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 2005-07-05

00:00:11HClthats just odd..
00:00:14HClit wasn't working
00:00:21HCli checked whether the runtime database was turned on
00:00:24HCli went to usb mode
00:00:27HClcame back from usb mode
00:00:29HCland now it suddenly works..
00:00:43*HCl dislikes bugs that "disappear"
00:01:09 Join webguest18 [0] (
00:01:13Bagdermy rockbox.rundb is 8 bytes
00:01:23Bagderseems... small
00:01:29hardeepmight need a new hd for that Bagder
00:01:37*HCl commits
00:01:46HClBagder: archos or?
00:01:54HClamiconn added an option that has to be turned on
00:01:57HClits in playback options
00:02:04Bagderplayed for ~90 mins today
00:02:06HCl8 bytes means it just has a header
00:02:15BagderI enabled the option
00:02:21HClthen i think we're having the same bug
00:02:24BagderI'll doublecheck
00:02:27stripwax_btw - i think my hdd IS knackered, you know.. :-( i'll see if i can get a warranty replacement for it, if not I'll upgrade to 30GB :-)
00:02:40 Join PaulJ [0] (
00:02:56webguest18pill: that feature already exists, look up the definitions of 'Insert', 'Insert Next', 'Insert Last'
00:02:57HClthere's something wrong with the rundb init
00:03:02HClit works after plugging in usb
00:03:06HClbut it doesn't get initialized on bootup
00:03:15*HCl goes to check
00:03:39 Join Infirit [0] (
00:03:41BagderI have the option enabled
00:03:42HClwho removed the rundb_init o.o;;
00:03:45 Part webguest18
00:03:46Bagderchecked now
00:03:50HClyea, the rundb init disappeared from the main init()
00:03:55*HCl goes to check cvs..
00:04:05 Quit Infirit (Client Quit)
00:04:59HClcan i check all the files that have been changed with a specific commit?
00:05:04HClamiconn moved the init somewhere, but broke it..
00:05:26*HCl checks cvs log..
00:05:30Bagdercvs doesn't really group files in a commit
00:05:51Bagderbut you can "guess" based on the comment and timestamp
00:06:18 Join webguest85 [0] (
00:06:36HClyea, i found it anyways...
00:06:53HCli don't see why amiconns change would be erronous, but it does bug for some reason..
00:07:03 Join LinusN [0] (
00:07:06HClmaybe its trying to init before the settings are loaded?
00:07:07HCli think thats it..
00:07:25hardeepokay, looks like moving resume update to mpeg/playback threads works fine...
00:07:33HClrundb doesn't belong in tree though...
00:07:40hardeepwould there be a reason it shouldn't be moved there?
00:07:50HClthe tagdb does because its needed for browsing, but the rundb init has nothing to do with the tree structure
00:08:15BagderHCl: yes, it is probably better off in a new file
00:08:29Bagderhardeep: not that I can think of
00:08:44Bagderamiconn or Linus might have opinions
00:09:01HClthe settings do get loaded before the call to the tree init though..
00:09:01 Quit webguest85 (Client Quit)
00:09:02LinusNi don't really remember if there was a reason
00:09:09HCli'm confused to why its failing to init at bootup
00:09:18LinusNperhaps a will to only update rtc from one thread?
00:09:41HClrundb_init *has* to be inited after audio_init or it won't work
00:09:58hardeepmaybe it had to do with the separation of apps and firmware and settings are only available in apps?
00:10:12hardeepi worked around that by adding a playlist function that the mpeg thread calls
00:10:15BagderHCl: surely the comments for that function explains that precondition? ;-)
00:10:52HCl :P i just added the comment
00:10:59HCli know, writing documentation is on my todo list
00:11:09HClfixed in the bleeding edge build
00:11:24Bagderyou and Slasheri write about the same amount of comments in your code
00:11:37HCl :P sorry, i'll try to do better
00:12:05Bagderit isn't really a complaint at this point, merely a remark
00:12:09HCl :p
00:12:15HCli'll try to do better anywho
00:12:24hardeepLinusN: If you wanted to look at it, I have a patch at
00:12:37hardeepif you have no problems with it, i'll go ahead and commit
00:15:03HClputting runtime data in the id3 structure was a really good idea
00:15:08 Quit Moxon ("leaving")
00:15:16HClits really easy to access the runtime data of the current song, and just as easy to change it
00:15:18LinusNhardeep: the rtc may be accessed from two threads, please check if there's a mechanism to handle that
00:15:46hardeepLinusN: yeah, I was just looking... doesn't appear to be any protection around it
00:15:56HClhow did i manage this..
00:15:56HClmake[1]: Warning: File `main.c' has modification time 2.4e+02 s in the future
00:16:17BagderHCl: using nfs?
00:16:37HCllocal file... just edited them with vim...
00:16:50HCli don't see why their modification dates would be in the future..
00:16:52Bagderbut vim is weird! ;-P
00:17:15hardeepLinusN: oh wait, rtc_write() calls i2c_begin() which takes a mutex
00:17:32LinusNhardeep: ah, yes
00:18:15Bagderhardeep: seems suitable you add a little comment about that
00:18:21HClwhen i add language strings, will english.lang do or do i have to add blanks for all the other languages..
00:18:37BagderHCl: english only is good enough
00:18:38hardeepBagder: comments??? blasphemy! =)
00:19:14BagderHCl: the other languages will get the english versions until they provide their own
00:21:47 Quit gromit` (Read error: 104 (Connection reset by peer))
00:22:41 Join gromit` [0] (~gromit`
00:23:40 Quit west-acre ("I-n-v-i-s-i-o-n 2.0 Build 3515")
00:27:39*HCl wonders where to add a tag for playcount and rating
00:27:54HClmost trouble i'm having is finding a spot where i can still put at least a sensible letter to use for it :P
00:28:16 Join cYmen [0] (
00:28:48 Quit cYmen_ (Read error: 104 (Connection reset by peer))
00:29:28*Bagder fades away
00:30:26n0bbyBADGER, NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOoooooooooooooooooooooooooooooooo!!
00:31:10 Nick courtc_ is now known as courtc (
00:31:11LinusNhmmm, the fading backlight made the boot loader hang/crash
00:32:02 Quit ghostiger (Remote closed the connection)
00:33:13HCli had to ask bagder a question
00:33:17 Join jlee [0] (
00:33:21HClwho knows stuff about wpses
00:33:21LinusNabout what?
00:33:28HCli need a very simple thing
00:33:36HCli just need to make a tiny own wps with a special tag i just made
00:33:38HClto test whether it works
00:33:44LinusNthen do it
00:33:55*HCl guesses he'll go read the wiki
00:33:57HCli dunno how they work..
00:34:06LinusNget the default wps from wps-display.c
00:36:39HClkind of scary how gcc does format string checking on snprintf..
00:38:09 Join TCK [0] (
00:39:03 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
00:39:33 Quit PaulJ (".")
00:43:22LinusNhardeep: i think you should commit the patch
00:43:42stripwax_HCI scary in what way?
00:44:13hardeepLinusN: okay... I was just double checking that calling settings_save() from two different threads wouldn't be a problem and it looks like it's fine
00:45:45HClthats pretty cool
00:45:52LinusNhardeep: in fact, periodic updates shouldn't be necessary anymore
00:45:56*HCl commits wps support for playcount and rating
00:46:08hardeepLinusN: why not?
00:46:22LinusNbecause we have "soft" poweroff for all platforms
00:46:37hardeepoh, right
00:46:54HClmy wps shows playcount and rating now :)
00:46:55hardeepshould i even bother with the settings_save() then?
00:47:02HClair - all i need; playcount: 4 rating: 9
00:47:04HCl :)
00:47:10LinusNhardeep: i dunno
00:47:28hardeepi'll leave it in for now
00:50:05LinusNHCl: how do you rate it?
00:50:14HClin the context menu :)
00:50:24HClwps context menu
00:50:25HClthat is
00:51:03*HCl goes to rate all the best songs of his collection with a 9 :)
00:51:59HClmight it be possible to add certain "graphic" tags?
00:52:14HCllike a dot in the middle
00:52:15thegeek_is the scale 0 to 9 ?
00:52:19 Nick thegeek_ is now known as thegeek (
00:52:22HCl0 meaning no rating
00:52:32LinusNHCl: i don't get it
00:52:34thegeekoh well
00:52:37HClLinusN: mm?
00:52:43HClthegeek: why?
00:52:46thegeekthat's a big scale for something that subjective
00:52:46HClLinusN: what don't you get?
00:52:47thegeekI mean
00:52:52LinusNgraphic tags
00:52:58HCli'm being a bit vague
00:53:03HClhold on
00:53:05thegeekhard to choose between 4 and 5;)
00:54:00HCllike that one is using - | and + in order to do some minor graphics
00:54:29HCland the one of jonnyDr (there's no link to it, but its above that one) has small dots
00:54:33HClbut that dot isn't in all fonts
00:55:17*HCl tried that one but it ended up showing as some weird u
00:55:19HCl :/
00:55:35HCland people using ==== to create a seperator line
00:55:49HClit'd be better if we had simple graphic tags for that
00:55:57HCllike draw line from there to there
00:56:18LinusNaha, now i understand
00:56:34LinusNbut i don't see anything in the link you gave me
00:57:03HClsee how he uses |-+ to mimick some sort of folder structure..?
00:57:06HClin his screenshot?
00:58:12HCli just noticed
00:58:14HCli posted the wrong link >.<
00:58:32HClthats the one i meant :X
00:58:36HClgomen nasai
00:58:48HClhm.. actually
00:58:54HClthe wps gallery needs some work
00:58:59HClthere are two html markbright tags in there
00:59:02HCland its taking the first one..
00:59:12HCllook at the one in the iriver section......
00:59:33LinusNi see
01:00:15HClit'd be nice if we would be able to offer a way to do some minor graphics instead of forcing people to ascii art
01:00:26HCljust a simple draw line from there to there would go a long way
01:00:58HClbut i gotta go to sleep..
01:07:42 Quit Sucka ("a bird in the bush is worth two in your house")
01:09:58*HCl has some bugs with id3 reading, wonders whether he created it..
01:09:59HClah well
01:11:12 Join Data_OverLoad [0] (
01:11:17 Part Data_OverLoad
01:12:35 Join austriancoder [0] (~austrianc@
01:12:40austriancoderhi all
01:12:46HCllong time no see
01:14:04LinusNyou want the 3.3V one, not the 5V
01:14:38austriancoderfine.. this was one of the questions i wanted you to ask ;)
01:16:11 Join Peuc [0] (
01:25:57 Quit Aison ("( :: NoNameScript 3.72 :: )")
01:27:00n0bbyi'm doing a serious overhaul on the formatting of BTW
01:27:16n0bbycould take a while
01:27:30n0bbyanyone need a quick edit lock release, just tell me :)
01:27:37 Join gromit`` [0] (~gromit`
01:27:44LinusNwhat are you doing with it?
01:27:51n0bbyits all in <pre>
01:27:55n0bbyim changing it to tables
01:27:58n0bbylooks far better
01:28:17 Quit gromit` (Read error: 104 (Connection reset by peer))
01:28:24n0bbybut it needs <pre> around the wps tags
01:28:32n0bbyso its taking a lot of changes to do
01:29:04 Join xen` [0] (
01:29:52 Quit cYmen ("zZz")
01:30:20 Quit hicks (Remote closed the connection)
01:32:16LinusNthe SYS_POWEROFF seems to work fine
01:32:57LinusNthe viewer plugin now automagically saves the settings before the power is cut
01:32:59tvelocityhow much usable is the recording function in the debug menu? can it be used in the real world safely?
01:33:21LinusNtvelocity: sure, but it seems to overwrite the same file every time
01:33:30tvelocitythat's no problem to me
01:33:52tvelocityi'm currently testing for how long it can record withe the battery full
01:35:25tvelocityany known issues with it? my roommate wants to record a live event
01:37:50LinusNlooks like it doesn't split long recordings into multiple files yet
01:38:28 Quit markun ()
01:38:43LinusNso you will probably hit the 2gig limit before the battery runs out
01:39:31n0bbysilly microsoft
01:39:31tvelocityi think 2gigs are enough though, he wants to record for ~3 hrs
01:39:52n0bbyi wonder what the largest available HDD size was when the fat32 format was designed
01:40:08tvelocityis there any record glitch, like in iriver FW, or was that a software problem?
01:41:05LinusNthe current test in rockbox has a "glitch" in the beginning of the recording, but the rest of it should be nice and quiet
01:41:34tvelocityit suits my need perfectly ;)
01:41:39LinusNhaven't tried it myself though, i'm just repeating what others have said
01:42:29tvelocitywhat will happen when it hits the 2gig limit? will it stop, or will it crash and destroy mankind?
01:44:26LinusNiirc, rockbox will go on until it hits 4gig, but the results are, so to say, "undefined"
01:44:42n0bbywiki is annoying
01:44:46tvelocityheh let's find out:P
01:44:46 Join Cassandra [0] (
01:44:54LinusNtvelocity: backup first
01:45:10n0bbyi cant use | in a table without it forming a border, EVEN WITH <pre> OR <verbatim> AROUND IT!!!!!!!!!!!
