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 2005-11-02

00:16:07 Quit Midgey34 (Remote closed the connection)
00:21:17preglowworks just fine here
00:21:40preglowcompiled again, works just fine
00:21:50preglowi'll try a make clean and see if it's a dep problem
00:22:44preglowi know what it is
00:22:55preglowi forgot to commit the macsr change
00:23:43preglowi've placed it in NeAACDecInit2, do you think that'll suffice?
00:23:50preglowany function that is always called will do
00:25:04 Quit ender` (Read error: 113 (No route to host))
00:28:37preglowlinuxstb: try it now
00:29:39preglowperhaps NeAACDecOpen would have been a better place?
00:32:19preglowi've also disabled SSR, LD and LTP locally, but this results in: no major code size changes, and several warnings in the code
00:32:51 Join ashridah [0] (
00:35:57 Quit DangerousDan ("Miranda IM! Smaller, Faster, Easier.")
00:40:08 Quit paugh ("brb")
00:40:18linuxstbpreglow: Yes, that's fine again now.
00:40:52 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
00:42:11linuxstbIt's definitely more than 50% realtime now.
00:43:17preglowthen hooray
00:43:23preglowi'm looking into having it output 32 bit numbers now
00:43:25linuxstbSSR is already disabled in libfaad - see the "#ifdef FIXED_POINT" lines in common.h. "MAIN" is also disabled.
00:43:28preglowand deinterleaved
00:43:42preglowso dsp doesn't have so much work to do
00:43:58linuxstbIs the output buffer in IRAM?
00:44:21 Join DangerousDan [0] (
00:44:30preglownot yet
00:44:33preglowthat is, it is here
00:44:35preglowbut not in cvs
00:44:44linuxstbBut that doesn't help the speed much?
00:44:51preglownothing seems to help the speed much
00:45:16 Quit DangerousDan (Read error: 104 (Connection reset by peer))
00:45:19preglowso i've just diverged into something i know will help: 32 bit deinterleaved data
00:45:40 Join Sandking [0] (
00:51:34preglowi'm not THAT far from realtime for 128kbps here
00:52:12 Quit tvelocity ("Leaving")
00:52:42linuxstbHow much of the IRAM are you using now?
00:52:45preglownot veru
00:53:09preglowabout 30k
00:53:16preglowmore or less arbitrarily allocated
00:53:26preglowoutput buffers are iram, at least
00:53:35preglowbut that might not be the wisest choice for this code
00:53:39preglowcodec <-
00:54:09preglowwhen you did your iram allocating, did you have any reason for using it the places you did, or was it completely random?
00:54:17preglowjust so i know i can remove them
00:54:22linuxstbcompletely random.
00:56:25preglowit doesn't have any option for non-interleaved audio, it seems
00:57:24TiMiDhi there
00:57:58TiMiDI just tried to add a new file to CVS (cvs add) and now when I commit I get this error :
00:58:01TiMiDcvs update: warning: `textarea.c' was lost
00:58:13TiMiDwhat does that means ?
00:58:49preglowthat you've deleted it
00:58:57preglowahh, riight
00:59:05preglowit deletes it, then updates it to get the id string right
00:59:15TiMiDhmm ok :)
00:59:22preglowyou probably commited it with just $id:$ in the header, right?
00:59:27TiMiDso everything is ok ;)
00:59:30TiMiDyes :)
00:59:36preglowyes, and now it's magically filled out ;)
00:59:42TiMiDhuhu :)
01:00:36preglowwhat the flaming hell is this?????
01:00:49preglowfloating point math in our output routine?
01:00:55TiMiDuh ?
01:01:25preglowforget it, just me being stupid
01:04:43 Quit cYmen ("zZz")
01:05:01linuxstbTiMiD: onplay.c is still utf-8 :)
01:06:36preglowthere's a bit of processing in the output driver i'll just go ahead and bypass
01:07:08linuxstbpreglow: Where is that?
01:07:55preglowlinuxstb: output.c
01:08:50TiMiDmy editor must be cursed, that's the only possibility :)
01:09:52TiMiDouch a warning
01:10:04linuxstbpreglow: Pretty obvious place for it.
01:11:09preglowi'll just bypass the entire thing
01:11:12linuxstbCan we set FAAD_FMT_FIXED somewhere to make it output native format?
01:11:38preglowin practice, that's what i'm doing
01:12:07linuxstbSo are you bypassing that function completely?
01:12:53linuxstbI forgot that AAC is used for multi-channel sound sometimes. But I don't think we need to support that.
01:13:38preglownow,let's see
01:13:42preglowlinuxstb: nor do i
01:13:52preglowlinuxstb: i'm bypassing it completely and just accessing time_out as is
01:14:42 Quit dpassen1 ()
01:16:33preglowa wee bit closer to realtime again with that
01:17:42linuxstbDo you still copy the data, or do you just return the data directly from time_out?
01:18:36preglowi use time_out directly from the codec plugin
01:18:39preglowjust to see if it works
01:18:45preglowi don't know if i'll bother to do anything cleaner
01:19:03linuxstbI don't mind cleaning that up. Is time_out in IRAM?
01:19:12preglowit is here
01:19:23preglowbut i got some nice clipping free with this one
01:19:24preglowi wonder why
01:19:43preglowi've updated SET_SAMPLE_DEPTH to 29
01:19:51preglowbut it clips at around 0.2 full scale
01:21:02preglowplus, i did some assumptions when putting time_out in iram, namely: no more than two channels, and max frame size is 2048
01:21:27preglowhmm, or 1024, it seems...
01:21:36***Saving seen data "./dancer.seen"
01:21:38preglowwhich must be perfectly valid, seeing as how it works
01:22:51 Quit muesli_- (Read error: 110 (Connection timed out))
01:25:01preglowseems clipping limits aren't automatically set
01:25:04preglowi wonder why
01:32:06preglowlinuxstb: but ok, what i've basically done now is comment away the Out_To_PCM call in aac_frame_decode, and then just access hDecode->time_out directly from aac.c
01:33:06linuxstbI think that's fair enough.
01:34:20linuxstbSo that means we don't need to malloc sample_buffer ?
01:35:01preglowit's never used, so no
01:35:23preglowiram or no, i can see no reason for us to not just simply access time_out as is
01:35:26preglowwe have no conversion needs
01:35:39linuxstbI agree.
01:36:30preglowTiMiD: what's up, i can't build anymore
01:37:03preglowgui/statusbar.h:113: error: `NB_SCREENS' undeclared here (not in a function)
01:37:29TiMiDit compiles very well here
01:37:31TiMiDlet me see
01:37:43preglowi just did cvs update, make clean, and reconfigure
01:37:56linuxstbSame problem here.
01:37:57Moosred builds
01:38:27TiMiDI just added a .h
01:38:42TiMiDand corrected a UTF-8 pbl
01:39:10preglowdo a cvs diff and check what's different on your part
01:39:13preglowit's got to be something
01:39:50TiMiDI try to recompile on my box
01:40:22TiMiDbut look my other CVS changes just did some yellow because I forgot a .h (my gcc didn't told me that)
01:41:16TiMiDI'm investigating this
01:42:03preglowNB_SCREENS is defined in screen_access.h, so it obviously isn't included when it should be
01:42:09preglowi can't imagine something else being wrong
01:42:15TiMiDI see
01:42:17preglowi'll just go brush my teeth in the meanwhile
01:42:23TiMiDit's a circular include problem
01:47:52preglowany fix forthcoming?
01:48:00TiMiDno :/
01:48:06TiMiDI've spotted the pbl
01:48:20preglowyou really need to test everything before you commit
01:48:25TiMiDanyway you just have to remove the #include "textarea.h"
01:48:31preglowif you can't fix it now, revert it until you can fix it
01:48:42TiMiDin screen_acces.h if youi need to work
01:49:00TiMiDbut I'm a little lost here
01:49:03preglownono, if you're not going to fix it now, you need to revert your changes
01:49:21TiMiDok how do I do this ?
01:49:38TiMiDif I revert my changes there will be a warning
01:49:47preglowa warning is better than a broken build
01:50:12preglowyou revert things more or less just by removing what you commited, then commiting again
01:50:16preglowat least that's the only way i know of
01:50:35 Quit DMJC-sleep (Read error: 110 (Connection timed out))
01:50:58linuxstbTiMiD: Just remove that troublesome include from your local copy and commit that change.
01:51:05linuxstbI agree that yellow is better than red.
01:52:21TiMiDthe problem is that there are other changes in my local copy
01:52:29linuxstbI'll commit it then.
01:52:38TiMiDI'm testing if it doesn't break anything (it shouldn't) and I commit)
01:53:00preglowif you can't, there's no trouble for us to do it
01:53:04preglowit's just a matter of removing a line
01:53:41preglowbut didn't this show up in your build, or didn't you test it at all before commiting?
