux-based devices
17:32:20preglowSlasheri: we could also of course do all dsp on the 16 bit data, but that would break our idea of an internal format and might give precision problems
17:33:44linuxstbIs the current PCM buffer as small as it can be, or could we reduce it?
17:34:41rasher19 pages left.
17:34:56IanPMShame, the HD5 could use some work in the firmware department
17:36:30 Quit Seedy (Read error: 110 (Connection timed out))
17:43:32 Join ze [0] (
17:51:33preglowlinuxstb: can probably reduce it, but it's nice to have it largeish
17:51:41preglowto compensate for sudden load spikes
17:53:24 Join noC|andY`fRa [0] (
17:54:01linuxstbpreglow: I only asked because at the start of the iriver port, we set everything much larger than needed.
17:54:08preglowlinuxstb: it's been reduced since then
17:54:19linuxstbI should pay more attention :)
17:54:28preglowcan't follow everything
17:55:01amiconnrasher: Is the attachment problem solved now?
17:55:23amiconnAh, apparently it is :)
17:55:30rasheramiconn: I believe so - the new TWiki version was handling attachments differently
17:55:38*amiconn is going to upload fresh voice files
17:56:11preglowamiconn: excellent
17:56:34HClqueen rocks
17:57:15preglowamiconn: can't fix stereo width and channel config stuff at the moment thanks to apps/firmware clash
17:57:29preglowamiconn: and btw, where do you actually set the 'custom' channel configuration?
17:58:09preglowSlasheri: oh, and yes, why are the DSP_* constants put in playback.h of all places?
17:58:09amiconn(1) Hmm, what clash? (2) "Custom" uses the "Stereo width" percentage
17:58:40Moosamiconn: what's about the recording bit-shift problem, any news?
17:58:44preglowamiconn: stereo width and co are placed in sound.c, and i can't call dsp code from firmware code
17:58:59amiconnAll other modes ignore that, and use their own config (mono = 0%, stereo = 100%, karaoke ~ infinity)
17:59:48 Nick Asku_ is now known as Asku (
18:02:47amiconnpreglow: All channel configurations are handled in sound.c/set_channel_config()
18:03:22amiconncase SOUND_CHAN_CUSTOM: calculates the scaling factors for custom mode
18:04:01amiconnsound_set() calls this function in two cases
18:04:23amiconnThe clash is in fact a difference in architecture
18:05:01amiconnFor hwcodec, stereo width is part of the hardware settings just like volume, bass etc
18:05:13preglowyes, indeed
18:05:25amiconnFor swcodec it goes into an entirely different module
18:05:30preglowall the settings we've implemented for iriver so far have been hardware based
18:05:33preglowso no clashes
18:05:38amiconnPerhaps we should have a dsp_set() in apps/dsp.c
18:06:12amiconn...and use that insetad of sound_set for "soft" settings
18:07:27amiconnIn fact we could do so for all platforms, but that's a thing I'd like to discuss with Linus too
18:07:58amiconnAfter all, the stereo matrix settings of the mas are dsp settings, as is the pitch setting
18:08:08Slasheripreglow: ah, i think there is no real reason for that. They can be moved to dsp.h
18:08:16preglowSlasheri: ok, i'll try
18:08:19amiconnThe mas is divided in 2 parts that are somewhat independent.
18:08:20Slasherigood :)
18:08:43amiconn(1) The audio codec (2) The dsp
18:09:08preglowamiconn: pitch change on mas isn't "gapless" ?
18:09:19amiconnVolume, balance, bass, treble, loudness, MDB are part of the dsp
18:09:27amiconnErr, part of the audio codec
18:10:07amiconnpreglow: No, because the pitch shifting is done by "cheating"
18:10:12preglowamiconn: yeah, i saw
18:10:18preglowiriver pitch change will be seamless, at least
18:10:21preglowamiconn: how long is the gap?