01:45:15tvelocityi haven't anything i want in it
01:45:33 Join EoS [0] (
01:45:42tvelocitybesides, it's allready at 2hrs and 50 minutes... i won't stop it now!
01:45:47LinusNEoS: hi
01:45:50stripwax_Hmm.. bit of a noob question but, .. I just tried adding support for pcm vu meters on iriver, and with the new code and plugins installed on my player, the unit turns itself off moments after turning on. Anything obvious? Clearly I just bungled somewhere..
01:46:22EoSi just saw a pic of the rockobx loading on a gmini 120, is it fully funtional for those players now?
01:46:31 Quit Cassandra (Read error: 131 (Connection reset by peer))
01:46:34n0bbynot even slightly
01:46:35LinusNstripwax_: perhaps not that obvious
01:46:52n0bbyi think...
01:47:06EoSah too bad..
01:47:16LinusNEoS: the gmini developer(s) have been away for quite some time
01:47:18stripwax_LinusN - I can't think of anything that would cause it to just turn itself off :-(
01:47:35LinusNstripwax_: how quick?
01:47:35EoSyeah that pic on the rockbox page is from jan 05
01:48:00EoSthx n00by
01:48:20LinusNstripwax_: is the patch online somewhere?
01:48:48 Nick n0bby is now known as _Mark (
01:48:50LinusNi could run it with the bdm
01:48:55stripwax_LinusN - no, sorry. About three seconds? don't get as far as the rockbox logo
01:49:02***Saving seen data "./dancer.seen"
01:49:41_Markthats a pretty nasty bug just for a VU meter :P
01:50:08stripwax_LinusN - I'll take another look tomorrow night, and I'll try and get a patch together then
01:50:10stripwax__Mark - heh
01:50:49EoSlol oops sry i called u n00by i misread :)
01:51:04_Marki've used the nick "nobby" for years and everyone does
01:51:28EoSits just too tempting to misread
01:51:43_MarkIt was "nobby nobbs", the discworld character
01:51:55_Markbut that was too long for repeated typing so i shortened it
01:52:00_Markbad move
01:52:03 Part stripwax_
01:52:19EoSim prolly getting a iriver h140 tomorrow :)
01:52:58 Join Cassandra [0] (
01:53:14thegeekdiscworld <3
01:56:20_Markthis bug in the wiki parser is pissing me off
01:56:35EoSok thx for the info guys, im going to bed
01:56:52 Quit EoS ()
01:57:21CassandraLinus: You know the patch for alignment in the WPS? Is that suitable for inclusion in 2.5?
01:57:41LinusNhave you tried it?
01:57:42tvelocitywow... 3hrs and still recording
01:58:13CassandraNope. I didn't feel that I knew enough about the WPS display code to evaluate it.
01:58:30 Part Peuc
01:58:52 Quit Hadaka ("leaving")
01:58:54LinusNit looks simple enough, so if it works, i think it's 2.5 material
01:59:16 Join Naked [0] (
02:00:57CassandraWe seem to be pretty much decided on a freeze. Are you waiting for the chip8 patch before calling it?
02:01:20LinusNi'd like that patch too, yes
02:01:30LinusNand i want the tag database stable
02:01:52 Quit jlee (Read error: 110 (Connection timed out))
02:01:53CassandraDo you mean rundb? AFAIK tagdb is stable.
02:02:36_Mark"the chip8 patch"??
02:02:59Cassandrachip8 plugin. It plays old HP-48 games.
02:03:01Rickwas chip8 ever fixed?
02:03:05Rickon iriver
02:03:17CassandraApparently someone has a patch.
02:03:23_Markit works for some people
02:03:27_Markwhats the patch do?
02:03:34CassandraFixes it.
02:03:44_Markit works on my ihp
02:03:44CassandraThat's all we know. No-one's seen the patch yet.
02:03:46Rickiirc the main problems with display+speed issues
02:04:06_MarkROFL @ waiting for a mystery patch that no-one knows what it does
02:04:49_Marknice patch
02:05:15RickI need to finish z
02:06:56Rickyou guys really cut down the amount of bug reports
02:10:34LinusNthis makes me sad:
02:11:26austriancoderLinusN: Could you look at this?
02:11:39austriancoderi cant compile and run it at the iriver
02:11:52austriancoderand i dont know if it works
02:12:49LinusNdid you adapt the radio screen as well?
02:13:06austriancoderits based on this:
02:13:46austriancoderi think.. i must look at the other computer... but i can do this tomorrow... now i am not at home
02:15:15LinusNupload what you have as a patch in the tracker, so i/we can continue working on it
02:16:00 Join ashridah [0] (
02:18:21*Cassandra thinks thoughts about mysticriver users.
02:21:12CassandraYou know, I'm beginning to think Rockbox on iPod would be a dreadful idea, since it would dilute the average user inteligence even further.
02:21:28LinusNCassandra: my thoughts exactly
02:21:57LinusNgoes in line with my "wma user with a clue" argument earlier
02:22:33LinusNthe guy who chooses to buy an ipod doesn't need, and will probably not appreciate rockbox
02:23:00 Quit gromit`` (Read error: 110 (Connection timed out))
02:23:14CassandraI put a lot of effort into making Rockbox comprehensible for Joe Average. Shame a lot of them don't put in even the little bit of effort required to read the manual/faqs/whatever.
02:25:16 Join amiconn_ [0] (
02:25:18hardeepare we leaving the h300 build red so that people don't ask about it? the fix is simple
02:25:33LinusNhehe, i simply don't care
02:26:28LinusNbut you have a point
02:26:43LinusNit might be better if it stayed red
02:27:40CassandraI think it should stay red till the LCD driver is in place.
02:27:57LinusNat least until we can boot some code
02:27:58CassandraWhat needs fixing with tagdb, Linus?
02:28:20LinusNCassandra: i just feel like there's so much happening in that area
02:28:37LinusNwith both a perl and a java version of the creation tool
02:28:55CassandraWish someone'd write one in C.
02:29:14LinusNand lately i have seen some "it doesn't work with the latest builds anymore" posts in the forums
02:29:31CassandraI think the perl one is out of date, yes.
02:29:50LinusNso we need so sort that out
02:30:41CassandraIt's also kind of buggy.
02:31:06CassandraThe java one seems OK apart from the minor problem that it's written in bloody java.
02:31:16Moosw00t austriancoder patch it's ready :)
02:31:46austriancoderits only a try... i cant test it.. broken iriver
02:31:55*Cassandra rebuilds her Archos toolchain in preparation for looking at the alignment patch.
02:32:41 Join gromit` [0] (~gromit`
02:35:00CassandraWhat's the patch do, austriancoder?
02:36:16CassandraCustomise the crossfader? Where do people get these ideas from.
02:36:19LinusNCassandra: i2c stuff, plus fm radio
02:36:34CassandraOh, cool. What's the i2c stuff do?
02:36:42LinusNfm radio
02:36:55LinusNthe fm chip speaks i2c
02:37:18austriancoderthis patch does only i2c.. radio will follow, when i am back home
02:39:10LinusNi gotta go to bed
02:40:24CassandraOh, arse.
02:40:38CassandraMy gcc-3.3.6 build choked.
02:41:31 Quit amiconn (Read error: 110 (Connection timed out))
02:41:32 Nick amiconn_ is now known as amiconn (
02:41:53CassandraOh, hold on. I didn't apply the workaround.
02:42:22 Part LinusN
02:42:29 Quit austriancoder ("using sirc version 2.211+KSIRC/1.3.12")
02:43:10Moostime to go to sleep to here
02:43:15Moosgood night all
02:43:21CassandraWay past time to sleep here.
02:43:27CassandraInsomnia strikes again.
02:43:54 Quit Moos (" HydraIRC -> <- Go on, try it!")
03:10:34tvelocity4hrs and still recording
03:10:35 Quit _Mark (Read error: 111 (Connection refused))
03:11:13 Quit Stryke` (Read error: 60 (Operation timed out))
03:14:24*Rori just watched The Matrix DeZionised. Much better.
03:14:53 Join wladston [0] (wlad@
03:15:26wladstonppl .. anyone here worked with Z80 before??
03:15:37 Join _Mark [0] (
03:16:11_Marklinus, still around?
03:16:36CassandraHe went to bed.
03:17:28wladstoni'm in lack of good disassembler ... ida does not solve the problem .. =(
03:20:26CassandraSorry, no idea.
03:23:46wladston=( ok ... thanks ...
03:28:02 Quit xen` (Read error: 110 (Connection timed out))
03:35:08 Quit wladston ()
03:49:05***Saving seen data "./dancer.seen"
03:51:34tvelocityrecord-0.wav, 0 bytes :(
03:55:20 Join CheeseBurgerMan [0] (
04:02:13_Markhow long was it going for?
04:05:51 Join QT_ [0] (as@area51.users.madwifi)
04:19:30 Quit QT (Read error: 110 (Connection timed out))
04:22:31 Join jlee [0] (
04:23:16tvelocity_Mark, it was recording for over 4 hrs when the battery went down
04:26:39 Quit tvelocity ("Leaving")
04:28:42 Quit _Mark (Read error: 104 (Connection reset by peer))
04:33:36 Part CheeseBurgerMan
04:38:37 Nick Febs is now known as Febs_away (
04:45:28 Join Stryke` [0] (
05:10:34 Quit Stryke` ("Friends don't let friends listen to Anti-Flag")
05:38:26 Quit thegeek (Read error: 104 (Connection reset by peer))
05:46:54 Quit hardeep ("BitchX: so real, you'll wet yourself!")
05:48:29 Join _Mark [0] (
05:49:06***Saving seen data "./dancer.seen"
05:54:43 Quit ashridah ("Leaving")
06:00:22 Quit RotAtoR ()
06:04:08 Join webguest32 [0] (
06:04:11 Quit webguest32 (Client Quit)
06:08:06 Join Zoom2 [0] (
06:22:11 Quit Zoom2 ("CGI:IRC (EOF)")
06:39:27 Quit jlee ("leaving")
06:47:57 Join ashridah [0] (
07:01:57 Quit _Mark (Read error: 110 (Connection timed out))
07:33:09 Join courtc_ [0] (
07:34:02Bgermorning all
07:34:18 Quit courtc (Connection timed out)
07:48:37 Quit ashridah ("Leaving")
07:49:08***Saving seen data "./dancer.seen"
07:53:54 Join kaouete [0] (
07:59:52kaoueteif i want to change the hdd in my archos recorder 15, is it better to have one with a big cache (about 8mo) or will it just drain more power for no more performance ?
08:00:33kaouetei just read a topic with the same question on the speed of the hd
08:01:52kaouetebecause for the same price you have hd with double size and 8mo instead of 2mo, i'm wondering if it is a good idea to take one like that
08:02:17 Join LinusN [0] (
08:05:52kaoueteand another question : is the next stable rockbox (fox archos) planned to be rleased soon ? :]
08:07:38LinusNkaouete: rockbox 2.5 will probably be feature-frozen in a couple of days
08:08:01LinusNthen expect it within 1-2 weeks
08:08:12LinusNbut 2.5 is only for Archos devices
08:08:29kaoueteok :]
08:08:39kaouetethat . .. . . "rocks"
08:09:58Rickwhat does that mean?
08:10:16Rickonly bugfixes?
08:10:27kaouetethere will only be bugfixes
08:12:43LinusNonly bugfixes, but iriver development may continue
08:20:30LinusNkaouete: regarding the hard drive, a large cache won't help at all
08:21:01amiconnLinusN: Someone had the idea to go to 11 MHz while on USB on iriver. What do you think?
08:22:00LinusNit's a little tricky with the current solution though
08:22:09kaoueteLinusN: ok, thanks
08:22:10LinusNespecially since the backlight boosts it
08:23:26amiconnAh, yes
08:23:44amiconnIt boosts only while fading though...
08:24:31amiconnPerhaps the base frequency could be switchable
08:25:07amiconnDuring normal operation it would be 48 MHz (or whatever we change it to) and during USB it would be 11 MHz
08:25:43SlasheriHmm, what about third boost level? So 0 would be 11 MHz, 1 48 and 2 120 MHz
08:25:58amiconnIt may be easier to let the fading not boost while on usb
08:26:17amiconn...or not
08:26:44Slasherior disabling the cpu_boost while on usb?
08:27:32Slasherisomething like locking the cpu to the current frequency and cpu_boost should not have any effect while it's locked
08:28:14amiconnI don't think this would be good
08:28:42amiconnImagine the fading locks it at 48 MHz while playing a .flac...
08:29:08Slasherino, i mean that the usb handler may lock it :)
08:30:59amiconnThere would be some (very minor) problem with backlight fading and USB
08:31:27amiconnIf USB is plugged or unplugged while fading, it would glitch
08:31:44amiconnI'd say this can be ignored
08:32:48Slasherihmm, that's true..
08:34:12Slasherii think so too, it's not very common that usb is plugged in while backlight is fading :)
08:34:22LinusNi think the base frequency idea is the best one
08:35:04HClmrf. 8 am is too early
08:35:30SlasheriHmm, but then the backlight fading would still boost from 11 MHz -> 48 MHz?