01:53:53preglowif it's the last: don't do it again ;)
01:54:02TiMiDit's the last :p
01:54:10preglowyes, everyone needs to get burnt once on that
01:54:12TiMiD(I didn't recompiled it for a small include :) )
01:54:42preglowbut get in the habit of building everything, even the smallest change
01:54:44TiMiDeverything is working well with my last changes so I commit
01:56:32TiMiDanyway the problem is still here :(
01:56:44TiMiDI need to find a way fo fix this warning
01:57:18preglowsure, and we'll allow you to do that in peace and quiet as soon as we can continue on our coding :)
01:57:22linuxstbYes, it compiles fine now - with that single warning.
01:57:47TiMiDonce again I'm dumb :(
01:58:05TiMiDI included the .h in screen_access.h instead of .c :p
01:58:12preglowso it's fixed?
01:58:21TiMiDI hope
01:58:26preglowthen hooray!
01:58:30TiMiDbut I have no way to see if there is a warning
01:58:39TiMiDsince gcc doesn't warns
01:58:47preglowwhat gcc version ?
01:58:52preglowit warns here
01:59:12preglowi've got 3.4.1 and that warns
01:59:41TiMiDmaybe because I compile it for simulator
02:00:46preglowwhy do that?
02:00:53linuxstbI think you should always compile for the target before comitting - it's much more important than the sim.
02:01:27linuxstbEven though with the part of Rockbox you're working on, it shouldn't matter - but it obviously does.
02:01:40preglowyou haven'
02:01:51preglowyou haven't got an actual unit?
02:01:56TiMiDI 'll do it by now, I had problems with a plugin working fine on simulator but not on target
02:02:13preglowahh, forget it
02:02:25preglowi'm so used to never using the sim myself, i just assume no one else does as well :)
02:02:51TiMiDIf I didn't owned a unit, I wouldn't be there I think :p
02:03:14preglowohh, it's not the first time someone coded without a unit
02:03:28linuxstbI've started using the sim a little recently (AAC only worked in the sim for a while). It makes debugging a lot easier - DEBUGF is very useful.
02:03:45TiMiDthey code without getting the taste of their work ?
02:04:00TiMiDlinuxstb: yes, the sim also saves your USB cable :)
02:04:46linuxstbIt would be useful for a H3x0 owner to get a the H3x0 sim working. That could speed up development in advance of the bootloader.
02:05:36preglowperhaps i should try making speex work in the sim first
02:05:43preglowi'm so used to being unable to use the sim i don't even consider it
02:06:08linuxstbpreglow: Yes, I think you would find it easier.
02:06:40linuxstbDo you use cygwin or a Linux box for your compiles?
02:07:42preglowremote linux box
02:08:05preglowi've got cygwin installed, so i've got an x server handy
02:08:10preglowaudio will be another issue, however
02:08:50TiMiDI never managed to make audio work on the sim :(
02:09:55preglowi could of course just use a win32 sim and copy it all the time
02:09:57preglownot too much of a bother
02:12:39linuxstbEven without sound, it could still be useful. Once you get to that stage, then the sim probably isn't that useful any more.
02:15:20preglowman, my h120 reacts violently to usb unplugging now
02:15:41preglowflashes up an illinstr, and then the backlight starts _strobing_
02:16:18preglowiriver firmware loaded after restart
02:17:28preglowi just might be writing outside some buffers here...
02:18:25 Nick paugh is now known as AliasCoffee (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
02:19:17linuxstbwtf. Playing a 24-bit WAV files causes the CPU to boost...
02:20:37preglowheavy stuff, that
02:22:56linuxstb24-bit wavs have a higher boost ratio than wavpack...
02:23:38linuxstbI think that proves we shouldn't use the boost ratio to measure CPU usage.
02:23:41preglowany obvious reasons for this...?
02:23:54preglowi've always said we shouldn't do that
02:23:56linuxstbI'm guessing playback.c is confused.
02:25:08preglowwhatever important arrays i put in iram: no dice
02:25:46preglowlinuxstb: those arrays aren't marked const
02:26:15linuxstbWhich arrays?
02:26:53linuxstbThe wav ones?
02:27:04linuxstbTrue. They should have been.
02:29:33preglowtalking libfaad here
02:30:03preglownow i'm using 37k iram
02:30:12preglowsolely for imdct arrays
02:30:42linuxstbHow close are you to realtime now?
02:31:31preglowam i the only one with the usb disconnect glitch?
02:32:42preglowit's a bit hard to say, i can't say i ever hear any improvement
02:33:08linuxstbpreglow: I just got an illegal instruction when disconnecting from USB.
02:33:49preglowthat's what get as well
02:33:52linuxstbYes, it's repeatable.
02:33:52preglowwhat i get as well
02:34:03linuxstbI04: IllInstr at 00000002
02:34:37linuxstbWas it us or TiMiD?
02:34:44preglowi have no idea
02:34:46linuxstbI'm guessing it's a screen problem.
02:34:49preglowi don't see how it can have been any of us
02:35:17preglowat least i don't see how it can have been me ;)
02:35:32preglowbut ok
02:35:35preglowthis is frustrating
02:35:55preglowi just leeched aac.codec from the latest daily build, and there is almost _no_ difference
02:36:00preglowi can't hear any difference at all
02:36:08preglowi'll just toss my aac.codec up so you can have a listen
02:36:18linuxstbYou mean difference in speed?
02:38:48 Quit Moos ("Glory to Rockbox")
02:38:52linuxstbI think it is slightly better - more music between gaps.
02:39:59preglowi somehow expected more than sligtly better after getting rid of the sample conversion and buffer copying, plus placing all important imdct buffers in iram
02:40:21linuxstbBut did you remove other buffers from iram?
02:40:40linuxstbIt's definitely better though.
02:41:01preglowi removed the ones you put there, if that's what you mean
02:41:40preglowright now it contains around five buffers, and each is around 8kb large
02:44:04linuxstbBut at least your sample conversion removes the need for a set of mallocs.
02:45:10preglowsure, but i would have expected more from the iram
02:45:25preglowperhaps i'm just mistaken as to what arrays are actually used the most
02:46:33preglowthere are buffers flying around everywhere, so it's a bit hard to know without reading the code more thoroughly
02:48:29preglowin any case, i think we can just plain forget decoding anything but LC files
02:48:34preglowperhaps main
02:48:50preglowi'd love to decode sbr files, but we can more or less just forget that
02:49:35TiMiDlinuxstb: does the unit crashes ?
02:49:44preglowTiMiD: yep, on usb deplug
02:49:46linuxstbTiMiD: Yes - when removing USB.
02:49:54linuxstbI've just broken the sim :).
02:50:05linuxstbWe're having a bad night.
02:50:59TiMiDI test on my unit
02:53:24preglowi love the fact that DEBUGF and LOGF don't take the api struct as a parameter, like LOGF(ci, "hehe");
02:53:26linuxstbOK, sim should be fixed.
02:53:31preglowthey just simply assume 'rb', which is braindead
02:53:33linuxstbpreglow: I know.
02:53:49linuxstbI'll try and fix that one day.
02:53:59linuxstbAt least it should be "ci", not "rb".
02:54:01preglowa lot of code needs updating if you do :P
02:54:14preglowlinuxstb: well, same mechanism is used for plugins, where they're often called rb
02:54:47TiMiDcrash :( bad
02:57:10*preglow hopes LOGF is still working
02:57:19TiMiDthe crash happens only when dircache is enabled
02:57:39preglowTiMiD: it didn't happen before today, i think
02:57:46TiMiDI don't know
02:58:08 Join actionshrimp [0] (
02:58:09TiMiDwhat should I do ?
02:58:22linuxstbYes, it started late tonight.
02:58:25preglowignore it
02:58:39preglowand ask some of the big boys tomorrow
02:58:47TiMiDhmm ok
02:59:00TiMiDmaybe it's not dorectly related to my code
02:59:07linuxstbYou could try and narrow down which commit broke it.
02:59:11preglowunless you can be bothered to find out which commit broke it
02:59:52preglowi'm pretty sure it's not one of mine, the only core part i've touched is dsp.c
03:00:28TiMiDhow do you get the code source for a given date ?
03:01:11linuxstbI think it's something like "cvs co -D '2005-11-01 23:20' apps"
03:02:23preglowremote support breaks remote logf
03:03:36TiMiDpreglow: just set NB_SCREENS to 1 in screen_access.h
03:04:00TiMiDit will not write on the remote with that
03:04:03preglowTiMiD: isn't there a way for that be done automatically if ROCKBOX_HAVE_LOGF (or whatever it's called) is set?