18:10:25amiconnWe tell the mas a different clock frequency than it really has
18:10:45amiconnWhen pithcing up, we're effectively overclocking the MAS dsp
18:10:57bobTHCbut on the terratec who use the MAS too , the pitch feature work well
18:11:15amiconnThe gap is a fraction of a second
18:11:29amiconnbobTHC: With or without gap?
18:11:41bobTHCwithout and without disto
18:11:58preglowbobTHC: well, yeah, but i'm certain they've got a proper pitch changing effect
18:12:06bobTHCfor sure
18:12:16preglowremember that rockbox has to do it by cheating because it has no intended real pitch change function
18:12:23***Saving seen data "./dancer.seen"
18:12:48amiconnbobTHC: Would be interesting to know how they do it
18:12:49bobTHCiirc we've got some info on the MAS programming ?
18:13:50preglowyes, but please don't remind us ;)
18:13:57 Quit Maxime (Read error: 104 (Connection reset by peer))
18:14:28 Join Maxime [0] (
18:15:16bobTHCwhy amiconn ?
18:15:57bobTHCiirc it was a [IDC]dragon announce...
18:17:04bobTHCi found the thread =>
18:17:43preglowyes, it's true, but programming the mas is very hard
18:17:54amiconnThis is about the pcm codec. I have that available
18:18:12amiconnIt's not about general MAS programming info
18:18:33 Join Seed [0] (
18:18:40amiconnThe pcm codec is binary & accompanied by a datasheet similar to the standard MAS datasheet
18:19:20bobTHCso pcm wave playback is possible ?
18:19:37amiconnYes, and recording
18:19:43bobTHCbut what the deal with the bandwith ?
18:20:13amiconnPlayback should be a no-brainer, as the serial bandwidth is 3MBit/s and it's fed via DMA
18:20:40amiconnRecording is a different thing though. The interface is parallel, but we can't use DMA
18:20:46 Join dpassen1 [0] (n=dpassen1@resnet-233-61.resnet.UMBC.EDU)
18:21:14amiconn...and have to use a routine that "mimicks" DMA as the MAS parallel port is designed to use DMA
18:21:36amiconnThat's the one I'm struggling with...
18:23:14amiconnIn case the bitshift problem multiplies with the higher transfer speed needed for pcm recording, it might be impossible
18:23:21 Join markun [0] (
18:24:10amiconnFor pcm playback/recording on archos the HD will probably be spinning all the time
18:24:56bobTHCand that can imply a quality limit...
18:25:34amiconn(for internal mic it's obvious, but otherwise?)
18:25:42 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
18:27:45bobTHCif when we record the bandwith requiered for 44Hz is higher than the parallel port
18:28:19amiconnAh, you mean the parallel port bandwidth
18:28:48amiconnIt _should_ be sufficient for 44.1/16/stereo _if_ we manage to solve the bitshifting problem
18:29:41bobTHCoki, nice
18:31:43rasher6 wikipages left
18:33:11rasherall the worst ones, of course
18:33:40preglowwell, well
18:33:47preglowwhat am i doing wrong, no codecs are working anymore, heh
18:36:01 Join DrMoos [0] (
18:36:37 Quit Moos (Read error: 104 (Connection reset by peer))
18:41:38 Join Gibbed [0] (
18:41:38 Quit Rick (Nick collision from services.)
18:41:48 Nick Gibbed is now known as Rick (
18:52:37 Join tvelocity [0] (
18:54:14preglowamiconn: no eta on release yet?
18:54:16*rasher considers declaring WpsGallery "lots"
18:54:36rasherThere have been 150 updates since the backup :-\
18:54:46preglowrasher: i'd say let people fix that themselves
18:55:16bobTHCit's the easyest to refill imho
18:55:29rasherThe ReleaseTodo is back - a version from 2005-09-06
18:56:32 Nick DrMoos is now known as Moos (
18:57:11 Quit bobTHC ("Smoke Weed Every Day !")
