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

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

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

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

#rockbox log for 2010-02-21

00:02:17 Join Rob2223 [0] (
00:06:08 Quit Rob2222 (Ping timeout: 272 seconds)
00:06:09 Quit ecio1 (Read error: Connection reset by peer)
00:06:39 Join ecio [0] (
00:17:41 Quit MaadMan (Quit: Verlassend)
00:20:03 Join phanboy4 [0] (
00:23:39Unhelpfulsaratoga: that's what i was thinking, if the CPU is running at about 2x bus speed then presumably it can only load about every other cycle... perhaps a few fetches can queue depending on how things work but obviously it can't transfer data 2x as fast as the bus.
00:24:41Unhelpfulbut nicely-scheduled loads on arm9tdmi should only take 1c... i wonder if you can even avoid stalling for memory latency if the load is scheduled early enough?
00:26:09 Quit Llorean (Read error: Connection reset by peer)
00:27:53saratogaUnhelpful: i know on ARM7TDMI the best you can hope to do is hide the load latency by doing a slow multiply right before hand
00:27:57saratogai don't know about arm9
00:28:41saratogai can't remember how well seperated teh cache and load/store hardware are
00:28:56Unhelpfulsaratoga: i thought arm7tdmi was pretty much strictly in-order?
00:28:58saratogai guess it probably does keep fetching the rest of the cache line after a load though
00:29:24saratogaunhelpful: its in order yes, but a mul takes 5 clocks or so, so if you issue a mul, then a load, both can be done in parallel and still retire in order
00:29:37saratogai don't think it locks the load/store hardware while doing the mul
00:30:00Unhelpfulalso load on arm7 has 3c throughput :/
00:30:09saratogaisn't it 2c?
00:30:26Unhelpfuldon't think so, no.
00:30:31saratogaone clock to load the instruction, then one clock to load the register?
00:30:38saratogai thought it only stalled because of the 32 bit wide bus
00:30:45saratogaand a load only takes 64 bits
00:30:55saratogaso i would expect it to sustain a load every other cycle
00:30:58saratogabest case
00:31:36UnhelpfulASDG gives load as being 3c to execute
00:32:02saratogalatency or throughput?
00:32:10Unhelpfularm7tdmi doesn't have nearly as many nice 1c pipelined instructions, aside from basic ALU stuff.
00:32:11saratogai thought on the last cycle you could start issuing the next load
00:32:31Unhelpfulthe ASDG tables are all throughput, latency if any goes in a notes column.
00:32:38saratogaif not thats really stupid, since the LDM takes 1 +N cycles IIRC
00:33:10Unhelpful2+N - so LDM 1 reg == LDR
00:33:55saratogawhats the link for this I should review it
00:34:02saratogai obviously can't remember what i've read
00:35:01Unhelpfuldocstoc and google books both have ASDG, google books as rather a lot of pages not available though.
00:35:50 Quit ender` (Quit: We know the speed of light... so, what's the speed of darkness?)
00:35:51Unhelpfulthe fact that timing tables for everything from arm7tdmi to arm11 are all in a row at the end makes it a really wonderful reference for this
00:36:04saratogahmm my player turned off mid test_codec
00:36:27saratogais there a way to tell rockbox to not use the idle power off while in USB charge mode?
00:37:41 Quit petur (Quit: zzzzz)
00:38:38Unhelpfulsaratoga: it shouldn't during test_codec, but amiconn and i both had trouble with this happening on PP during high-profile APE testing.
00:39:29 Join karashata [0] (
00:41:23saratogayeah maybe I should have removed APE
00:41:43 Quit bmbl (Quit: Bye!)
00:41:46Unhelpfulhis theory was that 1) power management gets confused when codec doesn't yield often enough or 2) sustained running at max clock might actually be draining the battery, even on charger.
00:42:37saratogaits not doing that on e200v1
00:42:48saratogai can run it with no battery at all off of USB power
00:43:03saratogaand it only draws 120mA with the screen on and the CPU boosted
00:44:00Unhelpfulper ASDG arm7tdmi has a 3-stage pipeline and generally instruction execution times don't depend on preceding/following instructions. can't do other things during those up-to-5c multiplies. :/
00:44:22stripwaxsaratoga/mt - looks like cook uses clip_to_15 but I thought that wasn't needed (I thought dsp did all that)
00:45:18saratogastripwax: its not
00:45:33saratogawe have a patch somewhere to remove it but it made the codec prefetch abort
00:45:37saratogawe had no idea why
00:45:51saratogabut yes it should be removed
00:46:07stripwaxhah. ok. also the implementation of clip
00:46:11Unhelpfulis this a range limit on output or are some internal values being clipped and then used for further calculation?
00:46:24stripwax_to_15 in dsp is only two instructions; in codeclib it's 5.