03:04:30preglowperhaps there should be a new option for remote logf
03:04:46preglowfor people who use the remote and wants logf file writing support only
03:04:57TiMiDmaybe unset HAVE_REMOTE_LCD
03:05:12TiMiDhmm not a good idea
03:05:27TiMiDanyway if you just change NB_SCREENS it will be good
03:06:22linuxstbI think you should set NB_SCREENS to 1 when ROCKBOX_HAS_LOGF. The other case will be the minority.
03:06:45linuxstbThat can be done in screen_access.h
03:07:37TiMiDI'm recompiling a build from this morning
03:07:44TiMiDI will see if it crash
03:07:50TiMiDthe I go to sleep :)
03:08:27TiMiDok the build from this morning doesn't compile
03:08:43linuxstbpreglow: Try changing line 33 of screen_accesss.h to #if defined(HAVE_REMOTE_LCD) && !defined (ROCKBOX_HAS_LOGF)
03:08:58TiMiDlinuxstb: I will commit that
03:09:13linuxstbTest it first :) I don't have a remote, so I'm only guessing.
03:10:05TiMiDanyway I'm very annoyed by the crash
03:10:17TiMiDwhat does the error means ?
03:12:33preglowillegal instruction encountered at an iram address
03:12:34preglowi think
03:13:32TiMiDnever encountered that :(
03:14:27TiMiDwell I saw it once when I compiled rb for a slightly different model of cpu
03:15:49preglowthink i need to start calling it a night
03:16:10preglowi was suspecting my iram arrays were never used, and overwritten with some malloced ones
03:16:14preglowbut turns out that's not the case :/
03:19:11linuxstbpreglow: Does point 4c) in this forum post mean anything to you?
03:19:40linuxstbThe sentence about Rockbox not supporting large delays in LAME headers
03:21:22preglowlinuxstb: well, no
03:21:38***Saving seen data "./dancer.seen"
03:21:41preglowlinuxstb: the only delay i know of in mpeg audio, is encoder delay, and that doesn't vary too much
03:22:15preglowlinuxstb: it doesn't say anything about a lame header that i can see
03:23:09linuxstbThis post could be useful:
03:23:24linuxstbIt refers to "encoder padding"
03:23:28linuxstbas well as delay.
03:23:40preglowlinuxstb: btw, your remote change worked
03:23:43preglowthink i'll commit it
03:24:53preglowlinuxstb: right, they call that delay
03:25:11preglowlinuxstb: in that case, he's right, i always did wonder if delays that large existed
03:25:15preglowand now i know
03:26:38 Join pokecapn [0] (
03:27:07pokecapnI want to make a really simple change to the code, should I just bug a developer or should I figure out how to do it myself
03:27:50linuxstbpokecapn: What's the change?
03:28:11pokecapnincreasing the maximum encoder delay that's allowed in the lame tag
03:28:25linuxstbhehe. We were just discussing that.
03:28:27pokecapni'm assuming the code there was ganked from foobar 0.8.3 but foobar does it wrong
03:28:31preglowpokecapn: you'll need more than just a one line change
03:28:41pokecapnwell if i wanted to do it properly yeah
03:28:55preglowi was planning to fix it tomorrow
03:29:27pokecapnthe proper way would be removing that line entirely and looking at the crc, but i have no idea how to do that so i was just going to increase the limits in the if statement
03:29:33pokecapnbut yeah if it's being worked on already that's great
03:29:33preglowno, not just that
03:29:40pokecapnwhat else?
03:29:48preglowthe code just supports skipping one frame at the moment
03:29:51preglowone frame is 1152 samples
03:29:53 Quit actionshrimp ("a bird in the bush is worth two in your house")
03:30:07pokecapnoh i see
03:30:09preglowbut like i said, i'll see about fixing it tomorrow
03:30:34pokecapncarry on then
03:30:35 Quit pokecapn (Client Quit)
03:31:03preglowi need to verify that mpa.c does proper gapless as well
03:31:09preglowi suspect it doesn't
03:31:56linuxstbI wonder if foobar attempts gapless AAC.
03:33:07preglowgod knows
03:33:15preglowif it does, you should be able to so it from the plugin source
03:39:37preglowbut i need to call it a night
03:39:38preglowlater, all
03:41:38linuxstbme too. night.
04:56:33 Join Lost-ash [0] (
05:03:51 Quit ashridah (Nick collision from services.)
05:03:58 Nick Lost-ash is now known as ashridah (
05:11:39 Quit AliasCoffee ("zzz")
05:21:39***Saving seen data "./dancer.seen"
06:36:02 Quit RotAtoR ()
07:21:43***Saving seen data "./dancer.seen"
07:23:22 Join ender` [0] (i=ychat@
07:38:27 Join Kohlrabi [0] (
07:47:28 Join B4gder [0] (
07:49:54amiconnB4gder: Could you perhaps have a look at fixing ?
07:50:09B4gderyeps, I will
08:09:48 Quit ashridah ("Leaving")
08:12:46 Quit Kohlrabi (Read error: 104 (Connection reset by peer))
08:17:43 Join solexx_ [0] (
08:29:39 Quit solexx (Read error: 110 (Connection timed out))
08:30:53 Join ashridah [0] (
08:42:52 Join einhirn [0] (
08:55:56 Quit Rick (Read error: 104 (Connection reset by peer))
08:57:51 Join Rick [0] (
09:07:27 Join _FireFly_ [0] (
09:20:25 Join Kohlrabi [0] (
09:21:47***Saving seen data "./dancer.seen"
09:27:23 Join Zagor [0] (n=bjst@pdpc/supporter/sustaining/Zagor)
09:30:04 Quit ryan_j ("leaving")
09:30:34 Join DangerousDan [0] (
09:49:01 Join DMJC [0] (
10:02:09 Quit ghode|afk (Read error: 110 (Connection timed out))
10:13:20 Quit Seed (Nick collision from services.)
10:13:25 Join Seedy [0] (
10:22:02 Nick solexx_ is now known as solexx (
10:45:46 Join cYmen [0] (
10:46:27 Quit Kohlrabi ("Leaving")
10:50:57 Join cYmen_ [0] (
10:56:07 Join arkascha [0] (
10:56:11 Join cYmen__ [0] (
10:59:26 Join thegeek_ [0] (
10:59:27 Quit thegeek (Read error: 104 (Connection reset by peer))
11:01:22 Join cYmen___ [0] (
11:01:22***Alert Mode level 1
11:01:22DBUGEnqueued KICK cYmen
11:01:22DBUGEnqueued KICK cYmen_
11:01:22***Alert Mode level 2
11:01:22DBUGEnqueued KICK cYmen__
11:01:22DBUGEnqueued KICK cYmen___
11:01:22***Alert Mode level 3
11:03:45 Quit cYmen (Connection timed out)
11:06:32 Join cYmen [0] (
11:06:32***Alert Mode level 4
11:06:32***Alert Mode level 5
11:06:32DBUGEnqueued KICK cYmen
11:06:32***Alert Mode level 6
11:06:32***Alert Mode level 7
11:06:32***Alert Mode level 8
11:07:30 Join tvelocity [0] (
11:08:46 Quit cYmen_ (Connection timed out)
11:11:18 Join Paul_The_Nerd [0] (
11:13:45 Quit cYmen__ (Connection timed out)
11:14:40 Join Kohlrabi [0] (
11:16:33***Alert Mode OFF
11:18:50 Quit cYmen___ (Connection timed out)
11:21:50***Saving seen data "./dancer.seen"
11:40:23 Join ghode|afk [0] (
11:44:36 Join lamed [0] (
11:45:29_FireFly_what's going on with this cYmen guy
11:46:42 Quit lamed (Client Quit)
11:47:04ashridahcrappy isp,
11:47:09 Join lamed [0] (
11:47:33lamedwho is working to complete the new gui?
11:48:32 Join Mxm`Pas`Bien [0] (
11:49:29_FireFly_what is with gui_list ??
11:49:45lamedwell, it'
11:50:18lamedit's not complete isn't it. it doesn't use it to show the playlist nor menus. and i guess that is being worked on.
11:51:29_FireFly_gui_list widget itself is complete but not all gui elements use a widget like gui_list currently
11:51:52_FireFly_i'm currently trying to dig into wps to create a wps-widget
11:52:31_FireFly_lamed: do you use the latest cvs-version ??
11:52:51_FireFly_because the menus are also "ported" to the new gui-system
11:53:34_FireFly_only showing and changing settings on the remote doesn't work currently but the menu's itself woks on the main-screen
11:53:46lamedit is indeed corrently missing. i can assume that every list will use the gui_list? like the menus? firefly - acctually no. I'm just asking a bunch of questions to make sure that my in progress patch is heading in the right direction
11:54:16_FireFly_which patch ??
11:55:08lamed(sssh... a seceret... it's working on my iriver but it's not totally done yet...)
11:57:22lameddo you think it is apropriate
11:59:06 Join LinusN [0] (
12:00:10 Join webguest82 [0] (
12:00:21webguest82Is hard to connect by irc program.