08:35:38LinusNno, to 120
08:35:42LinusNfrom 11 to 120
08:36:00Slasherithat doesn't sound very good either.. :)
08:36:32Slasheribecause we jump to 120 when we don't need to do so.. but maybe it's quite minor problem
08:37:07LinusNquite minor
08:37:33LinusNnot many users play with the buttons during usb transfer
08:37:46Slasheriand yes.. changing the base frequency would prevent glitches if the backlight is fading while plugging the usb
08:38:03LinusNand it's a pretty small change in rockbox
08:41:22amiconnLinusN: Any news for the MFDR(2) change?
08:41:46 Join DJ_Dooms_Day [0] (
08:43:25LinusNamiconn: how about this:
08:44:07LinusNlet cpu_boost() check usb_inserted() and select CPUFREQ_DEFAULT if true and CPUFREQ_NORMAL if false
08:44:30DJ_Dooms_DayWhats news guys?
08:45:36LinusNDJ_Dooms_Day: not much, just hacking along
08:45:48amiconnLinusN: That would be Slasheri's locking idea, and could work, but there are 2 things to consider
08:46:14LinusNyes, it needs to go back to the correct frequency when unplugging
08:46:17amiconn(1) How would the transition 48MHz -> 11MHz happen? cpu_boost() might not be called
08:46:27LinusNthat too
08:46:31DJ_Dooms_DayAnymore big leaps made? I heard mp3 playback is now all good. But there were still a few freezing bugs out there
08:46:35amiconnSame for the transition back at the end, yes
08:46:59amiconn(2) You should not check usb_inserted(), but ask the internal status of the usb thread
08:47:00LinusNthe lame approach would be cpu_boost(true);cpu_boost(false;
08:48:01LinusNamiconn: that's what usb_inserted() does, doesn't it?
08:48:05SlasheriDJ_Dooms_Day: Currently i don't know any bugs that can cause a crash (except resampling to too low frequency)
08:48:42LinusN return usb_state == USB_INSERTED;
08:49:47amiconnLinusN: Ah, yes. I confused it with usb_detect()
08:50:00LinusNthoght so
08:50:12*LinusN can't spell today
08:50:30 Nick QT_ is now known as QT (as@area51.users.madwifi)
08:51:46LinusNamiconn: the other approach would be to introduce a new system function, let's say cpu_idle_mode(bool yesno) or something
08:51:47DJ_Dooms_DaySlasheri - Cool, so if i put it on i could browse and listen to music just fine? Has the browsing code been completed?
08:52:01LinusNSlasheri: i hear complaints about "ticking" when upsampling
08:52:34DJ_Dooms_DayYeah i heard that too
08:52:34amiconncpu_idle_mode() sounds good, imho
08:53:05DJ_Dooms_DayHas the 'ticking' problem been fixed?
08:53:07amiconnThe ticking while upsampling is caused by roundoff errors
08:53:31DJ_Dooms_DayWhich means what?
08:53:55amiconnI had an idea how to fix it, but I didn't have the time to do it. Still busy with gfx...
08:54:18LinusNamiconn: any gfx progress?
08:54:23DJ_Dooms_DaySpeaking of which, has greyscale been properly implemented?
08:56:58SlasheriLinusN: yes, that's a known problem too..
08:57:23SlasheriLinusN: btw, i just calculated that CURRENT_NORMAL 80 could be more accurate for iriver
08:57:46LinusNSlasheri: good, change it
08:57:59Slasheriok :)
08:58:24Slasheribut that's only when playing 128k mp3 files..
08:58:44Slasheriogg takes much more current
08:59:16Slasherimaybe we could somehow use the cpu boost ratio to calculate accurate current
08:59:20LinusNso the boost ratio might be a good measure of current consumption
08:59:25Slasheriyes :)
09:03:41 Join ghostiger [0] (~ghostiger@f5764a7ca1cfeb36.session.tor)
09:04:07LinusNhmm, should cpu_idle_mode have the same type of counter? i don't think it should be necessary
09:05:16SlasheriHmm, where is that cpu_idle_mode?
09:05:37LinusNit will be in system.c
09:05:43Slasheriah :)
09:06:32LinusNi mean if it should behave like cpu_boost()
09:06:35amiconnLinusN: Didn't you check the logs? ;)
09:06:36amiconn22.38.13 # <amiconn> Hmm, 4-grey core done, screendump adapted, x11 sim (temporarily) adapted. Win32 sim and bmp2rb still left...
09:06:49amiconnThat was y'day
09:06:53LinusNah yes
09:07:03LinusNso, how slow is it btw?
09:07:09Slasherinow iriver shows remaining 16h 15 min.
09:07:17amiconnI don't notice a difference in the UI
09:07:44amiconnIt's only that the LCD is slower when actually showing greyscales
09:07:46Slasherii think that's better although it's still not accurate on every type files
09:08:09amiconnThe screendump works very nice, producing 16-colour bmps
09:08:12LinusNSlasheri: better than before at least
09:08:18LinusNamiconn: kewl
09:08:23amiconn(bmp only allows 1, 4, 8 and 24 bit depth)
09:08:49LinusNi see
09:08:49amiconn...and I fixed the bmp header with some advanced preprocessing :)
09:09:04DJ_Dooms_DayDoes the current rockbox compile have greyscale in it?
09:09:10amiconnDJ_Dooms_Day: Nope
09:09:19amiconnMy local version has, though
09:09:33DJ_Dooms_Dayheh, nice
09:10:01DJ_Dooms_Dayi think i might put rockbox on my iriver
09:11:18 Join cYmen [0] (
09:12:56SlasheriCrossfade has stopped working in the latest cvs i think :/
09:13:08LinusNhow nice
09:13:24Slasherii don't know yet why, just noticed it when i did update
09:16:53amiconnGapless is still not smooth with lame −−nogap encodings
09:17:18amiconnThese should be the easiest imho, but for some reason some frames are still dropped
09:25:10 Quit Seed (Read error: 104 (Connection reset by peer))
09:26:21Slasherifixed, there was a problem that crossfade_init was called twice
09:26:41 Join B4gder [0] (
09:37:57 Quit ghostiger (Remote closed the connection)
09:40:00 Join Seed [0] (
09:42:14B4gderanyone here checked it out?
09:49:12***Saving seen data "./dancer.seen"
09:49:34Maximeerm, i've got a question
09:50:03Maximewhen in the WPS, if we push one button, we have the file list, how do we go back to the WPS ?
09:50:21B4gderthe play button
09:50:32B4gder(on iriver)
09:51:03Maximebut this won't play the highlighted file?
09:55:21LinusNMaxime: not very intuitive, i know
09:56:33MaximeLinusN: i've looked as with the original firmware so.. I've thought play started the selected file so.. lol ^^
09:56:58LinusNone *could* expect that Play plays a file... :-)
09:57:18Maximeyep that's why.. lol
09:57:24LinusNwe might change that behaviour as the iriver version matures
09:57:40Rickreally need to stick with a naming convention
09:57:46Rickwhy not cpu_boost_mode() or cpu_idle() ?
09:59:04 Quit Seed (Read error: 104 (Connection reset by peer))
09:59:06LinusNcpu_boost() is more of a request to boost the cpu
09:59:23Rickis idle not also a request?
09:59:25LinusNcpu_idle_mode is a mode change
09:59:40Rickhmm, ok
09:59:41LinusNthat affects the cpu_boost() behaviour
09:59:50 Join Seed [0] (
09:59:56Maximeis it normal that a 22Khz mp3 is playing -slowly- ?
09:59:59Maximei think so
10:00:12 Nick Lynx_awy is now known as Lynx_ (
10:00:16LinusNMaxime: no, that's not normal
10:00:35LinusNis it stereo or mono?
10:01:18MaximeI think the tag is'nt well done, rockbox detects it as a 22Kz, winamp as a 44
10:01:48Rickmy guess, send it to one of the codec devs
10:01:57Rick(the mp3)
10:02:11LinusNMaxime: put it up somewhere so we can dl it
10:02:19Maximei'll try with the latest cvs version first
10:03:31LinusNRick: do you have a suggestion to make the boost api clearer?
10:03:50RickLinusN: not really
10:03:55LinusNor to explain it better in the wiki?
10:04:34Rickwell, it sounds like something that would conflict with each other (boost() then setting idle mode)
10:04:53LinusNhmmm, yes
10:05:36Rickis it possible to get the current state?
10:06:12LinusNnot officially :-)
10:06:20LinusNdo you think it's needed?
10:06:40Rickmm, dunno, just a random thought that popped into my head
10:06:56Maximewhat's happens if we ask two times cpu_boost(true) ?
10:07:02RickI guess the boost tracking would take care of it
10:07:12Rick(assuming it removes the boost afterwards)
10:07:40LinusNMaxime: it will boost as usual, but it won't unboost until cpu_boost(false) is called twice
10:07:54Maximeerm ok
10:08:48Maximeso if cpu_boost(true) is called four times, to unboost you have to call cpu_boost(false) four times.. maybe a verification might be "useful" .. or no.. lol
10:08:57 Join markun [0] (
10:09:19RickLinusN: does iriver have some sort of 'timeout' for unboosting?
10:09:27Ricknot iriver
10:09:43*Rick shrusg
10:09:58LinusNrockboy wouldn't appreciate that
10:10:45RickDid rockboy ever get the key sequence stuff yet?
10:10:54*Rick hasn't been keeping up with development
10:11:04LinusNkey sequence?
10:11:14Rickit was a feature I suggested
10:11:27Rickfor repeating key presses at the same time so you can simulate having extra keys
10:13:06Rickbasically the idea was that you register a key sequence using play (hold play, press first key, then second key) then the first key would be bound to simulate those keys being pressed at the same time
10:14:01LinusNwee, would be hard to play sonic that way :-)
10:14:09Maximethe mp3 is detected as a "mp2 22050Hz, 48Kbps CBR"
10:14:25*LinusN realizes that sonic is a Sega game :-)
10:15:00LinusNMaxime: can i see the file?
10:15:17MaximeI'll upload it, gimme a sec
10:15:36*LinusN thinks the vorbis seek patch looks ok
10:16:28RickLinusN: howso?
10:17:10MaximeLinusN: <
10:17:30 Join MisticJeff [0] (
10:18:03MisticJeffLinusN: all Rockbox FAQs removed from MisticRiver
10:18:15LinusNMisticJeff: sorry for being grumpy
10:24:27LinusNi have a nice debugging idea, a good project for the iriver wannabe hacker
10:24:57LinusNi want a cpu frequency field in the iriver status bar
10:25:18Rickiriver has a status bar?
10:25:24LinusNgood to have during development
10:25:34Rickyou mean that top bar thing?
10:25:54CassandraWe call it "the status bar". ;)
10:26:01LinusNshould be a pretty simple hack in status.c
10:26:55Rickin about a month i'll probably start working on z again
10:26:59Rickin preperation for college ;P
10:27:10LinusNround up FREQ to MHz, snprintf() it to a buffer and put it in the status bar
10:28:45LinusNwee, ogg seeking works, and vorbis comments too
10:28:56Rickcool :)
10:29:39Slasherioh, when that has started working?
10:29:49LinusNa patch came in this morning
10:29:56Slasherigreat :)
10:29:57LinusNi'm just about to commit it
10:29:58Ricka mystical magic patch that is
10:30:16LinusNThe Magical Mystery Patch
10:39:57 Join leftright [0] (
10:40:00LinusNi'm getting negative playtime in wps
10:40:10Maximecool lol
10:40:24Maximeah, may rockbox show the remaining time instead of elapsed,
10:40:25leftrightyes me too, whenever I pley the last track in a file only
10:40:32 Join Godeater [0] (
10:40:56B4gderMaxime: I've added your problematic song to my collection
10:41:06Maximek ^^
10:41:30LinusNit seems to occur after skipping
10:41:35LinusNto the next file
10:44:35leftrightI get negative time display and the file playback stalls at 00:-4, this inly occurs if I select the last track in a file for playback
10:44:37LinusNvorbis seeking and comment support committed, thanks to Ryan Jackson
10:48:48 Join Aison [0] (
10:49:54 Join ashridah [0] (
10:51:42B4gderHCl: when will we get that java parser thing in CVS?
11:01:58LinusNlunch time
11:03:09 Quit Seed (Read error: 131 (Connection reset by peer))
11:08:44 Join Seed [0] (
11:11:17 Quit Seed (Read error: 54 (Connection reset by peer))
11:11:44 Nick Naked is now known as Hadaka (
11:15:08 Join Seed [0] (
11:23:04 Join niobos [0] (
11:23:12niobosmorning all
11:23:28nioboshmm... whatever
11:24:05niobosi've just finished my exams, so I decided I could sleep until lunch...
11:25:33*niobos is off to breakfast... or lunch...
11:28:29 Quit Godeater ("CGI:IRC (EOF)")
11:37:55MisticJeffcan someone guide a noob through getting the tag database up and running using the .jar file?