00:47:51stripwaxUnhelpful - output.
00:49:04Unhelpfulstripwax: i count 3c - y = x >> 15, test y = x >> 31, do clip
00:49:15 Quit Omlet (Quit: ( :: NoNameScript 4.22 :: ))
00:50:18stripwaxUnhelpful - I have no idea what you just said :) teq %0, %0, asr #31 ; eorrne %0, %1, %0, asr #31 ; where %1 = 0xffff7fff
00:50:58stripwaxoh. um. hang on. maybe it is three :-p
00:51:12stripwaxanyway, point is, it's less than five :-)
00:51:49Unhelpfulstripwax: i'm looking in dsp_arm.S. you can do this clip operation in 3c if you can maintain a constant in a register for it, or 4c if not, but i could *not* get gcc to achieve either of these w/ C code.
00:52:29stripwaxah. yep. by '3c' you mean 3 cycles? sorry, just didn't understand what you were saying that's all. yep
00:54:04Unhelpfulstripwax: sorry, yes. 3 cycles. it looks like the inline in codeclib could be made at least 1 cycle shorter.
00:54:40stripwaxright. it's only used in one tight loop, i'm sure you can keep the mask in a register
00:55:03stripwaxi.e. output_math in cook_fixpoint.h
00:55:19stripwaxbut eliminating it altogether is a better approach in any case
00:57:31***Saving seen data "./dancer.seen"
00:57:55Unhelpfulthe one in codeclib seems to use some rather non-obvious methods... two tests? mov %1, %0, asr #15; teq %1, %0, asr #31; eorne %2, %0, asr #31; w / mask in %2, input/output in %0, tmp %1.
00:58:36stripwaxsaratoga - i couldn't see any patch in tracker, somewhere else?
00:59:26stripwaxUnhelpful - yeah we just want to eliminate it. afaict only cook uses it, and that shouldn't.
00:59:26Unhelpfulalso, why subs/adds when the output is discarded? cmp/cmn would make the codeclib 5 cycles but avoid use of other registers.
00:59:46 Quit BlakeJohnson86 (Quit: Leaving.)
00:59:55 Join Rob2222 [0] (
01:00:26 Join BlakeJohnson86 [0] (~bjohnson@2002:1876:a27b:0:227:13ff:fe65:1262)
01:01:31 Join Guest23293 [0] (
01:01:58 Quit n17ikh (Disconnected by services)
01:02:00 Quit Guest23293 (Client Quit)
01:02:16 Join n17ikh [0] (
01:02:28 Quit stripwax (Quit:
01:03:46 Quit Rob2223 (Ping timeout: 272 seconds)
01:08:02saratogastripwax: probably passed on IRC or gtalk
01:08:09saratogamt and i both had commit so we didn't use the tracker
01:08:29saratogathe patch should have been trivial though, just change the DSP bits and drop teh clip function
01:08:40 Quit koniiiik_ (Ping timeout: 276 seconds)
01:10:19saratogaprobably going to commit this
01:10:23saratogavery helpful for test codec
01:10:27saratogaor maybe just put it in the wiki
01:14:36saratogaheh wiki adds a .txt to my perl script
01:20:18 Join iq [0] (~chatzilla@unaffiliated/iq)
01:20:27saratogaTheSeven: would you mind posting the raw test_codec logs so I can update the wiki?
01:21:07saratogabah this is making me want to fix aac-he so that test_codec is quicker . . .
01:21:08TheSevenr24711 noeabi:
01:21:20TheSevenr24786 noeabi:
01:21:48TheSevenr24786 eabi:
01:27:53saratogaTheSeven: added
01:28:31saratogaopps bug in my parser
01:30:16 Quit avacore (Ping timeout: 252 seconds)
01:30:18iqMy clip+ booloader build is failing, any known issues ?
01:30:38 Quit hebz0rl (Quit: Ex-Chat)
01:32:33TheSevenfailing in terms of what? compilation errors? can you pastebin them?
01:32:42 Join avacore [0] (
01:37:56saratogadelete the build directory and try again
01:38:49saratogai am baffled by this perl problem with TheSeven's log
01:39:40saratogaprint "$key"; does nothing but print "$key\n" works fine
01:40:46TheSevenwhat if you just remove the quotes?
01:41:44TheSevenwhy should [u]modsi3 use floating point?
01:42:51saratogawhat revision is this?
01:43:29 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
01:43:36gevaertsiq: was that compiler built by
01:44:46iqgevaerts: I downloaded the VMWare image. So whatever was in there.
01:44:49saratogaheh perl's chomp can't handle windows line endings
01:45:16saratogaiq: try running to rebuild the compilers
01:45:30saratogatheres directions on the wiki
01:45:35iqsaratoga: ok
01:45:37gevaertsyes, it does look like a dodgy compiler install
01:45:48saratogaits possible the vmware tool chain can't build for the clip+ for whatever reason
01:46:39iqI was able to build the rockbox without errors.