12:01:44 Quit ghode|afk (Read error: 110 (Connection timed out))
12:05:02 Quit Maxime (Read error: 110 (Connection timed out))
12:06:05amiconnhi LinusN
12:06:59linuxstbDoes anyone else get an illegal instruction error when unplugging usb on an iriver with the current cvs?
12:07:11_FireFly_i will try
12:08:06linuxstbi.e. install the latest bleeding edge build, then reboot to run it, then enter USB mode, then leave USB mode
12:08:49markunwebguest82: You are from korea, right?
12:10:13lameddevs, wil you object a patch tha adds LCD_PUTS_OFFSET & LCD_PUTS_SCROLL_OFFSET in all drivers for some specific usage?(and changes the old calls to one-line functions) I mean, nobody would mind if there are some more lcd functions right..? I know it sound really stupid question, but that is to be intended my first serious patch, and the other way i could do it (instead of making a new function and replacing older to macros) is to change lcd_puts & lc
12:10:47_FireFly_linuxstb: i become I04: IllInstr at 00000002
12:10:57lamed-inside the functions
12:11:02webguest82markun: Yes
12:11:04linuxstb_FireFly_: Yes, that's what preglow and I found last night.
12:11:42linuxstbThe most likely culprit is one of TiMiD's gui patches last night, but I haven't looked at the changes in detail.
12:12:01_FireFly_maby in the text-widget
12:12:07markunwebguest82: I looked in the logs and see you have a H3xx. If you had a H1xx I could build you a with unicode and korean.
12:13:28webguest82Because there is no money impatiently, can not buy H100.
12:13:36lamedplease try to answer my weird question.. it's important for my prograss
12:14:16webguest82Thank if make.
12:14:16LinusNlamed: what would lcd_puts_offset() do?
12:16:17_FireFly_linuxstb: or maybe it has something to do with the problem mentioned in this comment which i have found in some plugins
12:16:25_FireFly_ otherwise you will get lovely "I04: IllInstr" errors... :-) */
12:16:36_FireFly_ otherwise you will get lovely "I04: IllInstr" errors... :-) */
12:16:36_FireFly_ rb = api;
12:16:47lamedlinusn: puts the text offseted. actually, it will pass the integer to lcd_puts_style_offset like it does now to lcd_puts_style, and there, instead of using lcd_putsxy it will use lcd_putsxyofs with that integer.
12:17:00_FireFly_if you are using a global api pointer, don't forget to copy it!
12:17:00_FireFly_ otherwise you will get lovely "I04: IllInstr" errors... :-)
12:17:06webguest82markun: Can you give now?
12:17:31linuxstb_FireFly_: No, I don't think this has anything to do with plugins. Same error, but I think a different reason.
12:17:31markunwebguest82: I can, but what use is it if you don't have a H1xx?
12:18:07linuxstbTiMiD said that it only happens when dircache is enabled.
12:18:16webguest82Can do.
12:19:36webguest82How do you give?
12:20:29LinusNlamed: should work, methinks
12:22:45lamedlinusN: it does work.. i'm wondering if that might be accepted to the CVS? it's indeed needed (will be done porbebly this~next week
12:22:47webguest82markun: Will you give?
12:23:04markunwebguest82: ok, give me a few minutes
12:23:30LinusNlamed: i can't say that it would be accepted, since you don't tell me what the rest of the patch does
12:24:00lamed:D thanks a lot for the answers
12:24:52lamedGday all
12:25:02 Quit lamed ("CGI:IRC")
12:25:59 Join Febs [0] (
12:26:55 Join Maxime [0] (
12:31:23 Part LinusN
12:34:47webguest82markun: %u3147o you use MSN messenger?
12:34:59webguest82 D
12:37:21_FireFly_linuxstb: i had just run a logf version of the build
12:38:14_FireFly_and then i become a I00: failure and in the remote the last displayed call is "Building directory cache"
12:39:08linuxstbThat seems to confirm what TiMiD said last night - that it's only a problem when dircache is enabled.
12:39:26_FireFly_but what changes tricker this
12:39:31linuxstbIt's probably just trying to access a null function pointer.
12:40:06webguest82markun: Added.
12:41:24 Quit Kohlrabi (Read error: 104 (Connection reset by peer))
12:44:00 Quit Mxm`Pas`Bien (Read error: 110 (Connection timed out))
12:51:53DMJCanyone know what model of LCD screen the iriver has?
12:52:04linuxstbI don't understand why, but removing the call to "gui_textarea_clear(&screens[i]);" at the very end of apps/tree.c fixes the illegal instruction bug.
12:52:08DMJCmodel name etc, mate of mine is an electrical engineer
12:52:23DMJChe's looking at ordering the part/fixing the busted lcd screen for me
12:53:11amiconnIt seems that TiMiD's latest changes triggers this IllInstr exception after USB, but only on iriver and only if the dircache is enabled
12:54:03amiconnIt looks to me like some uninitialised function pointer is called (derived from the address of the IllInstr)
12:54:07linuxstbamiconn: See my previous message.
12:54:27linuxstb(2 minutes ago)
12:54:27amiconnThe question is why this pointer is only uninitialised when dircache is on
12:55:05linuxstbIt's obvious now.
12:55:18amiconnMaybe it's not uninitialised in the first place, but gets overwritten somehow
12:55:20linuxstbIt accesses screens[i] OUTSIDE the for loop.
12:55:35linuxstbi.e. screens[2] if NB_SCREENS is 2.
12:56:01linuxstbThis is the tree_restore() function at the very end of tree.c
12:56:30linuxstbBut I don't understand what that code is for in the first place.
12:57:16amiconnI think it's for removing the "rebuilding dircache.." info when the rebuild is done
12:57:48amiconnIf so, then it should be in another for() loop
12:57:50linuxstbSo we need a second for loop to iterate through the screens after the dircache is built?
12:58:20_FireFly_or if it works remove it
12:58:34_FireFly_if it works without explicitly clean the area
12:58:36linuxstbNo, I think the correct solution is a second for loop.
12:58:52linuxstbTimid: You around?
13:01:39linuxstbAdding the second for loop works fine. Should I commit it?
13:02:53_FireFly_why not i think this was a mistake from him
13:04:33TiMiDI'h here
13:04:36TiMiDreading the logs
13:05:40linuxstbTiMiD: I've just committed a fix to the I04 error.
13:05:54linuxstbWe're pretty sure it's correct.
13:06:18TiMiDthe problem was that it accessed screens[2] ?
13:06:36linuxstbYes - accessing screens[i] OUTSIDE the for loop. So i would be equal to 2.
13:07:09_FireFly_a c'n'p error
13:07:24_FireFly_i think
13:07:29TiMiDmaybe I c'n'p at the wrong place :p
13:07:44_FireFly_the place isn't worng at all :)
13:07:45TiMiDouch :)
13:07:50TiMiDI see the error now :p
13:07:54TiMiDI forgot the loop :)
13:08:05_FireFly_yepp and fixed in cvs
13:08:32TiMiDthat's what happens when you are tired but still programming
13:08:57TiMiD_FireFly_: about your wps widget, there is gui_textarea that could be useful
13:09:23TiMiD(I changed the API a little, but it's more logical and easy that way)
13:09:35TiMiDIm' afk for 30mn
13:10:15 Join amiconn_ [0] (
13:10:27 Join Moos [0] (
13:11:41MoosHello all
13:12:07markunHi Moos!
13:14:25preglowbloody gapless mp3
13:15:09 Quit webguest82 ("CGI:IRC (EOF)")
13:20:47DMJCanyone know where you can get a replacement lcd screen for iriver iho-140?
13:21:28B4gderit'll come attached to an iriver ;-)
13:21:53***Saving seen data "./dancer.seen"
13:22:01DMJCshould be able to do it much cheaper than that
13:23:46B4gderthen do that
13:24:32tvelocitymy ISP sucks cock
13:24:36 Quit amiconn (Read error: 110 (Connection timed out))
13:24:37 Nick amiconn_ is now known as amiconn (
13:24:39tvelocityi'd like you all to know that
13:25:33preglowfor free?
13:25:52DMJCI need to know what the lcd display's part number is
13:26:01DMJCit's not made by iriver is it?
13:26:05preglowDMJC: i doubt it'll be easy to get that way
13:26:13B4gderDMJC: seen the URL I showed you?
13:26:17DMJCif I can get the part I can get it fixed
13:26:23DMJCsaw the url on the website
13:26:33B4gderit mentions the specific epson model
13:26:43B4gderEpson S1D15E06
13:27:02B4gderwhich it most likely is
13:32:45DMJCsomeone's claimed that screen is also used on ipod
13:34:54B4gderthe 4g ipod as an lcd with the same resolution
13:36:08 Join Kohlrabi [0] (
13:37:56DMJChmm so is the lcd different from the controller?