11:38:52MisticJeffcopy to root>from DOS run java -jar d:\songdb.jar d:\ and absolutely nothing happens
11:40:29leftrightHCl is the pro, its his baby
11:42:01 Join ep0ch [0] (
11:48:15 Part MisticJeff
11:48:47 Join [-AIR-] [0] (
11:48:51 Nick [-AIR-] is now known as west-acre (
11:49:13***Saving seen data "./dancer.seen"
12:13:40 Nick Febs_away is now known as Febs (
12:13:57 Quit ep0ch (Read error: 60 (Operation timed out))
12:18:52 Join Moos [0] (
12:20:31 Join hicks [0] (
12:25:25 Quit Maxime (Read error: 104 (Connection reset by peer))
12:27:45 Quit DJ_Dooms_Day (Read error: 104 (Connection reset by peer))
12:32:02 Join Maxime [0] (
12:53:06 Join rwlogix [0] (
13:04:38*B4gder takes the plunge
13:04:47B4gderinto 2.6.13-rc1
13:09:15 Join thegeek [0] (
13:10:11*HCl encountered several crashes while using rockbox today...
13:10:18HClalso some bugs in runtime database, will look at it later
13:10:19*rwlogix tries out his new "Venini"-Pipe ,)
13:10:23*HCl goes to sleep x.x
13:10:47rwlogixgood night, sleep tight
13:10:52west-acregood night, sleep tight
13:22:31Mooshe it's not the night in Netherlands, no?
13:23:06B4gderyou never know with those dutch guys ;-)
13:24:08amiconnHCl: I remember there is a bug in your java database creator. It fails if the destination file doesn't exist, but works if you supply it an emty file to work with
13:24:44B4gderwe want the java code in CVS
13:25:04B4gderor at least access to the source code
13:25:13Bgeriirc the code is in the .jar file
13:25:27Bgerso you can take it out from the .jar in the wiki
13:25:35B4gderwill do!
13:25:50amiconnHCl: Btw, sorry that my init move broke the runtime db. No wonder my runtimedb didn't get populated...
13:27:07B4gderso how do I extract the source code from it?
13:27:14B4gderI thought jar used tar format?
13:27:21amiconn.jar is .zip
13:29:30B4gder92 .java files !!
13:29:50Bgerwow :)
13:30:21B4gderand 312 .class files
13:31:12B4gderlots of .class files without souce
13:34:17B4gderno docs, no mention how to build it, no license texts
13:34:51BgerB4gder : license : as is
13:34:59B4gderI doubt that
13:35:06B4gderI think he has "borrowed" quite some code in there
13:35:55Bgeranyway, i think that we need something else for future for db creation
13:36:08MoosBagder: put the .jar in your player and java -jar SongDB.jar / for creat DB, scuse i'm not java user it was Cassandra and HCl who helped me
13:36:35BgerMoos : he knows how to use it, he wants the source code...
13:36:44Moosa ok
13:37:56niobosBger: I'm working on a C-version
13:38:20niobosbut with the exams I didn't do much
13:38:21Bgerworking on the target ?
13:38:42niobosjust a C-version to create the DB
13:38:50niobosnot rea/y on the target
13:38:58niobosbut it might work too
13:39:28niobosit can write the DB
13:39:37niobosbut doesn't sort yet
13:39:45niobosand doesn't read tags yet
13:40:41niobosIt should work on the device, since it's able to store the arrays in file instead of in-mem if malloc() fails
13:41:21niobosI'll continue the dev once I'm back from vacation
13:41:31niobosi.e. in 3.5 weeks...
13:41:52B4gdersounds good
13:41:56B4gderthe more the merries
13:42:15niobosI might put the current version online, in the mean time
13:42:33B4gderperhaps we should put it in CVS, to allow others to join in
13:42:35Bgeri just was to suggest it
13:42:50niobosok for me...
13:43:22B4gderit would be neat to re-use rockbox code for meta tags reading for that
13:43:31B4gdersince that is C
13:43:58niobosI was planning to do that, but haven't had the time to reod the rb-code
13:49:15***Saving seen data "./dancer.seen"
13:50:46*niobos is off to lunch
13:51:03nioboslet me know what to do for CVS
13:55:44 Join bipak [0] (
14:13:13 Quit bipak_ (Read error: 110 (Connection timed out))
14:21:49 Join Maxime`Mrn [0] (
14:21:49 Quit Maxime (Read error: 104 (Connection reset by peer))
14:23:43 Join RotAtoR [0] (
14:27:48B4gderniobos: its actually up to you, since your the guy with the code!
14:28:06B4gderwe could add it to cvs and hand you cvs commit access to allow you and others to work on it
14:43:40 Join Yokalosh [0] (
14:44:10 Quit Yokalosh (Client Quit)
14:50:30HClamiconn: its fine, i hadn't documented it
14:50:44*HCl goes back to sleep
14:57:27 Nick courtc_ is now known as courtc (
15:10:23 Join webguest50 [0] (
15:12:54 Quit Febs (Read error: 110 (Connection timed out))
15:13:31 Join E-P [0] (E-P@
15:14:22 Quit E-P (Remote closed the connection)
15:16:27niobosB4gder: ok with me. I guess I need a sourceforge account?
15:17:05B4gdernope, we host the cvs repo ourselves
15:17:37nioboswhat do you need then?
15:17:42B4gderI need your full name, user name and a password in a /msg
15:18:19nioboscan I change my passwd later?
15:18:36B4gderyes, but only via humans
15:18:54B4gderme, zagor or LinusN
15:19:17 Join Febs [0] (
15:19:59*amiconn now knows how to handle different dpeths in the win32 sim, and how to add grayscale lib support :)
15:20:40amiconnSomething to do for this evening...
15:21:04amiconnGrayscale lib support will come later, but it will be prepared.
15:21:39amiconnRemote LCD support shouldn't be hard as well
15:22:57west-acrehey amiconn
15:23:01west-acrewot IS grayscale?
15:23:07B4gderniobos: this code is meant for both host and target then in the end, right?
15:23:10west-acrei never really figured it out
15:23:16niobosin the end, yes
15:23:17west-acredoes it just show what colours the LCD can use?
15:23:32nioboscurrently only host is tested
15:24:00B4gderniobos: then I suggest making a new dir in apps
15:24:22B4gderusing a clever name
15:24:29B4gderany suggestions anyone?
15:24:49ashridahhmm. google earth images of my place are surprisingly recent
15:25:16B4gderI guess 'tagdb' would be a reasonable name
15:25:21B4gderor songdb yes
15:25:38niobosthat's what HCl called his Java-thing
15:25:50B4gderI was just thinking how to avoid confusions
15:25:59B4gderwith the perl version, the java version and now this C version
15:26:41 Quit lostlogic (Client Quit)
15:27:27B4gdergo with songdb, that might in fact reduce confusion
15:28:05niobosyour cvs uses pserver, I guess? or ssh?
15:28:12B4gderplain pserver
15:30:14niobosso I'll commit it to apps/songdb
15:30:21amiconnI don't know whether apps/songdb is suitable. For the host version, it should reside in tools/ and the target version will most likely be (archos: has to be) a plugin
15:30:32*niobos is off... his girlfriend just arrived
15:30:49B4gderwe currently have no code that is for both host and target, do we?
15:31:02B4gderI would assume that we make a sort of lib
15:31:12B4gderso that a target version is a plugin using that lib
15:32:41amiconnOur dir structure is not really prepared for host & target shared code
15:33:04BgerB4gder i don't think *song*db is the best name....
15:33:42B4gderit is a songdb
15:33:48nioboswell, make up your minds and msg me the result
15:33:58*niobos is off
15:34:06B4gderwith the runtimedb, it even contains songs totally without tags
15:34:28B4gderor rather _for_ the runtimedb purpose
15:34:30BgerB4gder some peoples are using their jukeboxes not only for songs
15:34:41B4gderyes, but this db is for songs
15:35:08CassandraI think of it as the tagdb myself.
15:35:10B4gderor can you forsee other uses?
15:35:47B4gderthose aren't songs
15:35:57Bgerbut you can tag them as well
15:36:15B4gderok, so perhaps tagdb is the better name
15:36:45Bgerat least i think so
15:37:14B4gderand regarding the path, I think it is better to get it in now and possibly move it later
15:40:30 Quit leftright ("CGI:IRC (Ping timeout)")
15:49:18***Saving seen data "./dancer.seen"
15:50:52CassandraI spotted a bug in the iRiver resume code. If you let a playlist run to the end, rather than clearing the resume point, it gets set to the beginning of the last track in the playlist.
15:56:36 Quit webguest50 ("CGI:IRC")
15:59:22 Join webguest00 [0] (
16:00:32webguest00as i can see an the daily builds the iriver h300 SIM has no errors, can i run rockbopx on my iriver h300 now ?:;D
16:00:52B4gderthe sim is a sim
16:01:06B4gderand it simulates the wrong lcd too
16:01:54webguest00but the other one only havev 4 errors ?
16:01:59webguest00can i use that ?:D
16:02:08B4gderyes, once you've installed the bootloader
16:02:14B4gderand fixed the missing lcd and adc code
16:02:17webguest00is it safe :D?
16:02:24B4gderwhat is?
16:02:28B4gderthere is no bootloader yet
16:02:53B4gderso it is safe ;-)
16:03:39webguest00aaa when does the bootloader comes out?
16:04:03B4gderwhen someone has written it
16:04:16B4gderthat someone is likely to be LinusN
16:04:30B4gderand no, we don't have any schedule
16:05:21webguest00but to make a bootloader wont take long?
16:05:28B4gderit won't?
16:06:20B4gderfeel free to write it
16:08:03webguest00in not so good at such things;)
16:11:10 Quit Maxime`Mrn ()
16:12:00 Quit webguest00 ("CGI:IRC")
16:17:45SlasheriHmm, does ata_spin() spin up the disk?
16:19:13amiconnIirc ata_spin() just resets the spindown timer, to keep it spinning. There is no separate spinup function.
16:19:54SlasheriAh, ok. I just would like to spin up the drive before initiating a track change
16:20:06amiconnWhy that?
16:20:28Slasherito prevent audio buffer going empty (required for crossfade to work well)
16:20:28amiconnWe don't want to spin up at each track change...
16:20:47Slasherino, only when it's necessary :)
16:21:03B4gderSlasheri: should use instead use the spinup time to caclulate when you should start reading data
16:21:07amiconn...we only want to spin up when the buffer goes empty, and that's what the low watermark is for
16:21:25amiconnThe low watermark is calculated dynamically on archos
16:21:29Slasheriamiconn: yes, of course that's not a problem
16:21:52B4gderso what is?
16:22:06Slasheribut if user selects a new track, we want the hard disk to be ready before stopping codec and flushing old track entries
16:22:47amiconnWhy that?
16:22:48B4gderI don't understand
16:23:03B4gderthis works perfectly well on archos, I don't see any new need here
16:23:06Slasherithat guarantees that crossfade can work.. :)
16:23:19Slasherino other reason for this
16:23:26amiconnI wouldn't expect a crossfade when selecting a new track, even with crossfade enabled
16:23:35B4gdercrossfade just increase the low watermark requirements
16:23:55B4gderI like crossfade but not for next
16:24:34*amiconn would *very much*prefer a correctly working gapless over any fading
16:24:46ashridahi agree
16:24:56ashridahi almost never use randomized playlists
16:25:48amiconnThe strange thing is that mp3s with a lame tag containing gapless info work, while mp3s encoded with lame −−nogap (which should be the easier case imho) don't
16:26:27FebsI thought that LAME −−nogap was never properly implemented.
16:26:43FebsI remember reading something about that at Hydrogen Audio recently.
16:26:49FebsI'll see if I can find it.
16:26:57*LinusN runs the a-b patch on archos
16:27:03amiconnlame −−nogap is working perfectly if it is used correctly
16:27:18LinusNa few visual glitches
16:27:44amiconnlame −−longhelp tells everything
16:28:07amiconnI have some mix albums encoded that way, and they play perfectly gapless on archos
16:28:23amiconn...but iriver rockbox doesn't manage it
16:28:38amiconnThere is no added gap, but instead some frames are swallowed
16:29:06 Join webguest88 [0] (
16:30:05webguest88I find that with crossfade disabled it still merges the tracks, which i dislike, its nice to have the pauses between tracks
16:30:05LinusNgotta go
16:30:22 Part LinusN
16:31:19 Quit RotAtoR (Read error: 131 (Connection reset by peer))
16:31:28FebsCoincidentally, I'm listening to Dark Side of the Moon right now. The transition from Breathe to Time produced a small but audible pop. The transition from Time to Great Gig in the Sky was perfect.
16:31:28webguest88unless the album is supposed to be gapless (Dark Side Of The Moon),
16:32:06FebsThis is a HUGE improvement from the iriver firmware, for which I and many others are grateful, even if it isn't yet perfectly implemented.
16:32:13amiconnwebguest88: Rockbox doesn't change the tracks, i.e. doesn't remove gaps that are present *in the file*
16:32:32webguest88it does with crossafade off
16:32:44B4gderit just doesn't add any gap
16:33:07amiconnHowever, adding gaps that aren't present in the files (like the iriver fw does, always) would be evil
16:33:13webguest88the songs in the playlists dont have the 'natural gaps'
16:33:40webguest88but its ok I'll get a new set of ears at best buy
16:33:43ashridahamiconn: uh, i thought it had silence detection and removal
16:33:43 Quit ashridah (Remote closed the connection)
16:35:14amiconnsilence detection and removal would be another feature I'd deactivate immediately (like fade in/out/cross and whatever audio fading there might be)
16:36:54 Join ashridah [0] (
16:39:01 Join t0mas [0] (
16:39:24*t0mas is back :)
16:39:30B4gderat last!