01:46:50saratogabut not for the clip+
01:46:56iqAnd I was able to build Clip bootloader as well. It is just problem with Clip+ bootloader.
01:47:07iqsaratoga: yes for Clip+
01:47:56 Quit nimak (Ping timeout: 252 seconds)
01:48:42 Quit mt (Ping timeout: 240 seconds)
01:48:48saratogaiq: i don't even think its possible to build for the clip+
01:49:19iqsaratoga: okay, I will take your word for that, I just got involved with it today :)
01:50:01iqrunning - will report later.
01:50:21 Join nima [0] (
01:55:58saratogaWMA on the 2G is a full 20% faster with the new mdct lib
01:56:00saratogathats amazing
01:58:21saratogastrangely most other codecs got a much smaller improvement
01:58:25saratogai wonder why WMA is so special
01:58:49TheSevenwasn't wma extraordinarily slow before? (compared to other targets)
01:59:07TheSevenhow is the vorbis : mp3 : wma ratio for nano2g, and for other players?
02:00:09TheSeveniirc we did a test once where we removed the mdct calls from the wma codec in order to check what's causing that, and removing them made it behave much better in comparison
02:00:27TheSevenso it's rather obvious why this gains most from a faster mdct
02:01:30 Join Strife89 [0] (
02:04:02 Join AsaelReiter [0] (
02:07:11 Quit mirak_ (Quit: Ex-Chat)
02:07:30iqthat's funny. I ran out of disk space on my Virtual Machine :)
02:08:21UnhelpfulTheSeven: it needn't actually use the instructions, object files are flagged for certain features which may cause incompatibilities, whether those feature are exercised or not. try linking object files that use short enums with ones that don't, even if neither is built from code that uses enums.
02:08:47iqI will try checking out code in my Fedora and run Thanks for your help. bye for now
02:08:50saratogaTheSeven: theres 6 codecs all using the same MDCT
02:09:04saratogai don't see why WMA should have been different then vorbis for instance
02:09:14TheSevensaratoga: looks like wma uses more of them / more often :-)
02:09:43Unhelpfulsaratoga: if a larger portion of WMA's time is spent on mdct then it will show a larger % increase from mdct improvement...
02:10:23saratogathey both use the MDCT more or less the same
02:11:31 Quit iq (Quit: ChatZilla 0.9.86 [Firefox 3.5.5/20091102152451])
02:11:54Unhelpfulsaratoga: in terms of number of MDCTs to decode X samples? how complex are the non-mdct parts? even if they spend as much *time* in MDCT if vorbis spends more time in its other code it will see less improvement in % terms
02:12:12saratogathey started taking almost identical decode times
02:12:43saratogabut WMA apparently got 11MHz faster, while vorbis only about 5
02:13:13saratogaonly thing i can think of is that long blocks are somehow a lot slower then short blocks, and vorbis uses a lot more long blocks then wma
02:14:22saratogatime to compare PP and see if thats teh same . . .
02:14:49saratogawhat the fuck I didn't get a test_codec log
02:14:57saratogaand it ran through to completion
02:15:16*TheSeven also had that sometimes
02:15:29TheSevenseems to happen if you insert usb while test_codec is still running
02:18:50saratogadoes fdprintf not write to the file immediately?
02:20:16Unhelpfulsaratoga: well surely complexity is >linear wrt block size?
02:20:46TheSevensaratoga: iirc it had something to do with writing the directory entry for it
02:20:46saratogayes, n*logn
02:20:55TheSeventhe log itself is probably somewhere on the storage
02:21:08saratogabut i wouldn't expect two codecs at the same bitrate to use tremendously different block sizes
02:21:34TheSevenafter disconnecting, rockbox might panic because it tries to finish whatever was left storing it, and things have changed during the usb connection
02:21:43Unhelpfulthey might, i'm sure their long/short block selection is entirely different
02:24:08saratogaTheSeven: thanks, found it on the disk in a lost fragment
02:24:08 Quit ecio (Read error: Connection reset by peer)
02:25:21 Join ecio [0] (
02:25:23saratogaTheSeven: any idea how to make test_codec more robust against lost data?
02:25:31saratogathe plugin didn't even crash so I'm surprised this happened
02:25:35saratogait seemed to complete normally
02:25:47TheSevenit's probably still running in the background or something
02:25:54TheSevenor it just didn't close the log file handle
02:26:46TheSeveni don't know exactly what's going on inside, but this definitely isn't the first time such a thing happened...