13:38:04DMJCthe actual screen is cracked on mine
13:38:09DMJCthat's what I'm looking to replace
13:41:08 Join ghode|afk [0] (
13:42:22 Quit ashridah ("sleep")
13:44:33CtcpIgnored 5 channel CTCP requests in 5 minutes and 10 seconds at the last flood
13:44:33*B4gder feels good today
13:44:46B4gderBOINC converted to using libcurl
13:56:00preglowgod, how i hate adsl
13:58:24Zagorbeats modems :-)
14:00:00preglowdepends on who's supplying
14:00:11preglowthis connection is too moody for me
14:00:29B4gderthat's not quite adsl's fault ;-)
14:00:32 Quit ghode|afk (Read error: 110 (Connection timed out))
14:00:59preglowoh, no, but the lousy upstream is, and that's another of my gripes
14:01:02preglowi want sdsl :/
14:01:15B4gdervdsl is
14:01:27preglowbut you need to pay through the nose for that here
14:01:41Zagorthat's not adsl's fault either :-)
14:01:44B4gderand live above a "station"
14:02:26Zagorotoh you pesky norwegians will soon have bought most of our net infrastructure if you continue at the current pace :-)
14:02:33 Quit Strath (Read error: 104 (Connection reset by peer))
14:02:40B4gderbtw, I learned that wimax won't be that fun as it sometimes is said to be
14:02:58preglowZagor: hahaha
14:09:02preglowsent my h120 hurtling towards the ground
14:09:13preglowlet's hope this marks the end of my rockbox involvment
14:09:21B4gderwas it on?
14:09:32preglowno, and it starts just fine
14:10:33preglowi think the disk is designed to take quite a punch, as long as it isn't on
14:10:52B4gderyes, several Gs
14:11:14B4gderI remmber reading the number before, but I've forgot
14:12:08Zagoriirc, the disk is encased in a nice rubber holder on the iriver that should lower the Gs quite a bit
14:12:14 Quit Sandking ()
14:13:41preglowyes, assuming i replaced it properly :P
14:14:19preglowB4gder: more than several gs, i think it was something around 40
14:15:40 Join amiconn_ [0] (
14:16:37*amiconn_ wonders what's going on with his home connection
14:16:44markunB4gder: I see you committed the iso8859-*.bdf fonts. Do you remember where you got them from?
14:17:07B4gderno, I believe they came in a patch or similar
14:18:52markunI am replacing fonts with their unicode parents. The iso8859 fonts are made from the GNU Unifont I think.
14:21:28markunI am replacing fonts with their unicode parents. The iso8859 fonts are made from the GNU Unifont I think.
14:43:50 Join linuxstb_ [0] (
14:53:36 Quit Febs ("CGI:IRC (Ping timeout)")
14:54:09 Quit B4gder ("time to say moo")
14:54:47_FireFly_TiMiD: to use the text_widget maybe also for wps was a suggestion which i had made to you yesterday :)
14:59:36 Quit linuxstb (Read error: 110 (Connection timed out))
15:08:53TiMiDit doesn't really handles text anyway, just the area in which text can be typed
15:09:20_FireFly_yepp i saw it
15:10:40_FireFly_the first problem which i have to solve is, to make a good data-structur for each wps
15:11:25_FireFly_which all necessary information so that on the and only a draw_wps(<wps-struct> ) is needed to draw it :)
15:11:49 Quit TiMiD (Remote closed the connection)
15:21:11linuxstb__FireFly_: Isn't that data structure just a copy of the WPS text file?
15:21:56***Saving seen data "./dancer.seen"
15:22:25 Join LinusN [0] (
15:23:02 Join TiMiD[FD] [0] (n=TiMiD[
15:23:33 Quit TiMiD[FD] (Client Quit)
15:23:41_FireFly_what about the images, the lines and sulines
15:24:03_FireFly_which are currently separatly saved
15:24:26_FireFly_in ram
15:25:13preglowwhen i've got "move to next folder enabled", audio buffering doesn't continue once the last track in the folder has been loaded
15:25:16linuxstb__FireFly_: True. I suppose you need to gather them all together into a struct.
15:25:24preglowis this a feature or a bug?
15:25:38_FireFly_if i only save the format-buffer then i have to parse this buffer complete each time the screen is updated
15:25:47linuxstb_preglow: I would say that's a bug.
15:26:04linuxstb__FireFly_: I thought that's how it worked now (but I could be wrong)
15:26:08 Join TiMiD [0] (
15:26:36linuxstb__FireFly_: In fact just ignore me. I shouldn't talk about things I don't know anything about.
15:27:04LinusN_FireFly_: linuxstb_ is basically right
15:27:20linuxstb_I am? :)
15:27:43LinusNthe wps file is parsed and split into sublines, and the images are loaded
15:27:54_FireFly_i know
15:28:06LinusNbut the %-codes are reformatted each time the wps updates
15:30:11*markun gets a bit tired of talking to a translator program..
15:30:11_FireFly_the only problem i currently see is how to manage the 2dim arrays for the lines because MAXLINES depends on the screen-size
15:30:38 Join hshah [0] (
15:30:55LinusN_FireFly_: ah yes, you don't want the remote struct to be unnecessarily large?
15:31:10_FireFly_if i can avoid it
15:31:13TiMiDthat would bea easy if we had malloc ...
15:31:23LinusNdon't swear
15:31:36LinusNand we do have semi-dynamic memory
15:32:04linuxstb_Can't you just index the arrays manually - i.e. a[y*width+x]
15:32:47TiMiDhmm so the wps aray would be initialized at the very beginning
15:35:23_FireFly_a simple solution would be to have for each wps-struct the same size which will be the size from the largest display on the system
15:36:09_FireFly_and have a additional var which has the actual max-lines which can be displayed on the display which is represented by the wps-struct
15:36:11TiMiDbuffer_alloc is not a very complicated solution
15:36:15LinusNyes, or allocate the buffers with buffer_alloc()
15:36:48TiMiD_FireFly_: the var that represents tne number of lines that can be displayed is in the screen struct
15:37:15TiMiDyou also have char_width and char_height
15:38:12_FireFly_equals the value in this var this define: #define MAX_LINES (LCD_HEIGHT/5+1) ??
15:38:27LinusNTiMiD: when are char_height/char_width updated?
15:38:39TiMiDLinusN: that's the good question :)
15:39:10TiMiDthey are actually updated when a function needs them
15:39:32TiMiDwhen they are susceptible to change
15:40:05TiMiDbut the most convenient thing would be IMHO to do that in the code that changes the font for example
15:41:07 Join RiverFish [0] (
15:41:29 Join ghode|afk [0] (
15:41:48RiverFishSomeone mentioned here that some people think that RockBox sounds worse that iRiver firmware
15:42:13RiverFishIs there any likely reason for RockBox sounding different other than the different treble and boost behaviour
15:42:27LinusNwell, we have different codecs
15:42:55RiverFishdo the fixed point codecs produce the same output PCM as the reference floating point codecs?
15:43:02preglowRiverFish: no
15:43:23preglowRiverFish: every codec has sligthly different output
15:43:42RiverFishok that's news to me
15:43:54RiverFishI'd that all codecs had to produce the same PCM to be compliant
15:44:01preglownot so
15:44:04preglowit has to be the same within a limit
15:44:15preglowit depends on the codec
15:44:21preglowit is like this for mp3, at least
15:44:28preglowand other formally defined codecs
15:44:40RiverFish i mostly use vorbis and thinking of using musepack
15:44:41LinusNthere are also different levels of compliance for mp3
15:45:06preglowRiverFish: the iriver firmware might also be using 20 bit output
15:45:29preglowRiverFish: but anyway, i've personally decided to ignore these claims until i see hard proof (double blind tests)
15:45:53preglowshouldn't take long to conduct a proper test
15:45:56KohlrabiI consider gapless playback much more important than slight subconscious differences :)
15:46:04RiverFishwould comparing the SPDIF output from both firmwares work?
15:46:17preglowRiverFish: that would be the best approach currently, yes
15:46:35preglowLinusN: btw, did you notice what bit depth the spdif output from the iriver firmware had?
15:47:19RiverFishI'll do some tests if I can find some time
15:50:21linuxstb_One possible difference is that Rockbox doesn't do any processing on the audio if you don't ask it to. We don't know if that's the case for the iriver firmware.
15:50:57linuxstb_I read a complaint from someone saying that WAV sounded better than FLAC - the reason turned out to be winamp processing the audio differently in the two cases.
15:52:05preglowcan't say i miss winamp
15:52:07RiverFishRockbox not doing automatic processing without asking is a big IMO
15:52:16RiverFishbig plus that is
15:52:19preglowguess who's using the web client ;)
15:52:54linuxstb_+++ I'm not.
15:53:32preglowbut anywho
15:53:39preglowlame headers are only found on mp3 files, yes?