16:50:31t0mas847 new messages
16:50:50t0masand I haven't transferred all backup MX mail
16:58:19t0masall outgoing warning mails removed :)
16:59:03t0masawk rules the world
16:59:03t0maspostqueue -p | grep "MAILER-DAEMON" | awk '{print $1}' | xargs -n 1 postsuper -d
16:59:18t0masand all warning mails are gone :)
17:00:28 Join RotAtoR [0] (
17:01:01 Join Maxime [0] (
17:02:14 Quit B4gder ("go go go")
17:06:55*HCl wonders if he found his bug..
17:07:20 Join thegeek_ [0] (
17:08:39 Join _Mark [0] (
17:09:03_Markwas the evil crashbug of doom in the iriver vu meter code fixed yet?
17:09:17_Markme wantses my precious vu meter!
17:10:18*ashridah hands _Mark a text editor and a compiler
17:10:19ashridahhave fun :)
17:10:31_Markits not MY crashbug :P
17:12:11*HCl goes to fix the database generator..
17:12:20HClor should i fix the rockbox code
17:12:44HClashridah: what do you think, should the fileentries be sorted case sensitive or case insensitive?
17:12:53HClah nm.
17:13:05t0mas_Mark: you can fix someone else's bug too :P
17:13:40_Marki dont know C
17:14:19amiconnHCl: What's the problem?
17:14:35HClbinary search failing for lowercase directory names
17:14:52HCljust altered the generator to sort it case insensitive, since rockbox' binary search assumes case insensitive sort
17:15:20amiconnYes, case insensitive is preferable imho
17:15:23 Quit webguest88 ("CGI:IRC")
17:15:28HClyea, since fat doesn't make a difference
17:15:48HClpretty important bug
17:15:57amiconnImho it always makes more sense than case sensitive
17:16:00*HCl fixes
17:16:21ashridahinsensitive gets my vote
17:16:24amiconn(perhaps except for some system dirs on *nix)
17:16:34HCli dunno how the perl version handles it at the moment..
17:18:12*HCl uploads a new songdb.jar to the wiki..
17:18:40HCldoes anybody happen to know about crash bugs while playing solitaire while playing music?
17:19:55HClthats the first critical bug in the java generator, i think :)
17:20:18_Markanyone noticed that all the newbs submitting iriver wps's are overwriting each other's because theyre all called dump_001.bmp and theyre morons?
17:20:33_Markand fecking stupid opera cant render bmps in webpages
17:20:49HClbut i did see that using the same name twice in the wpsgallery ended up in bad wiki links
17:21:48amiconn_Mark: The 'overwritten' files aren't lost. Twiki does automatic versioning, so they should be still available.
17:22:13HClalso, what would be a better way to hash files? take the 32kb in the middle of the file?
17:23:31 Quit thegeek (Read error: 110 (Connection timed out))
17:23:57amiconnHCl: Did you fix the songdb bug that it doesn't work if the database file doesn't exist?
17:24:11HClnot yet, let me take a look at that
17:24:37HClah hrm.
17:26:18*HCl prods java and scratches his head
17:28:46HClbleh, i wonder why i can't find an canCreate() function, but i'll code around it.
17:31:35HClit does, but by the means of an ioexception
17:32:21 Quit ashridah ("sleep")
17:32:34Cassandrajava - it's evil
17:33:24HClamiconn: will you be able to give the version i'm about to upload a testrun with some bad directory permissions?
17:34:17HCltrying to create a database in a non writable directory and such
17:34:25HCli don't have java on a linux machine at the moment..
17:34:40amiconnI'm not running Linux, and I'm always running the generator on the unit == FAT32
17:34:53t0mascan you commit this one:
17:35:00t0masor have you got any open bugs?
17:35:01amiconnI'm not sure whether my Linux VW contains a java installation
17:35:03HClyea, well, linux has mounted fat32 readonly before.. but i'll search for someone else..
17:35:10HClprobably not
17:35:46HClbut yea
17:35:47HClbug fixed
17:35:52Cassandrat0mas: Just not got around to testing yet.
17:35:58CassandraExpect a commit later today.
17:36:46HCldoes anyone know why my iriver doesn't seem to store the wps file its supposed to use? everytime i boot it its set back to default
17:36:55t0masCassandra: i've read it (not tested) looks ok to me
17:37:02CassandraI prefer making sure I understand patches before I apply them, which in this case was a good idea, since I spotted a bit of cruft code that shouldn't have been in the patch.
17:37:07t0masmaybe check why he wants us to put %s before and not after it
17:37:33CassandraHCl: Storing it in /.rockbox/ ?
17:37:52HClnope, i was just about to move it there to see if that helps
17:37:54HClis it required?
17:37:55Cassandrat0mas: From my reading of the code, that ought not to matter. I'll test it.
17:38:13CassandraRockbox only remembers wps settings if they're in there.
17:38:34HClt0mas: did you read my proposed simple-graphics wps tags thing?
17:38:36amiconnHCl: Yes, for storing it permanently
17:38:40HClamiconn: k
17:39:04amiconnRockbox stores the name only, without path
17:39:22amiconnThere isn't much space in the config sector
17:39:24t0masCassandra: that was what I read there too...
17:39:25HClso i can't do .rockbox\wps\new.wps either
17:39:44t0masCassandra: but still... he writes it should be before... so I wonder why he did that
17:39:52t0masHCl: no, where is it?
17:40:14HClt0mas: well, irc logs, but my suggestion was to have a few simple graphic tags like draw line from x,y to x,y
17:40:21HClso we stop forcing people to resort to ascii art
17:40:29HClthink it would be possible?
17:40:31t0masHCl: that should be easy...
17:40:37HCli think it would help tons
17:40:42HClwith how the wps looks for people
17:40:45t0masjust add 1 virtual image... and paint the things on that
17:40:56t0masbut... people can add those lines to a bmp too...
17:41:18HClyea, but its less work with just a line tag for simple seperator lines and stuff like that..
17:41:43HClits just cause i was looking at the wps gallery and noticed a lot of people were resorting to ascii art
17:41:58amiconnThe ascii art is pre-bitmap-wps
17:42:05HClyea, i guess thats true
17:42:18HClah well, it was just a suggestion :)
17:42:26t0masit shouldn't be difficult
17:42:37*amiconn wonders what'll happen to bitmap wps when he commits 4-grey graphics core
17:42:37CassandraI actually have a BDF font that contains CP437 line graphics.
17:42:40HCldo you think you might have some time to add it?
17:42:44t0masamiconn, there are already line painting functions on your gfx lib right?
17:42:55CassandraUnfortunately, it doesn't seem to work in Rockbox, and I have no idea why.
17:43:17t0masamiconn: I'll update bmp loading code for grayscale... and some people will start using weird backgrounds and stuff like that
17:44:57HCldoes anyone have any suggestions on what to hash?
17:45:23CassandraDon't hash the middle. It'll really screw up with "hidden" tracks.
17:45:27HCli need to improve the hashing algorhythm before everyone uses the current hash extensively
17:46:09amiconnImho the best solution would be to hash the beginning, but after the tags
17:46:28Cassandrahash the id3 info and the first n seconds of non-silence (volume > delta).
17:46:42HClamiconn: yea, but how would i do that?
17:46:55amiconnHowever, that would require some more sophisticated parsing, might be undesirable on target
17:47:02amiconnDo we need to hash on target?
17:47:08HClCassandra: i was wanting to not hash the tags cause then files with different tags but the same song will still get registered as the same file
17:47:18HClamiconn: at the moment, no. but that might change in the future
17:47:20Cassandraamicon: I think it's desirable.
17:47:31HClbut even then we'd have an option to hash on target or not
17:47:37CassandraHCl: Hence the reason why you hash the tags and part of the tune.
17:48:00HClmaybe i don't understand what you mean..
17:48:04amiconnCassandra: The reason for my suggestion is to *not* hash on the tags
17:48:08t0masyou mean 2 hashes Cassandra?
17:48:14t0masor you're talking bs :P
17:48:34amiconnThe tags might change, but the track is still the same
17:48:37 Join fogcat [0] (
17:48:48HCli already had that occur on my player..
17:48:55Cassandrat0mas: No, it's all data. You can treat the info structure + a section of track as a single block of data for hashing purposes.
17:49:15t0masCassandra: yes, but then you didn't understand the reason why HCl doesn't want to hash the tags
17:49:20***Saving seen data "./dancer.seen"
17:49:47fogcatcan I ask a question about that alignment patch?
17:49:54fogcatDoes it allow different alignments on the same line i.e. one bit of info aligned left and one aligned right?
17:50:04CassandraOh, right.
17:50:07amiconnCassandra: I think it is desirable to find duplicates which only differ in tagging
17:50:15t0masfogcat: it doesn't
17:50:16HClshall i just hash the middle 32kb of a song?
17:50:21CassandraI agree with amicon.
17:50:31t0maswe all do Cassandra
17:50:34fogcatok - but I bet that's what is asked for nexy ;-)
17:50:35t0masbut we want a way to do that
17:50:42CassandraHCl: No. Think of tracks from the end of a CD with a "hidden extra" on.
17:50:48amiconnHCl: The middle wouldn't help in finding duplicates with different tags
17:50:54CassandraHCl: They'll have silence in the "middle"
17:51:01amiconnThe tags might change in length
17:51:03t0masfogcat: and I bet it's something I'll code if noone does it before the end of my vacation
17:51:09HClwhat should i do then ? :/
17:51:16HCli don't quite know how to parse tags
17:51:26HCland i doubt my backend provides info about it
17:51:46amiconnRockbox already includes tag parsing code
17:51:59amiconnToo bad it's C and your songdb is java :P
17:52:01HClwhat about ape and stuff like that?
17:52:16fogcatcheers - nice one t0mas, I really should see if I can set up a build environment and remember C - talking here makes me feel guilty
17:52:20t0masjust pumped about 1700 emails to my isp
17:52:24CassandraI have quite a few instances of two identical songs with differing tags.
17:52:28t0masdon't thing they liked it :P
17:52:39CassandraI think it would annoy me to have them treated as the same.
17:52:39 Quit _Mark (Read error: 111 (Connection refused))
17:52:51CassandraOr maybe it wouldn't. I'm confusing myself now.
17:52:54t0masfogcat: yes, that's the way I started here too ;)
17:53:07HClthe only way in which they'd be treated the same is filerating, playcount, and stuff like that
17:53:25amiconnCassandra: Why do you do that? I can't think of a reason...
17:53:58t0masHCl.. what about the last part of a song?
17:54:06HClshall i just do that?
17:54:10t0masbut not 32 kb
17:54:10HClbut wait
17:54:11Cassandraami: If I have a track on a greatest hits album. Say "Hey You" which I have on both "Echoes" and "The Wall" (and a third, different version on "Is there anybody out there?")
17:54:14t0masit might be silence...
17:54:16HCli should like
17:54:18t0masdo a bit more..
17:54:19HClno wait.
17:54:33HClt0mas: the more i hash, the longer it takes to generate a database..
17:54:49t0masHCl: yes, but 32 kb is not much...
17:54:56t0mas320 kbits = 40 kb/s
17:55:07CassandraI think you should parse the MP3 until you find the first 32kb where vol>delta
17:55:15t0masso... the last 1 second can be silence... and you'll hash a lot of files the same
17:55:15HClbut they're not all mp3
17:55:22HCland we want to support future formats like wavpack too
17:55:30HClt0mas: yea, exactly
17:55:30CassandraAh, yeah. :(
17:55:56HClwhat was there against hashing the middle of a file again?
17:56:07CassandraHidden tracks on albums.
17:56:08 Join Strath [0] (
17:56:09amiconnCassandra: I don't think tracks will ever be identical if they're encoded from different albums
17:56:34HCli don't quite understand that... what happens with hidden tracks?
17:56:58CassandraLast track is followed by a couple of mins silence then the "hidden track".
17:57:17HClif the cd was ripped properly, wouldn't the hidden track have its own file
17:57:18CassandraIt's a stupid idea, but we're stuck with it.
17:57:33CassandraNo, because it's all on the last physical track of the CD.
17:57:55CassandraI can think of at least 4 albums I have that do this.
17:57:59amiconnI also have one such album... with 3 bonus tracks all appended to the last listed track
17:58:10fogcatThe musicbrainz people hash music files - is it worth checking out how they do it?
17:58:14HCli don't like those albums x.x heh.
17:58:23amiconn...and ~2.5 minutes silence before the first bonus track
17:58:38amiconnTotal track length is 22 mins
17:58:51Cassandrafc: Not really. Too cpu intensive, plus they do generate hashes where the tracks I mentioned above have identical hashes.
17:59:10fogcatokey dokey - just a thought
17:59:13CassandraHCl: Me neither. It's annoying to have to FF through the silence.