18:57:27 Quit Seed (Read error: 104 (Connection reset by peer))
18:59:27 Quit Rick (Read error: 110 (Connection timed out))
19:04:29rasherSo.. now only ManualMainMenu, WpsGallery, ReleaseTodo and TrackScreen are out of date
19:04:38rasherManualMainMenu is a beast.
19:04:56rasherWpsGallery might as well be updated by people who put them there in the first place
19:05:22rasherReleaseTodo needs to be checked and updated. Shouldn't be too hard. It's as up to date as the searchengines allow.
19:05:27rasherTrackScreen is just lost.
19:05:35rasherThe update, that is.
19:07:02 Quit IanPM ("leaving")
19:15:04 Quit thegeek (Read error: 113 (No route to host))
19:15:57 Join thegeek [0] (
19:21:12 Quit rasher (Remote closed the connection)
19:23:42 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
19:26:04 Join crash3m [0] (i=crash3m@unaffiliated/crash3m)
19:32:26 Quit markun (Remote closed the connection)
19:34:06 Join Mxm`Pas`Bien [0] (
19:37:00 Join hshah [0] (
19:49:16 Quit hshah ("Leaving")
19:49:16 Quit Maxime (Read error: 110 (Connection timed out))
19:52:37 Join Rick [0] (
20:05:58 Quit paugh ("Leaving")
20:08:53 Join linuxstb_ [0] (
20:09:05 Join Lear [0] (
20:09:28preglowLear: yo, any reason for not thinking the ogg problem with the alloca is gone?
20:09:44preglowLear: and why not use builtin_alloca for the sims as well?
20:09:44 Nick Sucka is now known as Sucka`away (
20:10:22 Quit dpassen1 ()
20:12:24***Saving seen data "./dancer.seen"
20:13:05 Quit Mxm`Pas`Bien (Read error: 104 (Connection reset by peer))
20:13:22 Join Maxime [0] (
20:17:51LearWell, I have only had one ogg causing crashes, so I fixed that problem...
20:18:19LearAnd I tried using it in the simulator too, but I got link errors then, don't know why...
20:18:52Lear(Not exactly sure what you mean with the first question, I must say... :) )
20:18:56preglowvery strange, builtin_alloca is a compiler hint more than anything
20:19:10preglowit's not an actual function
20:19:59preglowyou said in the post you thought you're not sure how complete the fix is
20:20:06preglowjust wondering what makes you think it's not complete
20:21:04preglowwe're probably going to need a better malloc some day :/ i don't think we'll ever get rid of them, so might as well do it properly
20:28:53LearAh, well, that's simple. After having tested the problem a bit, I came to think of the alloca thing (didn't like it when I first saw it), so I gave it a try, not knowing if it would help.
20:29:31LearIt did turn out to help, and at that point I wrote the post. Later on I actually tried it with debug prints in the simulator, to see the real behaviour.
20:29:33preglowi actually spotted that bug while away the last weekend
20:29:36preglowbut you beat me to it, heh
20:29:50preglowthe previous implementation was right out wrong
20:29:58preglowand the current sim implementation still is
20:30:00LearAnd it was lots of small alloca()s from one place in Tremor.
20:30:40 Join solexx_ [0] (
20:30:59LearRegarding builtin, aren't they all intrinsic (sp?) functions, i.e., stuff that generates code inline? Happens on target at least.
20:31:06preglowthey are
20:31:09preglowso do alloca
20:31:13 Quit Sucka`away ("a bird in the bush is worth two in your house")
20:31:24preglowit just keeps track of how much to subtract from the stack pointer when the function exits
20:31:38LearAnyway, didn't look closely at the sim problem; audio playback is a bit "shaky" there anyway...
20:31:39preglowand that's how the automatic deallocation happens
20:31:45preglowprobably, never tried it
20:32:15preglowbut alloca for gcc is always #defined to __builtin_alloca via alloca.h
20:32:16LearYeah, I know. There's some alignment stuff too. Nothing complicated or magic though. :)
20:32:33LearYes, I was a bit surprised when it didn't work...