15:53:45preglowso i can hard code the mp3 frame size
15:54:16linuxstb_preglow: Are you talking about mp1/mp2 files? Or other formats completely?
15:54:45preglowi'm talking about lame headers only being found on mp3 files and not mp1/mp2, yes
15:54:50preglowbasically anything handled by mpa.c
15:55:13preglowi'm trying to fix the large enc_delay problem
15:55:22linuxstb_I can't think of a reason for people to encode their own mp2 files. They will normally be recordings from digital tv/radio broadcasts. In which case, there won't be any headers at all.
15:55:45preglowi prefer listening to mp2 files any day
15:55:59preglowso i might have wanted to do that instead of mp3, but i use vorbis ;)
15:56:45preglowforget it, i'll just assume i always see a layer 3 frame size and put in a warning for the people who know better than me
15:56:50linuxstb_I would say assume it will only be layer 3 - you can always fix it in the future.
15:57:38preglowanother problem is i don't have test files
15:58:16preglowthat use such insanely high enc_delays, that is
15:58:38RiverFishpreglow: suspect that iRiver firmware SPDIF output is 16 bit
15:58:50preglowRiverFish: do you have any idea of finding out?
15:58:57RiverFishsince it works with my cheap SoundBlaster /
15:59:34preglowi think i'll remove all these commented splash() functions from mpa.c
16:00:26 Quit ghode|afk (Read error: 110 (Connection timed out))
16:01:47RiverFishoops there I go again web clienting :)
16:01:59RiverFishmp3plus that should read
16:03:55_FireFly_should i made the size of the format-buffer always be 1600bytes or should this size be selectable by the programmer which use this widget
16:04:00RiverFishpreglow: is mp2 noticeably better. I've never tried it head to head with mp3
16:04:21preglowRiverFish: i prefer the encoding artifacts of mp2 to those in mp3
16:04:56preglowmp2 encoding artifacts sound more or less like those in musepack
16:05:53RiverFishon a brief listening musepack artifacts seem more natural and easier to live with
16:06:20preglowyes, exactly
16:06:55preglowsubband codecs operate on much coarser frequency bands, and very seldom eliminate entire frequency bands
16:07:02preglowso you'll pretty much only get more noise here and there
16:07:08preglownot so with transform codecs
16:07:26TiMiD_FireFly_: I would rather use a fiexed value
16:07:45TiMiD1600bytes should be enought for everyone :
16:07:53RiverFishi find mp3 really tiring to listen to. vorbis is better but I'm interested in trying musepack seriously
16:07:54 Part LinusN
16:08:18_FireFly_k and the same for the Maximages(52) ??
16:10:13TiMiDif you do it variable, it would need an additionnal entry in the config menu that is big enougth
16:10:22TiMiDbut it's as you wish :)
16:11:43markunRiverFish: about differences in mp3 decoder output:
16:12:24RiverFishmarkun: thx
16:15:21 Quit hshah ("CGI:IRC")
16:30:49RiverFishCan't see any discussion of allowed vorbis decoder differences in the spec
16:31:05RiverFishBut there's some talk about the low and high accuracy modes of tremor
16:33:42 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
16:34:37amiconn_preglow: Why do you need to assume a frame size?
16:37:25amiconn_You can calculate the frame size from the header, rockbox already has a function to do that
16:46:42preglowthink libmad has as well
16:46:53preglowam i the only one having some problems with the lower watermark on the pcm buffer?
16:46:57preglowit never boosts now
16:46:59preglowso i get skips all the time
16:47:38preglowRiverFish: vorbis doesn't have any formal accuracy spec yet
16:48:04preglowRiverFish: the low and high accuracy modes of tremor is something different, we use high accuracy mode
16:48:47RiverFishpreglow: yep just found // #define _LOW_ACCURACY_ commented out
16:54:09linuxstb_preglow: Which codec (for your watermark problems) ?
16:56:01 Quit RiverFish ("CGI:IRC")
16:56:06preglowlinuxstb_: i tried both mp3 and vorbis
16:56:23preglowthink i'll just try a daily to see if it's just me screwing up
16:58:36linuxstb_Maybe that's the reason for my 24-bit WAV boosting.
16:59:25linuxstb_Anyway, it works file with ALAC. Just trying some others.
16:59:58linuxstb_MP2 is fine.
17:00:59preglowgiving daily a try now
17:01:03linuxstb_MP3 seems fine as well.
17:01:38preglownope, daily works fine
17:01:42preglowi wonder how i managed to break this
17:10:01preglowhmm, could you try a debug build with logf enabled for me? that seems to trigger it here
17:15:53linuxstb_Yep, I'll try it now.
17:22:00***Saving seen data "./dancer.seen"
17:23:01linuxstb_Yes, I'm getting the same problem. Very odd.
17:23:22preglowi've used debug builds often before, so it's a new problem
17:24:18 Join Febs [0] (
17:26:30 Quit DMJC (Read error: 110 (Connection timed out))
17:35:11preglowamiconn_: besides, i don't think you'll ever see a lame info header on anything but mp3
17:42:30 Join ghode|afk [0] (
17:55:53 Join muesli_- [0] (
18:04:02 Quit ghode|afk (Read error: 110 (Connection timed out))
18:04:11 Join Jue_ [0] (i=Jue@
18:07:42Jue_(in advance: sorry for my bad english) i have got some questions about the iriver H1xx-Rockbox.
18:10:07Jue_thanks. As there are only to players that support FLAC (The H1xx with Rockbox and the X5) wich one should i choose. I heard many complaints about the x5-firmware and very good things about the soundquality of the H1xx-series. Am i right so far?
18:10:36 Quit Paul_The_Nerd ("Chatzilla 0.9.68a [Firefox 1.0.7/20050915]")
18:11:25preglowas far as i know
18:11:25Jue_I also heared, that the X5 does not support gapless replay. Will the a Rockbox-H1xx support gapless replay and replaygain with my Q8-Flac-files?
18:11:35preglowrockbox already does support gapless playbacj
18:11:44preglowand rockbox has a very efficient flac implementation
18:12:04preglowit SHOULD support gapless flac
18:12:10preglowas well as flac replaygain
18:12:34preglowgive me a second, and i can verify the gapless part, at least
18:12:49Jue_thank you very much.
18:14:21preglowjust need to do some ripping and encoding, give me five minutes
18:14:43Jue_no problem. ^^
18:14:55linuxstb_Yes, FLAC is definitely gapless.
18:15:03preglowthere you have it, heh
18:15:09Jue_and the replaygain-support?
18:15:21preglowi think we support replaygain for almost all our formats
18:15:24linuxstb_I'm not sure. But if it's not there we can easily add it.
18:16:19 Quit Febs ("CGI:IRC (EOF)")
18:16:22preglowand i'm pretty sure someone added replaygain support for flac
18:17:01Jue_how much battery duration could i expect if i play Flac-q8-files without using the display often?
18:17:30preglowlinuxstb_: god damn, wavpack is a fast encoder
18:17:55linuxstb_Jue_: I don't know. I plan to do such a test soon. Maybe overnight tonight to see if it's still running in the morning.
18:17:56preglowlinuxstb_: heaps faster than flac, and actually makes smaller files
18:18:24linuxstb_preglow: Yes, I know the encoder is fast. But I didn't think the files were significantly smaller.
18:18:42linuxstb_e.g. something like 30MB for Wavpack, 31MB for FLAC
18:18:54preglowlinuxstb_: no, not very much, but they are consistently smaller
18:19:26 Join Lear [0] (
18:22:02Jue_hrm... another question: does Rockbox feature the display of individual vorbis comments? For example: I added Vorbis comments for "Mood" and "Rating".
18:22:56linuxstb_I think someone has written a patch for that, but it's not been comitted to Rockbox CVS yet.
18:23:37preglowmost definitely gapless
18:24:27linuxstb_But I think I broke replaygain when I wrote the new FLAC decoder.
18:24:56linuxstb_I think I just need to add the codec_set_replaygain(ci->id3) call back into flac.c
18:25:07Jue_yeah. then it would be possible to sort them by rating and so on? When the H1xx's Soundquality is as good as i heared... this would be perfect.
18:27:06linuxstb_Replaygain should now be working for FLAC again. Just committed it to CVS.
18:28:08linuxstb_Jue_: No - Rockbox doesn't do anything with the tags - it just displays them.
18:28:41preglowi hate the way the seek bar skips back to where it was, before finally advancing to where it was when you seek
18:28:56linuxstb_There is a "tag database" in development - that may be able to do what you want. But not many developers use it, and it doesn't seem to be developed very much.
18:29:14Jue_would be great. A Flac-player with gapless replay, replaygain and individual vorbis comments (when will this patch be commited to the CVS?). great!
18:29:48linuxstb_The patch will be committed when a developer with CVS access takes an interest in it.