17:59:51Cassandrafc: Although it's not clear with the latter is a desirable property or not.
18:00:07HClmaybe i should just add an disable runtime database option to the wps context menu so people can disable it for those specific albums, and just hash the middle?
18:00:19CassandraI'm in favour of having a CPU intensive method and reading in old data first so you only generate a hash once.
18:00:25HCllike, the current design has no problem with disabling specific files
18:00:52CassandraHCl: I dislike that idea. You want those tracks in the db if possible.
18:00:53HClwhat do you mean old data?
18:00:55fogcatCassandra: as I understood it their aim was to hash based on how it sounds??? so guess they would want them to hash to the same value
18:01:11CassandraSongs you've already hashed once.
18:01:47HClthat would require the old tagdatabase though
18:01:54HCland i know some people who were against that
18:02:05HClit would make hashing on target practically impossible
18:02:25HClwhich isn't really something i want since at some point its nice to have it add new files to the runtime database on its own
18:02:26fogcatCould you grab 3 chunks from the; beginning, middle and end and hash those?
18:02:42HClno, but i could grab a chunk of the middle and the end
18:02:54HClbut we'd still get problems with hidden tracks with a silence at the end
18:03:05CassandraIt wouldn't require the tagdb. It'd just take a long time to generate a tagdb without the previous one.
18:03:06HClthe beginning has tags and isn't really suitable cause of that..
18:03:16Cassandra(Or for the first time.)
18:03:20HClyea, i guess..
18:03:21amiconnThe end may also have a tag
18:03:38fogcatbut then you could get hash collisions anyway
18:04:37CassandraThere is no good solution to this problem.
18:04:47amiconn...and formats like ogg are even more complex, allowing multiple streams, which might be cut in another file etc...
18:04:51CassandraI personally would favour accuracy over speed always.
18:05:11*fogcat shuts up cos he knows he doesn't know anything about the structure of music files
18:05:20*fogcat listens and learns
18:06:00HCli'm fairly tempted to just take the middle of a file for now :/
18:06:39thegeek_as long as editing header/editing tags won't change the hash......
18:06:57amiconnIt would
18:07:07amiconnThere is no simple solution to this problem
18:07:14thegeek_even editing tags?
18:07:21HClwhat tags happen in the middle of a file?
18:07:33thegeek_middle sounds fine
18:07:37HClit doesn't have to be perfect, as long as it gets it right 99% of the time
18:07:43thegeek_would avoid header and tags I guess?
18:07:43amiconnYes. The tags might change in length, so the old middle is different from the new middle
18:07:54DBUGEnqueued KICK thegeek_
18:07:58HClokay, that sucks.
18:08:00thegeek_how about a fixed offset into the file?
18:08:15thegeek_the header would change in size?
18:08:20HClfixed offset from the end of the file?
18:08:28thegeek_tags are stored at the end of the file are they not?
18:08:34thegeek_I don't know the ape format
18:08:36thegeek_but mp3 does?
18:08:47amiconnFixed offset from the end of file _might_ work, with at least 2 exceptions though
18:09:03amiconnthegeek_: Tags are usually stored in the beginning, except id3v1
18:09:16thegeek_I remember I did some id3 stuff a long time ago
18:09:18thegeek_was prob v1;)
18:09:27amiconn(1) If someone adds or removes an id3v1 tag
18:09:48amiconn(2) If someone cuts away garbage from the end of the file
18:09:56HClthis is really an annoying problem... how about different hashes based on the filetype?
18:10:08HCllike we'd parse the mp3 slightly to find the first frame and hash from there
18:10:11HCland i dunno about ogg
18:10:24HCli guess ogg doesn't have id3v1? or does it?
18:10:29HClno it doesn't
18:10:35HClso it wouldn't have the fixed offset from the end problem
18:10:37amiconnNo, ogg uses vorbiscomments
18:10:58amiconnmp3 can have id3v1, id3v2 and/or apev2
18:11:09amiconn(The latter we don't support yet)
18:11:17HCli think the best for mp3 is to just read the offset of the first mp3 frame
18:11:19HCland hash from there
18:11:27HCland we could use the other method for ogg..
18:12:28CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
18:12:28*t0mas wonders why we don't just use the middle + end...
18:12:35t0masand give warnings on collisions
18:12:39t0mas(on pc, not on target)
18:12:56t0masfor most files... I can see if they really are the same with just the filename
18:13:02 Nick Lynx_ is now known as Lynx_awy (
18:16:33amiconnHCl: Hmm, I just wondered what the hash is for after all....
18:19:02 Join hardeep [0] (
18:24:58 Quit Febs ("Chatzilla [Firefox 1.0.4/20050511]")
18:27:53 Quit fogcat ("CGI:IRC 0.5.4 (2004/01/29)")
18:29:11 Join Maxime`Mrn [0] (
18:29:31 Quit Maxime (Read error: 131 (Connection reset by peer))
18:32:51HClamiconn: its the link between the tagdatabase and the runtime database
18:33:33HClt0mas: there's already an option to show duplicates, maybe i should enable it by default..
18:35:23 Join TCK- [0] (
18:41:47HClt0mas: end will give trouble with id3v1.. i'll just do middle for now..
18:48:41HClhm wait.. :/
18:48:54HClah well >.<;
18:49:03HCli can't keep account of everything, someday we'll have to find out a better way
18:49:19 Join yyz [0] (
18:49:52 Join Febs [0] (
18:51:56HClwe need to think of a good way :/
18:52:03 Quit TCK (Read error: 110 (Connection timed out))
18:53:18HCli don't feel comfortable with introducing a hashing algorhythm thats barely any better than the old one
18:54:08CassandraI wonder if you should introduce a seperate unique key.
18:54:31CassandraThere's something in the back of my mind about non-unique keying in databases being a bad bad thing.
18:54:50HClnope, not in this case
18:55:08HClthe whole idea behind using a hash is having the same entry for the same songs
18:55:15HCleven when they're in different files
18:55:38 Join fogcat [0] (
18:55:44CassandraI don't think you're going to find a better solution than the one MusicBrainz have already come up with for that one.
18:55:49HCli guess i'll go look into the mp3 frame thing
18:56:00CassandraAnd I bet that involves parsing raw audio data.
18:56:24fogcatforgive the stupid question but what is the hash to be used for? Identifying duplicate songs?
18:56:43CassandraYou might want to look at the source for the MusicBrainz tagger and see if it gives you any ideas.
18:57:35fogcatif people have tagged their files with the Music Brainz tagger (mp3s anyway) the hash is kept in the ID3v2 tags
18:57:51CassandraAre they? Now that *is* interesting.
18:57:57HCliirc someone said music brains didn't hash any better than we did
18:58:10CassandraIn which case, all my MP3 files already have a unique hash sitting in their tags.
18:58:20fogcatSure they are... I use it on mine and I'm sure I've seen the data in there - not at home so can't check
18:58:54CassandraI'd certainly welcome an option to use MB tags as keys.
18:59:40CassandraStupid Windows ID3 browser doesn't seem to show anything.
18:59:56Cassandrafc: What do you use to examine ID3 tags.
19:00:16fogcatI've got the editor that you get with DBpowerAmp at home
19:00:21HCli need to talk to preglow, mabe
19:00:56HClhe probably knows how i should read the mp3 header to get the offset of the first audio frame
19:01:01fogcatI'll check when I get home anyway .. about time to go get the bus now
19:01:05 Quit fogcat ("CGI:IRC 0.5.4 (2004/01/29)")
19:01:27CassandraAs to MB hashing better than us, I'm certain they would claim that their hashing is better. (Although they do get dupes.)
19:03:46Cassandra*sighs* Winamp's tag editor is no better.
19:07:04 Join bagawk [0] (1000@bagawk.user)
19:09:17HCli wonder if i found something..
19:11:03HCli love java
19:11:27BagderHCl: I trust you've read my comments
19:11:47Bagderearlier today
19:11:59Bagderabout your songdb
19:12:01*HCl searches
19:12:03Bagderjava version
19:12:19BagderI lack code, docs and info on how/scripts to build it
19:12:27Bagderthere's also no license info
19:12:50HClthe code is inside the .jar at the moment..
19:12:54Bagderif this is to be the recommended tool, I think those things should be added
19:13:07Bagderis that really all code?
19:13:20Bagderok, goodie
19:13:48BagderI'm just lost in java so I didn't realize that
19:13:59HCli'll search for the licenses soon, i dunno about putting it in cvs like that
19:14:08HCli'll clean it up a bit too
19:14:16HClbut first i want to see if this hashing thing works..
19:16:17HCli love java
19:16:22HClit just hashed based on audio data
19:16:25HClrather than file data
19:16:31HClwith barely any code changes
19:16:48HCland it falls back on file data if it fails to hash on audio data
19:18:25 Join Chamois [0] (
19:20:18HClit doesn't say anything about the hash itself though? how unique is it? do files with the same audiodata get the same hash?
19:21:33CassandraHCl: That's the intention, yes.
19:22:21amiconnHCl: How are the hashes in the runtimedb created?
19:22:24HClwell, i'd have to write support for reading them, but once thats in it shouldn't be too hard to make musicbrainz the priority
19:22:30HClamiconn: copied from the tagdatabase
19:23:02amiconnIf they're just copied from the tagdb after looking up the path, and we don't verify the hash on target, why not just take the path for linking both databases?
19:23:36HClbecause that would eliminate having the same runtime db entry for the same songs though they are at different locations
19:24:45amiconnAh right.
19:25:12amiconnHowever, doing this properly would require hasing the pure audio data, without tags
19:25:28HCland thats what i just implemented in the java version of songdb :)
19:25:34HClwith extreme ease too :)
19:25:41HCljust a matter of changing datainputstream to audioinputstream
19:26:05HClit hashes on audio data now, if everythings correct
19:26:14HCland only falls back on file data when it fails to hash on audio data
19:26:20amiconnYou could verify that...
19:26:30HClyea, it'd be nice if i could have a setup to test
19:26:38amiconnJust copy a file you have, tag the copy different and then hash
19:26:38HClits all theory so far, but it seems to work
19:26:48HCli'll do that in a sec
19:26:56amiconnThey should give the same hash regardless how you change the tags
19:27:06HCllet me try that now
19:27:24amiconnI would try 2 different cases
19:27:32amiconn*5 cases
19:27:34amiconnfor mp3
19:27:37rwlogixonly a problem if the audio-decoder changes...
19:27:51rwlogix_if_ (i don't know if it ever will *g*)
19:28:33HClokay, didn't work :/
19:28:35HCli think
19:28:37CassandraHCl: Here's an explanation of TRMs (MusicBrainz tags):
19:28:37amiconn(1) File with both id3v1 and id3v2 (2) id3v2 only (3) id3v1 only (4) no tags (5) like 1, but with changed content in the tag(s)
19:29:03 Join webguest09 [0] (
19:30:16 Quit webguest09 (Client Quit)
19:32:01 Join Stryke` [0] (
19:32:28HCli wonder whether this is a bug in my backend or just because the java webpage isn't entirely clear on what you get when you call read() on an audioinputstream..
19:33:39HClwell, at least i can tell you that the tritonus backend that i use is gpl
19:35:51*Cassandra offers to have HCl's babies if he implements using MusicBrainz TRMs from the tags as a way of generating a hash.
19:36:33HClyou're scaring me o.o;
19:37:21CassandraOK. How about if I offer *not* to have your babies if you do it?
19:37:46amiconnCassandra: If there is no specific tool to examine file content like id3v2 tags, there's always the good old trusty hex editor...
19:38:31Cassandraami: It's OK, I found where the tag info the MB client creates is documented.
19:41:24 Join Coldtoast [0] (
19:41:42*HCl has to upgrade to a full-fledged tritonus backend..
19:42:15CassandraErm, you what?
19:42:39HCljava sound api implementation
19:43:03CassandraAh. Reads mp3s, oggs etc. I take it.
19:47:30 Quit hardeep ("[BX] Reserve your copy of BitchX-1.0c19 for Windows CE today!")
19:49:21***Saving seen data "./dancer.seen"
19:52:27Plugh_I can't believe people still use bitchx
19:53:18Plugh_I remember helping panasync test it in his first few builds of it, when it was private to our IRC channel. And there were much better clients available
19:54:48HCltrm seems to require internet
19:55:19CassandraTo hash, yes. To read from music that's already been hashed, no.
19:58:57HClSummary Java MusicBrainz TRM Generator
19:58:57HClCategories None
19:58:57HClLicense GNU General Public License (GPL
19:58:59HClthere we go
19:59:22HClunfortunately, they don't seem to have any files submitted
20:00:20CassandraCan I suggest making the first byte of your hash an indicator of the algorithm used to hash.
20:00:30CassandraThen you can support multiple hashing methods.
20:00:42HCloh wait
20:00:43HClthey do..
20:01:03*Cassandra ponders dinner.
20:01:06HClits just cvs only
20:01:25*Cassandra bounces.
20:02:04HClthey only have the initial commit though
20:04:03HClgot it :)
20:04:20HClthis should work fine
20:04:23HCland its gpl
20:04:56CassandraI assume it still requires net access.