20:33:34 Join rasher [0] (
20:33:53preglowbtw, do you know if the "rumours" of tremor on rockbox being more efficient than in the iriver fw is true?
20:34:41preglowif so, we should really beat them into the ground with the additional 16kb of iram we don't use yet
20:36:37LearI haven't read those rumors, and they are difficult to substantiate. Battery runtime, which is pretty much the only way to measure it (short of a BDM or something), shows more differences than CPU load...
20:36:54LearBut you need to do something useful with those 16 kb...
20:37:00preglowoh, indeed
20:37:06preglowputting the entire floor decode in iram, for example :P
20:37:36LearWhat I'd like to see is someone try out that "MDCT via FFT" thing, to see if it is faster on the Coldfire.
20:37:52amiconnWe're compiling with -ffreestanding which implies -fno-builtin
20:37:53LearMaybe the FAAD code can be used right away? :)
20:38:11preglowamiconn: for both sims and targets?
20:38:20Learamiconn: and that affects __builtin_alloc in what way?
20:38:26 Join Mxm`Pas`Bien [0] (
20:39:01preglowin no way
20:39:07Learfloor decode, the code you mean?
20:39:08preglowas long as you spell it out to gcc
20:39:11preglowLear: yes
20:39:14 Quit Maxime (Read error: 104 (Connection reset by peer))
20:39:14amiconnLear: __builtin_alloca() should be available under this name, but not as alloca()
20:39:30amiconn...unless you explicitly #define alloca
20:40:17preglowLear: i know code cache, etc etc, but it usually actually does make a difference
20:40:27preglowLear: and there's not much data left to be put in iram
20:41:07LearI've noticed too, but is the floor decode that large part of the decode? Don't remember seeing it in the quick profiling I did (not on target, true)...
20:42:01LearWindow lookups for more than 256 and 2048 can actually be useful. With different sampling frequencies, and with much compression, window sizes can range from 256 to 4096 with the current encoder.
20:42:06preglowi believe it is, that's where the entire floor construction and dot product is done
20:42:10preglownot 100% sure, of course
20:42:22preglowthere is some data malloced here and there that we can put in iram
20:42:27preglowa lot of constant size mallocs
20:43:39LearBtw, I think it was in vorbis_lsp_to_curve that did the alloca()s causing problems, so it is only used on low quality oggs or something.
20:44:31LearAnd that function has been written in arm asm as well, so it might benefit from some attention...
20:45:03preglowthat's a floor0 function
20:45:23preglowin which case it isn't just a low quality ogg problem, it's a very-old-encoder problem
20:45:49LearAnother thing: the alloca fix has one negative side effect: during decode setup, up to ~3.5 kB stack can be used by a certain function...
20:46:06 Quit solexx (Read error: 110 (Connection timed out))
20:46:15 Quit Moos ("Parti")
20:46:21preglowLear: yes, but it's still no way near libmad stack usage, no?
20:46:29preglowLear: in which case i don't see anything to worry about ;)
20:46:38LearDon't ask me. :)
20:47:28LearNo, the problem file I used to locate the problem (twit21) is encoded by libvorbis 1.1.0, which isn't particularly old...
20:47:29preglowlibmad has 90% usage
20:47:35preglowhmm, well
20:47:43preglowfloor type 0 is only used by really old encoders, afaik
20:48:01preglowthat's the bark scale thing
20:48:10 Join Febs [0] (n=upirc@
20:49:53*Febs found a way to access irc using his Treo smartphone.
20:50:07Learwait a minute, it might have been _01inverse too. Don't quite remember... :)
20:50:18 Quit Rick ("I… don't need to be here.")
20:50:27FebsWebsense be damned!
20:51:14Febsrasher, you were looking for me earler?