18:30:25Jue_oh. even without the database. this is really awesome. thanks, linuxstb!
18:30:53linuxstb_preglow: I agree with you about the seek bar.
18:30:55Learlinuxstb_: btw, there was a reason for using rb as api pointer in wav.c, as you noticed... :)
18:31:15linuxstb_Lear: Yes, but I don't think any of the codecs should be using "rb".
18:31:15preglowlinuxstb_: i wonder if it is that way on archos as well
18:31:52Learthen I think codec.h should redefine DEBUGF/LOGF...
18:32:04preglowi think it should take the api ptr as an argument
18:32:08preglowLOGF(ci, "w00t");
18:33:05linuxstb_We could do it in codeclib.h - defining "ci" as an external variable.
18:33:35linuxstb_I also want LOGF/DEBUGF inside the codec libraries, not just the main codec .c file.
18:34:19linuxstb_Lear: Did you try playing 24-bit WAV files. I tried one yesterday, and it causes the CPU to boost more than Wavpack does when playing 24-bit files.
18:34:28 Part Jue_ ("Verlassend")
18:35:16Learlinuxstb_: I think I did... Didn't check on CPU usage though.
18:49:39 Join _aLF [0] (
18:54:35preglownone if you fellas got any mp3 files with a large enc_delay?
18:56:57preglowthe wps is completely helpless for a series of small files
19:05:13 Quit muesli_- (Read error: 110 (Connection timed out))
19:12:16Learbut is that caused by enc_delay? doesn't similar things happen with oggs? (maybe you mean something different by helpless wps...)
19:12:25 Quit amiconn_ ("CGI:IRC")
19:12:39preglownonono, nothing to do with delay
19:12:58preglowit's got to do with the wps fucking up towards the end of all files
19:13:28preglowhmm, i need a mono mp3
19:19:42preglowi'm still not entirely satisifed with the gapless
19:19:51preglowsounds like foobar still does a slightly better job
19:22:04***Saving seen data "./dancer.seen"
19:23:18Learbut they do things pretty much in the same way...
19:23:29Learbiggest difference is the decoder.
19:23:38preglowLear: you did the if (stop_skip > 0) part in mpa.c, yes?
19:23:42preglowi don't quite get why it's needed
19:24:00Learyes, just looked at how foobar did it. :)
19:24:15preglowLear: the lame header handling in mpa.c is similar to foobars because that's where i found out how it worked ;)
19:25:03preglowsamplecount already does account for stop_skip
19:25:06LearI kind of guessed that...
19:25:07preglowso i don't see why it needs to be done again
19:27:09 Join tucoz [0] (
19:27:13Learbut samplecount wasn't really used before, was it?
19:27:23tucozHi, what is the proper way of deleting wikispam?
19:27:36tucozHanSolo has entered a few links in a clever way
19:27:58preglowLear: perhaps :)
19:28:34preglowLear: no, you most certainly seem to be right
19:29:33tucozI mean, how do I revert a change to a previous version of the wiki page in question?
19:29:34Learthat's what made me look at it to start with...
19:32:09preglowit just needs to be adapted somewhat, though, framelength == 0 is no longer a breaking condition
19:32:59preglowi wonder if samplesdone is calculated properly when seeking
19:33:19Learit is, as long as there's a stop_skip?
19:34:14preglowLear: there are files with a start_trim big enough to encompass several frames
19:34:26preglowLear: so the first couple of frames might end up with a framelength == 0
19:35:16preglowlead_trim, whatever
19:35:20tucozZagor, Bagder: You need to delete HanSolo from the wiki, spammer.
19:35:22Learhow could that be? the file starts with short frames or what?
19:35:41preglowLear: the first couple of frames might contain pure bitreservoir buildup
19:37:34preglowit's kind of a stretch, but i figured i might as well support it
19:37:39preglowi'll just remove the break condition
19:39:08preglowit would rather be better to break if samplesdone > samplecount
19:39:13preglowwhich is the case if max is negative
19:39:38preglowbut then we'll break files with an incorrect frame number in the header
19:42:00preglowsobut those are broken now anyway by the same code, it seems
19:55:01 Quit DangerousDan (Read error: 110 (Connection timed out))
19:59:02 Join DangerousDan [0] (
20:05:13 Join paugh [0] (n=kickback@2001:5c0:8fff:ffff:8000:0:3e03:6822)
20:05:15 Join pengo [0] (
20:07:19 Join ghode|afk [0] (
20:07:59 Join dpassen1 [0] (n=dpassen1@resnet-233-61.resnet.UMBC.EDU)
20:08:50 Quit tucoz ("CGI:IRC 0.5.4 (2004/01/29)")
20:20:26 Quit linuxstb_ (Read error: 104 (Connection reset by peer))
20:21:00 Join linuxstb_ [0] (
20:27:48 Quit ghode|afk (Read error: 110 (Connection timed out))
20:28:51 Quit dpassen1 (Read error: 110 (Connection timed out))
20:31:02 Join linuxstb__ [0] (
20:31:29preglowlinuxstb__: yo, should i commit my sample conversion bypass thing even though it isn't exactly elegant?
20:31:48preglowi might not have time to work on rockbox for some days
20:32:42linuxstb__Yes, commit it. I may have time to tidy it up a little.
20:33:32 Join adiamas [0] (n=adiamas@
20:33:39preglowit's just a little thing anyway
20:34:57*adiamas perks up
20:37:21preglowwhen i start passing 32 bit ints to pcmbuf_insert, shouldn't i do frameInfo.samples*4 and not frameInfo.samples*2 like in your 16 bit version?
20:38:16preglowi sorely wish slasheri would sanitize the argument format to that function
20:38:27preglowit just doesn't make sense as it is
20:39:26preglowhaha, it's realtime if i increase it to *4 :PP
20:39:30preglowbut with a nice time stretch
20:43:03pengoi wish you could change the speed of audio (without changing the pitch)
20:43:14pengopodcasters talk too slowly
20:43:57pengoand most music is better when 12% faster
20:44:11preglowpengo: i'm going to fix that some day
20:44:22pengopreglow: i love you.
20:44:24preglowpengo: i could have done it right now since it's not very hard, but some quirks in rockbox makes it hard
20:45:02preglowlinuxstb__: just be aware that this method might break for channels > 2 streams
20:45:11preglowsince libfaad does some internal channel mapping
20:45:38 Quit linuxstb_ (Read error: 110 (Connection timed out))
20:46:54preglowlinuxstb__: you wouldn't happen to have some isp troubles, would you? ;)
20:47:58linuxstb__preglow: Yep. ADSL....
20:48:08 Nick linuxstb__ is now known as linuxstb (
20:49:20linuxstbpreglow: As long as we add something like MAX_CHANNELS define (set to 2), I don't think it's a problem.
20:50:52pengoi wouldn't mind having 200ms (or 100ms) as an option for backlight fade in.. 500 is too long.
20:52:25 Quit tvelocity ("Leaving")
20:53:06preglowpengo: you know, i agree
20:57:56 Join linuxstb_ [0] (
21:07:41 Join linuxstb__ [0] (
21:13:38 Quit linuxstb (Read error: 110 (Connection timed out))
21:22:08***Saving seen data "./dancer.seen"
21:23:17 Join dpassen1 [0] (n=dpassen1@resnet-233-61.resnet.UMBC.EDU)
21:24:11 Quit linuxstb_ (Read error: 110 (Connection timed out))
21:29:13 Quit _FireFly_ ("Leaving")
21:30:00 Join RotAtoR [0] (
21:34:01 Join yngwi [0] (
21:35:14 Part yngwi
21:37:24 Quit adiamas ("Chatzilla 0.9.68a [Firefox 1.0.7/20050915]")
21:38:09 Join yngwi [0] (
21:38:56 Quit yngwi (Client Quit)
21:39:30preglowyes, very likely
21:41:15 Join godzirra [0] (
21:41:30godzirraHeya guys. Quick question... is there anyway to set rockbox to autoresume where it was before it turned off last?
21:42:36godzirraCan you tell me how? :)
21:42:36crwlthere's auto resume (or similar) in settings
21:42:39preglowyou can always resume by pressing 'play' when you boot
21:42:47preglowafter you boot, that is
21:43:04godzirraHrm.. how do I get to settings on an iriver then? I've looked everywhere I can and I dont see autoresume. And I have a fairly new release of rockbox
21:44:52crwlit's behind the a/b button
21:45:30godzirraHmm.. I get Shuffle Mode - Repeat Mode - and Show Files options when I try that.
21:45:47dpassen1don't hold it down long
21:46:09preglowit's in general settings->playback->resume on startup
21:46:26godzirraThanks very much :)
21:46:31godzirraI love rockbox.
21:46:36preglowyou should ;)
21:47:56godzirraI do. i've been using it for about 4 months now.
21:48:00godzirraIts fantastic.