20:05:23HClmost probably, yes
20:05:53HClfor future reference
20:06:05HClin case i forget the url or so
20:08:04 Join XandriX [0] (~slack@xandrix.user)
20:08:11HClseems easy enough, only problem is that the trm is a string, not an int
20:08:14XandriXrockbox wth ?
20:08:17 Quit bagawk ("Leaving")
20:09:44Rorierm...why does rockbox think Ogg is FLAC and MP3 is WAV?
20:10:01CassandraSatan hates you.
20:10:10CassandraOr it might be a bug.
20:10:16Roriit was working a few builds back :)
20:10:34CassandraSometimes fixing things breaks something else.
20:10:55amiconnI don't think the TRM generation itself requires internet. It's the TRM lookup in the musicbrainz database that does, but we don't need that
20:11:16HCli dunno
20:11:19 Part XandriX ("Leaving")
20:11:33HClthis is going to require a total rewrite of the filehash field in the database
20:11:39HClsince the trmid is a pretty long strin
20:11:58 Join Lear [0] (
20:12:24Rorican someone put that in the buglist (Incorrect filetype displayed)? Ta.
20:12:47amiconnHCl: We don't need to use that string as-is
20:13:12HClwhats an 406 response?
20:13:33amiconnIs there an example how such a TRM id looks like?
20:14:03RoriNext track works much much better now
20:14:12amiconnHCl: HTTP 406:
20:14:15RoriI can even scroll it no problemo
20:14:18 Join webguest40 [0] (
20:15:34HCljtrim gives me that response
20:15:44HClString trmId = "69ecc0fa-ef63-46ea-b6b7-531e44d8934c";
20:15:45Rorialthough it sometimes bugs out on Ogg when playing the first track some some Ogg's. Not a big deal though.
20:15:57Rorion some Ogg's rather
20:16:02HClanywho, java can hash a string no problem so thats not really an issue..
20:16:19HCljtrim is failing at the moment though
20:16:29webguest40playback gets very confused if I skip two tracks and then FF the third, it zooms throught the third and then stops
20:16:36HClC:\PROGRA~2\SONGDB~1\classes>java -cp . com.inzyme.jtrm.Main "C:\WINDOWS\Profile
20:16:39HCls\HCl\Desktop\music\01 - Children.mp3" Server returned HTTP response code: 406 for URL: http://trm
20:17:08amiconnHCl: So the TRM is a just a hex string showing 16 bytes (4 longs)
20:18:28CassandraLooks that way, unless the position of the -'s is significant.
20:18:44amiconnMy guess is that the musicbrainz server changed the request requirements somehow. -> Try to extract the hashing code and go without the lookup
20:20:32HClamiconn: then i wouldn't be able to use tagged mp3s..
20:20:36HCli'm looking at this log..
20:20:42CassandraThe TRM is generated by the server from internal hash data.
20:20:59HClthat guy had the same problem
20:21:04HClhe quickly said nevermind
20:21:08HClbut never said how he fixed it
20:21:38amiconnCassandra: Hmm, that's bad
20:21:53amiconnThis would be reason enough for me to drop TRM
20:22:28amiconnWe could use the audio signature part thouh
20:24:04 Join MisticJeff [0] (
20:24:10MisticJeffHi Gang..
20:24:25HClwe can't drop trm cause its WAY too expensive
20:24:43CassandraIt is rather sucky, isn't it?
20:24:43HClit took 15+ seconds on my single test file
20:24:50MisticJeffHC1: please walk this noob through getting the database setup... place SongDB.jar into root then what??
20:25:01HClnot root, .rockbox
20:25:10HCljava -jar SongDB.jar \
20:25:38MisticJeffdo i need to identify the location such as, D:\??
20:25:56crashdhey MisticJeff
20:26:06Coldtoasthey MisticJeff. heh
20:26:16crashdsigh, shouldnt rely on nick-tabbing and being in the right channel
20:26:48MisticJeffhi guys
20:27:08 Join webguest31 [0] (
20:28:35HClgot it
20:28:41 Quit webguest31 (Client Quit)
20:29:01markunI've got unicode working again. The only problem is the fontsize: the one I'm using right now is 180kB.
20:29:06HClC:\PROGRA~2\SONGDB~1\classes>java -cp . com.inzyme.jtrm.Main "C:\WINDOWS\Profile
20:29:10HCls\HCl\Desktop\music\01 - Children.mp3"
20:29:15HCl :)
20:29:48LearHuh? Where's the function set_rating?
20:29:52HClthe only problem is that this heavily relies on an external internet server
20:30:21Cassandrahcl: to which the source is proprietory/unavailable.
20:30:28HClCassandra: yea, heh.
20:30:32CassandraThis kind of sucks, like ami says.
20:30:46HCli know
20:30:51HClbut i got it working anywho
20:30:58amiconnmarkun: Font caching (!!!!!!!)
20:31:02MisticJeffguess i'll wait a bit... if anyone can walk me through that tag db thing step by step please email me jeff at
20:31:12markunamiconn: I know, but it's not so easy.
20:31:13 Part MisticJeff
20:31:26HClwhy didn't the \ work for him?
20:31:59amiconnmarkun: There's a patch for supporting chinese, on archos. Maybe you can take some ideas from that
20:32:02LearOdd... Wonder why I didn't get that through cvs update (or how I managed to get rid of it while hacking in screens.) :)
20:32:12 Join hardeep [0] (
20:32:37markunamiconn: right now every glyph that is not in the font is replaced by the default glyph. Would be useless to load a range of glyphs in the cashe if the are all just copies of the default glyph.
20:32:59amiconnYes of course
20:33:58amiconnThe caching would be necessary for fonts with more glyphs than which fit into the font memory space. We would only load the glyphs which are used
20:34:30LearGah, not even cvs diff picks up that set_rating is missing...
20:34:36webguest40Slasheri; the FF is buggy right after skipping a track
20:34:38amiconn...possibly even swapping out some glyphs loaded earlier if we need new, unloaded ones and there is no more free memory
20:34:55Slasheriwebguest40: hmm, what do you mean with that?
20:35:12markunamiconn: Yes, I've thought about all that.
20:35:20webguest40skip two track in a row, then FF as the track starts to play
20:35:31Slasheriok, i will try
20:35:46markunAnd I also have to keep an eye on the speed because looking up a glyph is something that is done a lot.
20:36:17Slasheriwebguest40: if you mean the problem that elapsed counter goes to negative, that has been fixed in the latest bleeding edge
20:36:24markunWhen a line is scrolling it could become a problem with the current implementation
20:36:30webguest40no that that problem new one
20:36:38amiconnStill, caching is simply necessary if we want true unicode support, that works for chinese/japanese/korean etc
20:37:15Slasheriwebguest40: are you using crossfade? and the latest bleeding edge build?
20:37:22amiconnmarkun: Not necessarily. We'd need to preload the glyphs for all characters in the scrolling line
20:37:26webguest40yes latest, no crossfade
20:37:34Slasheriok, i will turn crossfade off
20:38:29Slasheriwebguest40: and you are using mp3 files?
20:38:36webguest40yes mp3
20:38:40amiconnSome educated guessing on what glyphs might be needed might be needed as well. E.g. the ascii glyphs should always be available etc
20:38:49Slasherii tried start playing, then skipped few tracks and did FF.. no problems so far
20:38:56markunamiconn: For single quotes I use u 0x2019 in my ogg tags. Right now you cannot have it in a font without having the other 0x2018 as well (most of them copies of the default glyph)
20:39:18webguest40you need to start FF as soon as the next track starts playing
20:39:46Slasheriwebguest40: ah, thanks i found it :)
20:39:50CassandraSlasheri: You got around to looking at why the resume code doesn't clear itself properly at the end of a playlist yet?
20:39:53amiconnmarkun: The default glyph should only ever get cached once
20:40:02markunamiconn: yes, it should
20:40:02 Quit TCK- (Read error: 104 (Connection reset by peer))
20:40:03SlasheriCassandra: no, haven't looked that yet
20:40:10hardeepCassandra: are you seeing that in the latest build?
20:40:17hardeepCassandra: it was working fine for me
20:40:19amiconn...regardless how many code positions use it
20:41:01 Join TCK [0] (
20:41:11markunamiconn: I had some ideas but they involve a lot of pointers and with 4 bytes each it got very big.
20:41:14hardeepSlasheri: but there is a problem with resume −− offset is being updated too early (not taking latency into affect?) so we're resuming a few seconds later then where we stopped
20:41:36godzirraHrm.. I'm getting all sorts of odd problems now.
20:41:37amiconnmarkun: Yes, pointers for every glyph might not be the right solution
20:41:47godzirrawhen I change songs its back to about 4 seconds when I skip forward one song...
20:42:01godzirraand it plays part of the next song and "slips" back into the last song for half a second then plays the rest of the new song...
20:42:02amiconnAgain, I recommend looking at the chinese rockbox patch, perhaps there are some good ideas...
20:42:36markunI have the patch here, I'll take another look later.
20:42:55HCland then we should have a hash based on trm
20:43:05HClif −−usetrm is used
20:43:05HClthat is
20:43:12Slasherihardeep: Hmm, that's true..
20:43:44 Join thegeek [0] (
20:45:28Slasherigodzirra: that problems seems to appear with crossfade disabled.. will be fixed :)
20:46:28 Quit yyz (Connection timed out)
20:47:48HCltrm is slowwwwwwwwwwwwwwww
20:48:29*HCl stares at his +- 140 songs testcase..
20:48:37CassandraHCl: But if you write it to the ID3 structure, you only ever need to calc it once.
20:48:41markunHCl: It uses the decoded audio to make the trm tag, so it's not strange that it's slow.
20:48:52webguest40how would it take to do 37gigs
20:48:55HClCassandra: yea, i know... i don't have the infrastructure to write to the id3 tag at the moment though..
20:49:25HCli'll go see if my backend supports it and if not find a backend that does
20:49:35Cassandrawg: Quite a while. (Hours, I imagine.)
20:50:04CassandraI reran it on my entire music database a while back and it took a good 30 mins or so.
20:50:11CassandraFor about 4000 tracks.
20:50:34Hadakadoesn't trm calculation also require a network roundtrip
20:50:49CassandraBut that isn't the bottleneck.
20:51:11Hadakaoh, I always thought it was
20:52:51Hadakathe entire tcp session took 0.7 seconds here for a single file - 200ms for the actual POST to get a reply
20:53:25CassandraWell, you can calculate the next one in the background while the lookup is going on.
20:53:56Hadakawell yeah, the entire procedure spends close to 4 seconds real time, so it clearly is not the bottleneck
20:54:39 Join ghode|afk [0] (
20:54:46HCl30 minutes for 4000 tracks? seems fast to me
20:55:00HClits doing 7 minutes for just 170 tracks here
20:55:04HCland thats on my local harddisk
20:55:08 Join CheeseBurgerMan [0] (~Dan@
20:55:14HClon top of that.
20:55:21HClits even being a worse hash!
20:55:37HClbut that might be due to me hashing the hash
20:55:54Hadakaworse hash?
20:55:55HClit found two duplicates that are not duplicates
20:56:13HClit did properly find duplicates of two files with different id3v2 tags
20:56:24Hadakatrm collisions are rather common in my experience
20:56:27 Part webguest40
20:56:41HClthen whats the point of trm >.<
20:56:52HadakaHCl: to detect what *song* it is, not what *mp3* it is
20:57:02HClyes thats what we're trying to do
20:57:08Hadakaand different encodings vary so much that it is bound to be difficult
20:57:24HCli will check the trm hashes..
20:57:34Hadakayou don't want to detect different encodings of the same song?
20:58:35HCli know why it failed
20:58:38amiconnHadaka: If network roundtrip is a problem, parallelize the requests
20:58:43HCljtrm only works for mp3s so far
20:58:45amiconn(makes for a nice DoS)
20:58:46HClthe duplicates it got were oggs
20:59:15Hadakasorry, I entered the conversation midway, so I don't know what people are trying to accomplish
21:00:00 Quit thegeek_ (Read error: 110 (Connection timed out))
21:00:01amiconnARgh, archos fm/v2 rockbox isn't far away from the final size barrier :(
21:00:34 Quit ghode|afk ()
21:00:43Cassandra*nods to amicon* We might start having to make Rockbox light for the Archos.
21:01:07LearSlasheri: another playback bug (in build from yesterday at least): at the end of a playlist, playback stops when the codec buffer is empty; it doesn't keep playing until the end of the PCM buffer...
21:02:56Hadakaanyway, like I said, I have gotten numerous TRM collisions from my mp3 collection - and also different encodings generate different TRM ids so I end up with multiple TRMs for the same song
21:03:13*HCl kicks trm and song hashes :/
21:03:27HadakaTRM is just a heuristic
21:03:38CassandraLike we already said. There are no good solutions to this problem.
21:03:41HClheh. for a heuristic its pretty darn slow
21:03:53amiconnCassandra: There is quite a number of spots where we can decrease code size. Spots where we use duplicated (or nearly duplicated) code now, that can be converted into one function...