20:52:48 Join Rick [0] (
20:53:23 Quit rasher ("Ex-Chat")
20:53:54 Join rasher [0] (
20:54:37rasherFebs: It was something to do wiith WikiManual.. I figured it out - hadn't seen the mail at that point
20:55:52Febsok. I know that ManualMainMenu is going to be a nightmare ... I saw your earlier comment on that.
20:56:29LearThat MainMenu thing should be split into several pages, IMHO...
20:56:31rasherYeah. It's pretty much the only one left
20:56:45LearBit of a pain to edit something that large...
20:57:07Febslear, yes, I agree.
20:58:08FebsOriginally I had ALL of the menus in one page, but that grew too big so I split it into chapters.
20:58:40FebsMenu menu is big enough now to require subchapters.
20:58:41preglowi wish monty had used whitespace when he wrote tremor
20:58:47preglowi hate reading source code with no whitespace
20:58:58rasherreindent it?
20:59:30preglowwell, we should avoid modifying codecs
21:00:00rasheroh right, I forgot that
21:00:32Febsrather than simply restoring the previous ManualMainMenu page, I will take this opportunity to refine it.
21:01:26preglowi think i'll go play libflac on tremor as well
21:01:40preglowthrow some calloced shit into static variables instead
21:01:54*preglow croons for a wav writer
21:03:02preglowLear: i wonder how much we should strive to support floor0 files
21:03:10preglowno oggenc has made them since god knows when
21:03:13preglowand we waste good iram there
21:03:17goais or will there be any kind of "database browsing", such es listing by artists or genres?
21:03:28preglowgoa: there is
21:04:02goapreglow: ah sorry, the site seems to be online again, so i'll rtfm ;)
21:04:08preglowLear: ignore me, no idata in floor0 :>
21:04:17preglowgoa: do that
21:04:36goayeah, sorry.
21:05:25preglowno worries
21:05:33preglowi'm not being harsh ;)
21:05:54goaok =)
21:09:36goaehrm, just while I fly over it, there is "crossfading" under "Features we will not implement" −− for that, it works pretty good ;p
21:09:58rasherWhere are you reading this? Url?
21:10:17goapage 87
21:10:19rasherAh, that's outdated
21:10:29goamh :/
21:12:38goaerm, do i have a cache problem? the same sentence is there ;/
21:12:54goaoh sorry, missed the "Archos line of players"
21:13:10rasherYup, that page has been split for that exact reason
21:16:54 Join hicks [0] (
21:24:36 Quit linuxstb_ ("CGI:IRC (Ping timeout)")
21:27:06 Join Dma-Sc [0] (
21:29:36goapreglow: mh ok, so it seems to be this one there is no support for ape2 tags (yet), right?
21:30:07Dma-Scthe devkit downloadable in the wiki is really not corresponding to the instructions, as someone pointed in the forum
21:32:15preglowwavpack supports ape2
21:32:27preglowbut i guess the db doesn't
21:32:30Febswiki question: is there any equivalent to the html COLSPAN function?
21:32:31preglowa shame, 'tis
21:32:55FebsI want a 3 column table ...
21:32:55preglowwhat would be great would be the db gen being written in c and using the rockbox metadata parser
21:33:09rasherFebs: horizontal merging?
21:33:23Febswhere columns B and C are combined in some instances
21:33:31rasherthat'll make foo span two cells
21:33:46rasher(no space between ||)
21:34:12 Join linuxstb_ [0] (
21:34:27Dma-Scwell i was wrong about the devkit, it's correct, it's just two persons who can't read (me and the forum guy)... ;)
21:34:43Febsgot it. It was the 'no space' that was throwing me off.
21:34:48goapreglow: yup, i only have wavpack files here, that's why i asked ;)
21:35:13preglowgoa: well, rockbox supports ape2 in wavpack files, but the db generator doesn't, afaik
21:35:45goapreglow: mh, is there a donate button somewhere? =)
21:36:07rashergoa: under the menu
21:36:10preglowgoa: sure, on the front page, downmost left
21:36:28goaomg, i'm blind..