21:54:15 Join LinusN [0] (
22:00:47amiconnpreglow: The seek bar snap-back is an iriver-only glitch. On archos it's glitchless
22:08:41 Join ghode|afk [0] (
22:14:10preglowamiconn: i suspected as much
22:14:19amiconnLinusN: Do you know whether the SDRAM controller should notice on-the-fly changes of the page mode bit in DACR0? I tried that in a test plugin measuring copy time with and without movem for various alignments, and didn't observe any speed change?
22:15:20LinusNi'm not sure, but i think it should work
22:29:42 Quit Kohlrabi (Nick collision from services.)
22:29:46 Join Kohlriba [0] (
22:30:06 Join len0x [0] (
22:50:42 Part len0x
22:52:13 Join len0x [0] (
22:54:51 Join Midgey34 [0] (
23:00:37 Quit Lear ("Chatzilla [Firefox 1.5/undefined]")
23:01:09 Join DMJC-L [0] (
23:04:21 Quit dpassen1 ()
23:20:48preglowok, we got a user on the forum saying seeking messes up the samplesdone, so it's as i thought
23:21:04LinusNsaw that too
23:22:10***Saving seen data "./dancer.seen"
23:23:27preglowi'll see if i can get proper gapless mp3 going some day soon
23:25:51preglowi should probably just mail the lame guy who has a page about this and ask him exactly how to do it
23:26:17preglowthe foobar2000 mp3 plugin does some strange things including codec latency which i've tried to replicate, but don't understand completely
23:26:37 Quit RotAtoR (Read error: 113 (No route to host))
23:27:29preglowshould also do some proper lame header parsing, what i put in there some time ago only qualifies as a hack
23:33:23*Bagder pokes on glibc with a stick
23:33:40 Part pengo
23:34:09 Join tucoz [0] (
23:34:27amiconnLinusN: I had a look at the SDRAM init, and there are some points that don't look right to me...
23:35:01amiconnIt might not help to improve speed, but could you have a look as well?
23:35:23tucozHello, is there a reason to why wma is still listed as supperted in rockbox. Played one by mistake with somewhat of a failure.
23:35:40preglowtucoz: listed as supported where?
23:35:42amiconnFirst, firmware/crt0.S, line 194 and 196
23:35:45tucozIn rockbox
23:35:49LinusNthe only reason is to create playlists
23:35:53preglowtucoz: what linus said
23:36:12Bagderit isn't "supported" in any other way that it creates playlists with them
23:36:13preglowtucoz: you can rest assured that wma wont be supported in rockbox for a long, long time, if ever
23:36:29amiconnDCR is set to 0x8001 for h120, but to 0x8204 for h100. The low 9 bits are correct, that's the refresh cycle count
23:37:15tucozAh, get it. Thought I remember something like that. Still, isnt it a nice thing to simply remove that playlist functionality now that rockbox, well, rocks on the iriver.
23:37:16amiconnBits 9 and 10 (RTIM) are what is fishy. For h120 it's 00, which means a tRC of 3 bus clocks
23:38:03amiconn3 bus clocks are sufficient for 11 MHz and 45 MHz operation. For 124 MHz operation, one bus cycle is 16.1 ns, x3 is 48.3 ns
23:38:14amiconnOur SDRAM wants tRC >= 65 ns...
23:38:19preglowtucoz: i tend to agree
23:38:32LinusNamiconn: ouch
23:38:32preglowpeople can always just enable 'show all files'
23:38:43amiconn...which means to set RTIM to 01 (6 bus clocks) for 124 MHz
23:38:55Bagdertucoz: people with WMA files could still build the playlists with rockbox and then reboot to play them...
23:39:00tucozas these horrible wma files could be in a directory and rockbox will try to play them with unwanted results.
23:39:18Bagderrockbox should not play them "with unwanted results"
23:39:21Bagderthat is the bug
23:39:28LinusNtucoz: rockbox should handle that much better
23:39:31tucozBagder: ok, fair enough.
23:39:47LinusNamiconn: sounds fair
23:39:55amiconnFor h100 it's the reverse problem. RTIM is set to 10, which means tRC = 9 clks - a waste
23:40:20LinusNeither we keep it at 6x, or make it dynamic
23:40:27LinusNi vote for 6x
23:40:40tucozBagder: btw, have you deleted HanSolo from the wiki? I deleted some massive spam from him earlier on, but wasn´t sure what to do with the user.
23:40:46amiconnI'd vote for making it dynamic.
23:41:01amiconnWe already fiddle with DRC in set_cpu_frequency() anyway
23:41:12LinusNtucoz: i have personally deleted him twice
23:41:21tucozLinusN: hehe
23:41:34LinusNamiconn: true
23:41:36tucoz...well looks like he is back.
23:42:18amiconnLinusN: Next thing is the sequence of lines 229..238 in crt0.S
23:42:40Bagderhansolo removed
23:43:22tucozgoodie, I googled for HanSolo and twiki and that gave a lot of hits.
23:43:27Bagdernext time I'll start blocking his IP
23:43:34preglowhow about doing it now!
23:43:34amiconnMCF5249UM.pdf p. 109 says that the init sequence requires to enable refresh first, then wait for at least 8 refresh cycles
23:44:05amiconnThe code does it the other way round. It might be irrelevant if I read the Samsung datasheet correctly
23:44:22len0xguys, why do you run CVS on insecure pserver instead of a proper ssh?
23:44:30LinusNno it doesn't do it the other way around
23:44:39preglowbecause we're not afraid of password sniffing? :)
23:44:46Bagderlen0x: because we're not afraid
23:45:10LinusNamiconn: it's only missing the 8-refresh wait
23:45:11Bagderit really is very little risk involved
23:45:12amiconnLinusN: The sequence in the code is precharge -> wait -> enable refresh -> MRS
23:45:51LinusNand the sequence in the data sheet is precharge -> wait ->enable refresh -> wait -> MRS
23:46:17amiconnYes, but the first wait is so short that we don't need an instruction to cover it, at least not at 11 MHz
23:46:20LinusN"3. Issue a PALL command to the SDRAMs by setting DCR[IP] and accessing a SDRAM location. Wait the time (determined by tRP) before any other execution."
23:46:33amiconnYeps. tRP is 20 ns ....
23:47:02tucozI think it is time for me to say Hooray for rockbox. Cheers to all the developers :)
23:47:32*Bagder joins in the cheering
23:48:00LinusNamiconn: in any case, it takes quite a long time before the sdram is accessed after enabling the refresh
23:48:18LinusNbut yes, we should fix that
23:48:18amiconnLinusN: Finally, lines 247..248 of crt0.S seem to be superfluous. Quote from the cf manual: "The DRAM controller clears IMRS after the MRS command finishes."
23:48:21tucozI found myself looking through all the menus yesterday and found out just _how_ good it is.
23:48:33*preglow cheers loudly
23:48:39amiconnNo need to clear ourselves...
23:49:38amiconnI'll try the dynamic RTIM. I won't try the bootloader changes myself without a BDM...
23:50:18LinusNamiconn: file a patch, so i can fix it in the next bootloader version
23:50:20tucozBtw, the ihpos-guy seem to make some progress with his hacking. .
23:51:58preglowtucoz: persistent fellow
23:52:17amiconnLinusN: For proper wait fixing, how do you interpret the 8 refreshes we have to wait? Is that 8 full 8192-cycles (64 ms each), or simply 8 single refreshes?
23:52:18Bagderreinventing the wheel big time
23:52:31tucozBagder: haha
23:52:32LinusNindeed, and where is the source?
23:53:36amiconnSamsung says "issue 2 or more auto-refresh commands". I think that means 2 or more single refreshes (8 to play safe)
23:53:37LinusNamiconn: i interpret it as 8 single refreshes
23:53:49preglowoh well, he's probably having a blast
23:53:58preglowsomeone mail him and tell him to join us :)
23:54:05Bagderbut he mentions having problems to find info
23:54:10Bagderinfo we have since long
23:54:15preglowwell, he knows about us
23:54:17preglowhe's using our bootloader
23:54:32Bagderhe's on a solo raid for sure
23:54:43tucozprobably, he said in a comment that he is thinking of that, and that he uses this as a learning experience.
23:54:44amiconnLinusN: So our loop can be somewhat shorter. 8 refreshes means 8*32 bus cycles for h120, and 8*80 bus cycles for h100
23:54:46 Quit DangerousDan (Read error: 104 (Connection reset by peer))
23:55:06Bagdernot everyone is a team player
23:55:20LinusNapparently not this guy
23:56:32tucozwell, seems like he could be.
23:57:29 Join Strath [0] (
23:57:55preglowwell, i get what he's doing very well, he's probably learning a bunch
23:58:00linuxstb__Someone let him loose on a H340
23:58:06 Join matsl [0] (
23:58:22Bagdernow there's an idea

Previous day | Next day