21:04:13HadakaI personally detect my mp3's (as I want to differentiate by different encodings) by taking an md5 sum of the first 30 seconds of the decoded PCM audio data
21:04:27HClyea, thats what i'm wanting to do
21:04:34HClbut i'm getting really tired of this
21:04:41HadakaI could take the MD5 sum from the mpeg frames themselves, but this way it works for every audio format that is deterministic in decoding
21:04:42HCli'm spending way too much time on it..
21:05:06Hadakaand my point is just to detect mp3's that have had garbage prepended or suffixed or which have been truncated
21:05:33HCli think i'll just throw the musicbrainz approach away
21:05:43HClmainly because it depends on an external internet server
21:05:46amiconnHadaka: Doing that won't allow to hash on the target later.
21:06:21amiconnDecoded PCM will differ a little depending on the decoder used for lossy formats
21:06:57Hadakathe musicbrainz thing is nice to interactively detect which song a certain mp3 is - I think it as a layer on top of the mp3 detection
21:07:06CassandraI don't see why we can't rely on tag info and say mistagged files are the user's problem. Garbage in, garbage out.
21:07:25Hadakaamiconn: really? I know that's true when the stream has errors since decoders handle errors differently - but do they handle valid data differently as well?
21:07:42amiconnYes, definitely.
21:08:40Hadakaamiconn: ah right, then I have to ditch this approach in the long term (and make sure I use the same decoder always in the short term)
21:08:40amiconnThe first thing are roundoff errors. They depend on the algorithm, the math precision, whether they used fixed or floating point...
21:09:06amiconnThe next thing is that decoders may sacrifice precision for speed
21:09:16godzirraSlasheri: just making sure someone new about the probs :)
21:09:35LearSlasheri: that mp3 metadata thing I mentioned yesterday (if you saw it...): I just uploaded a patch fixing that (#1232957). Feel free to apply it. :)
21:09:48Hadakaamiconn: right
21:10:33 Join LinusN [0] (
21:10:46LearHrm... Preparing multiple patches, with changes in the same file, is no fun... :)
21:10:59HadakaLear: quilt is your friend
21:11:16CassandraHuh? cvs diff -u does it all for you.
21:11:51LearHadaka: Huh? What has quilt got to do with it. Goggle didn't show anything else that seemed reasonable...
21:11:59HadakaCassandra: except if you want to have a new patch on top of your old patch?
21:12:06LearCassandra: But here I have two different patches, both meddling with the same file...
21:12:14CassandraAh. Messy.
21:12:31HadakaLear: quilt makes handling a patch series easy, and makes it easy to rebase them and refresh them and such
21:13:31amiconnThe Win32 simulator is messy (less messy than the x11 sim, but still...)
21:17:29Rorican anyone tell me what I am doing wrong for next track in wps? I want it to pull info from the ID3 tag but it's not working. I can only pull the filename or directory name
21:19:05Rorioh ignore that I think I know what I am doing wrong (Impatient)
21:19:09CassandraAre you putting it on scrolling lines?
21:19:41RoriI keep forgetting the 2mb buffer up rule
21:19:55Coldtoastit works on scrolling lines Cassandra
21:20:13Rorithe new build has a quick fix to show filename and directory info...lets see if it changes after the 2mb buffer up
21:20:27Cassandract: If it does, that's pure luck.
21:20:39CassandraScrolling WPS lines don't support dynamic update.
21:20:51Coldtoastjust the first time you play a track and it has the info, it doesn't update the wps with the info til you seek or change tracks
21:20:57SlasheriLear: thanks, i will apply it :)
21:21:18Coldtoastjust that firsttime. After that, when you just let a playlist play, it updates correcrly
21:21:25HCli need to talk to preglow..
21:21:41HClor anyone else who knows how to get the offset of the first mp3 frame
21:22:17*HCl stares.
21:22:23Coldtoast%s%?Ia< %Ia| %Fn>
21:22:23Coldtoast%s%?It< %It>
21:22:25HClrockbox just has a field with the offset.
21:22:36*HCl goes to figure out how this field is calculated
21:23:15 Join TCK- [0] (
21:23:16Coldtoastthat's what I use. First track you play just displays the filename, which is instant. It works nicely for me
21:23:22SlasheriLear: what is your realname?
21:23:29hardeepHCl: you're not referring to the resume related offset are you?
21:23:34hardeepthat's something different
21:23:39Coldtoastjust the first track you play gets the filename instead. That's ok for me
21:23:45HCl entry->first_frame_offset = bytecount;
21:25:57LinusNgotta go, cu l8r
21:26:00 Part LinusN
21:28:06HClthat was easy, says man, and he procedes to prove black is white and gets killed at the next zebra crossing
21:34:05 Quit TCK (Read error: 104 (Connection reset by peer))
21:35:05HClhm, well, not as easy as i originally though.
21:40:11HCli don't know about ogg vorbis comments
21:40:22Learwhat about them?
21:40:23HClbut it skips id3v2 now and hashes the first 32kb of the first frames
21:40:31HCldunno how to skip them
21:41:32Learnot trivial, but if you read the vorbis spec, you will be enlightened. :) Short version: skip first two Ogg pages (or rather, first ogg page, and then first vorbis packet on the second page).
21:41:42LearThough I doubt that made much sense to you. :)
21:43:47HCli hate song hashing..
21:44:02HCli added the beeping skipping of id3v2, why doesn't it work..
21:45:35HClit doesn't seem like tritonus is very reliable
21:45:38HClbut i knew that already..
21:48:02HClwhy can't all song tags be at the end?
21:48:38zebecause if you're streaming it then you dunno wtf it is till its over
21:49:23***Saving seen data "./dancer.seen"
21:49:31ze...and if your file's truncated
21:49:39zethen you never know wtf it is
21:49:40HClwell i'll just put this on the wiki anyways
21:49:49HClits *slightly* better
21:57:27amiconn:) :) :) 4-level greyscale in the win32 simulator :) :) :)
22:02:26*HCl scribbles some documentation on how to use the java tool onto the wiki
22:02:38*HCl is still annoyed over song hashing..
22:02:51HCltritonus is a really bad backend..
22:03:04HCli should replace it :/
22:04:32HClt0mas: any chance that you might be able to implement a line drawing tag to the wps?
22:06:07HClamiconn: does rockboy utilize the 2bit grayscale yet?
22:06:17amiconnYes it does
22:06:23amiconn(In fact it has to
22:06:24HClk good :) enabled the define then?
22:06:27HCl :p k
22:06:34amiconnNo, changed the define
22:06:42HClwell, that works too
22:06:49amiconn#if LCD_DEPTH == 2
22:07:17*HCl ended up all unmotivated after messing with hashing for 4 hours without much result, goes to wander around :/
22:07:26amiconnThe only thing missing is a nice 4-grey ROCKbox logo
22:07:37HClmmm, why?
22:07:54amiconnI could commit without, but I don't want to
22:07:58HCl :p
22:08:08HClwhat about wps images?
22:08:35amiconnThe current b&w logo works with 4-grey as well, but obviously in b&w only. Dull.
22:09:15amiconnwps images should work the same as now, in b&w
22:09:32amiconnThere won't be much visible difference at the start
22:10:06amiconnOnly (1) rockboy uses 4-grey (2) The splash() boxes will have a light grey background (3) The logo, if I manage to do that
22:11:11 Quit RotAtoR ()
22:13:22 Join PaulJ [0] (
22:15:17t0masdoes anybody know what the english expression for ehm... "school french" or something like it is?
22:15:22t0masso really bad french?
22:15:38t0masin dutch: Huis tuin en keuken frans -> house garden and kitchen french
22:16:28hardeepschool is ecole in french iirc
22:16:43RoriI can't seem to find the perfect wps yet. Sigh...
22:17:14Stryke`anychance of that allignment patch for wps being committed?
22:18:00HClt0mas: there isn't one, afaik
22:18:06hardeepStryke`: someone was looking at it (Cassandra?)
22:18:43t0masStryke`: Cassandra is looking at ti
22:18:47Stryke`ok, sounds good
22:18:51*HCl goes to look at the submitted patches
22:18:52t0masand it will be committed tonight, or tomorrow
22:19:01Stryke`great news, thanks">
22:20:14HClnice :)
22:20:22HClfirst time i see rockboy run in a sim
22:23:31amiconnIn fact it should be more playable in the sim than on target
22:23:45amiconn(1) No speed problem (2) No button constraints
22:24:52 Quit PaulJ (Read error: 145 (Connection timed out))
22:26:23*HCl looks at the patch for the chip emulator
22:29:11LearHmm... Does anyone else get the wrong codec name in the WPS?
22:32:09HCltoo bad, missing file from the chip8 patch..
22:32:13*HCl registers for an sf account..
22:32:24LearYep, id3.c and id3.h aren't in synch...
22:41:50 Join tvelocity [0] (
22:44:30BagderLear: you want your cvs access fixed/restored?
22:45:17LearBagder: Linus fixed that about an hour ago. I've even started comitting... :)
22:45:25LearAh, you noticed...
22:45:26BagderI _just_ noticed that
22:45:38Bagderwhen reading mails waiting for your reply ;-)
22:46:01Bagderback to lurk-mode
22:46:11LearBagder: Btw, you've seen my oggvorbis patch? Needs updates, but is it still of interest?
22:46:32 Quit Febs ("Chatzilla [Firefox 1.0.4/20050511]")
22:46:37BagderI haven't checked it out, I trust your judgement
22:50:23HClBagder: added playcount and rating to wps.. if you're interested..
22:50:42BagderI noticed and I am interested, I'll upgrade and try it out soon
22:50:49HClyou're gonna have to regenerate your database though, there was a critical bug in the binary search..
22:51:44BagderI made mine with the perl script
22:52:29BagderI didn't see any updates to that, was it?
22:53:16HClthe perl script isn't gonna be capable of runtime database till it gets a hashing function
22:53:39Bagderand that is not gonna happen until you describe what it is and how it works
22:53:58HClvery simple, just a crc32 of the first 32kb of the files
22:54:34Bagdercan you from now on please clearly describe all file format changes
22:54:48HCli haven't made any file format changes..
22:54:49Bagderwe have two tools now, soon to be three
22:55:14Bagderwell, you made up the format and it isn't easy to track devel nor format
22:55:28HCl*checks the wiki*
22:55:38HCli'll add a comment about the hash field
22:56:06CassandraStryke`: Alignment patch is in my local source tree. I'm fixing formatting issues at the moment.
22:56:49Stryke`does it allow for partial formatting? alligning left some text and right others on the same line or not yet or not feasible
22:57:09CassandraThat's actually something I'm working on fixing too.
22:57:23Stryke`sounds great
23:00:06 Part Lear
23:01:14Rorimaybe someone will eventually make an app for designing wps in...maybe :)
23:01:23 Quit tvelocity ("Leaving")
23:02:02RoriJust been playing with trying to get bmp's to sit right with the text
23:02:12Rorifont sizes mess with that something awful
23:02:37HClBagder: pretty much, the perl tool has to implement">, and use CalcCRC32 on the first 32768 bytes of each file
23:03:24Bagder"and" ?
23:03:30Bagderisn't that code crc32?
23:03:52Bagdernow I see
23:03:57BagderCalcCRC32 is the name of it
23:06:15HCli marked the spot in the perl tool where the hash should go
23:08:30Bagderand the runtimedb uses that to figure out what song it plays?
23:10:06HClit uses that to link a song to a runtime database entry
23:10:32HClso if two files have the same hash they will also have the same (shared) runtime info
23:10:53HClso if you move files you won't lose your runtime info, among other things
23:13:41HCli'd better go sleep
23:14:44amiconnBagder: Where can I find that super-hires rockbox logo bitmap?
23:15:11HCltomorrow i'll rewrite the databasev2 wiki a bit to reflect the way things are now
23:16:14amiconnThanks... so it was in cvs once...
23:17:00BagderHCl: and things no longer correct should be removed from the TagDatabase I think
23:17:08Bagder(the wiki page I mean)
23:17:15Bagdergood night
23:17:16HCli'll do that too
23:23:55BagderHCl: I guess it is up to 32K and less if the song is shorter?
23:31:38*Cassandra swears at the WPS display code.
23:34:27*Rori swears with you
23:34:30 Join einhirn [0] (
23:39:07*t0mas swears with you both
23:39:21t0masCassandra: maybe we can try to move some things arount
23:40:33 Join stripwax_ [0] (
23:45:15 Part CheeseBurgerMan
23:46:18 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
23:47:04west-acrehey. there's a problem wit my WPS. it used to work, and now. the retreiving next song data doesn't disappear when the data is received. its the GeorgeCollins on the wps gallery :S
23:47:11Cassandrat0mas: It's OK. It's just very difficult to do multiple aligned strings on a single line with the current structure. I'm trying to think of a good way to solve this and failing.
23:49:27***Saving seen data "./dancer.seen"
23:52:04CassandraNo idea, sorry.
23:53:54stripwax_Hmm.. does the CF5249 have a branch predictor?
23:54:13CassandraNo idea
23:55:02stripwax_west-acre - between which versions of rockbox did it work and not work? ;-)
23:56:18amiconnstripwax_: A very simple one
23:56:50amiconnIt assumes backward branches are most likely taken, and forward branches most likely not

Previous day | Next day