21:36:53preglowgoa: but don't expect a donation to give you apev2 support in the db gen, only sucking up to HCl will get you that, hehe
21:37:44 Nick NibbIer is now known as nibbler (
21:42:05 Quit Febs ("Leaving")
21:42:50 Join Seed [0] (
21:43:04goapreglow: i'll just wait until it's done, it's the same with all opensource projects ;) but that doesn't mean that they could not use some support :) and if i cannot help with things i have, (hosting, server or bandwidth) then i'll drop a few EURs to paypal..
21:45:11preglowi don't think you'll see any complaints on receiving a donation, no
21:47:16preglowby god, is people unable to unsubscribe themselves these days?
21:48:08rasherYes, yes they are.
21:51:44 Quit Seed ("cu, Andre")
21:52:59 Join Seed [0] (
21:56:44goadone, 50$, not much, but maybe a small motivation to continue this project..
21:57:27 Join [IDC]Dragon [0] (
21:57:55Zagorgoa: thanks. $50 is lots, and very appreciated.
22:04:56[IDC]Dragonamiconn, have you located the recording loop in the archos s/w?
22:05:02preglowgoa: 50 is much for a donation, you're an angel
22:05:42[IDC]Dragonis that for backup tapes?
22:06:02*[IDC]Dragon tiptoes sideways, to avoud the worst flames
22:12:25***Saving seen data "./dancer.seen"
22:18:18 Quit rasher (Remote closed the connection)
22:37:54 Join Moos [0] (
22:40:04 Quit linuxstb_ ("CGI:IRC")
22:45:02 Quit Lear ("Chatzilla [Firefox 1.4/undefined]")
22:45:14 Quit Seed (Nick collision from services.)
22:47:31 Join hshah [0] (
22:47:45Dma-Scsimulator rule for designing wps :)
22:47:59hshahur the one who posted in my thread :p
22:48:03hshahread my reply...
22:52:23 Join stripwax [0] (
22:53:42Dma-Schshah, eheh yeah i replyed :)
22:58:40hshahcool thanks Dma-Sc
23:00:47Dma-Scpretty nice that this tutorial/devkit exists, it saves time :)
23:01:44hshahtook me a while to write it - but decided to do so coz i was stuck in the same position as many others
23:03:08 Quit noC|andY`fRa (Read error: 104 (Connection reset by peer))
23:05:15 Join burningrave1011 [0] (
23:05:21 Part burningrave1011
23:05:54 Join rasher [0] (
23:07:25Dma-Schshah : i can imagine that in packaging a clean devkit is quite apinful
23:07:30Dma-Scin > and
23:07:58hshahrasher- how is the wiki restore coming along
23:08:04hshahpackaging devkit?
23:08:17rasherPretty much done, I'd say
23:08:47Dma-Schshash : yeah the devkit package .exe i mean
23:10:23 Join linuxstb_ [0] (
23:11:22*rasher positions himself squarely on the fence in the neuros talk
23:12:35fuzziegood work.
23:20:22 Join Dan [0] (
23:26:39 Quit hshah ("Leaving")
23:30:07 Quit edx (Read error: 110 (Connection timed out))
23:47:12 Quit DangerousDan ("Miranda IM! Smaller, Faster, Easier.")
23:53:59 Join webguest59 [0] (
23:54:09webguest59Hola muchachos
23:54:19rasherWell hello there
23:54:41webguest59o scuse
23:55:32webguest59what is the status of release please?
23:55:53rasherWaiting for recording tests/fixes
23:56:03webguest591 month is so long :)
23:56:13webguest59what is wrong with it?
23:56:25rasherRecordings break after a few hours
23:56:47webguest59planed to fix it, or release with?
23:57:02rasherHard to tell - the ball is amiconn's
23:57:14webguest59ah ok
23:57:28rasherBut he's not around much this week
23:57:33webguest59he is alone in this bug?
23:57:53rasherPretty much.
23:58:03webguest59alone working on it, want to tell

Previous day | Next day