02:27:41TheSevenand iirc the fat code doesn't do any write caching (or it would at least flush it prior to an unmount caused by usb connection), so it must be the plugin that's at fault here
02:28:35saratogaUnhelpful: my recovered logs indicate that on PP WMA and Vorbis get identical speed ups
02:28:48saratogaso its not the transform, but something special about the nano2g
02:28:58 Join teru [0] (
02:29:09saratogaperhaps we inadvertently fixed something else
02:31:42 Join ecio1 [0] (
02:31:42 Quit ecio (Read error: Connection reset by peer)
02:36:06saratogaWMA is now 1.74 times as fast as it was on Stripwax's original test 4 days after it was committed
02:36:43saratogaand about 3.75 times as fast as the very first time I tried to play a WMA file in rockbox on my Sansa
02:42:14Unhelpful1.74 *times* as fast? yikes.
02:44:48 Join Uretti [0] (
02:47:42saratogathe first factor of two speed up was so much easier, took me a day or so and all I had to do was add some IRAM and assembly optimized MUL functions :)
02:47:57saratogathe next factor of 1.75 took me 2.5 years . . .
02:48:35saratogai just noticed theres an open source DTS codec, so in theory we could add libdca as a codec
02:48:43saratogathough i guess no one wants to use DTS in rockbox
02:50:07Unhelpfulsaratoga: on beast. in mpegplayer. ;)
02:50:51saratogaac3 would make a lot of sense for mpegplayer
02:51:00saratogasince its a valid mpeg audio codec and its way faster then mp3
02:51:14saratogaand also for doing dvds on the beast :)
02:51:21Unhelpfulmpegplayer on beast might conceivably play VOBs straight from DVD if it had AC3 and DTS audio support
02:51:27Unhelpfulyes, that is what i was thinking mostly.
02:51:43Unhelpfulmaybe support DVD-ROM drive via USB host? ;)
02:52:18*Unhelpful grumbles about how slow and/or useless mlas is on arm :/
02:52:29 Quit Tomis (Quit: Tomis)
02:53:03saratogawould closing the log and reopening it between files make it less likely to be corrupted in test_codec?
02:53:03Unhelpfulit's as slow as a mul and an adds, *and* adds would set carry on the add result... although it would not set anything useful from the mul result
02:53:13saratogai don't really understand why a file gets lost in the first place
02:53:26Unhelpfuli would think so, i looked into that earlier and couldn't find any evidence that we had an fsync :/
02:55:30 Quit Strife89 (Quit: Reboot to investigate BIOS fan settings.)
02:57:33***Saving seen data "./dancer.seen"
02:57:49 Join moos [0] (moos@rockbox/staff/moos)
02:59:06saratogaUnhelpful: think this will help?
03:00:51Unhelpfulsaratoga: i'd think so, but i don't really know why it should in the case that test_codec completes without hanging.
03:01:06Unhelpfuli don't really know why anything should go wrong in that case. :/
03:01:38 Quit Schmogel (Quit: Miranda IM! Smaller, Faster, Easier.
03:01:53saratogacould be a codec error or something
03:01:54 Join Strife89 [0] (
03:02:54notByanI just noticed that in my new bleeding edge build, my ipod 5G line out now uses volume control and EQ
03:03:02notByanwhats with that?
03:03:29saratogathat sounds like a bug
03:03:36saratogawhen was the last build that didn't do it?
03:04:20notByanI only have bleeding edge because I like to make a small edit on the source code
03:04:25saratogaactually I bet I know
03:04:28saratoga"Allow volume control down to -90 dB for WM8758 (e.g. iPod Video). (FS #10082)"
03:04:28notByanI don't really keep up to date with it
03:04:36saratogasounds likely to be the commit in question
03:04:56TheSevensaratoga: this one shouldn't have anything to do with the eq, but it would have been my first quess, too
03:05:13TheSevenbut this was before 3.5
03:05:27saratoga"When lowering the volume below -57dB the lineout is also affected."
03:05:43TheSevennotByan: does this also happen with the 3.5 release?
03:05:46saratogawell that seems easy enough to work around, don't lower the volume below -57dB
03:05:58notByansaratoga: but.. the EQ?
03:06:16 Join Tomis [0] (~Tomis@
03:06:33saratogaline out always used EQ
03:06:38saratogaif thats what you're asking
03:07:15notByanTheSeven: I have no idea, as I said, I've been using bleedge edge cause I need to build it to make the small cahgne anyway
03:07:18notByansaratoga: you sure...
03:07:22notByancause I could've sworn..
03:07:44saratogaEQ is done in software like a mile upstream of the DAC
03:08:05saratogachanges the output line on the DAC has no impact on how DSP software in rockbox works
03:08:06notByanI must be crazy then
03:08:24saratogaonly thing i can think of that might be confusing here is that theres some hardware bass boost on the DAC or something
03:08:26Unhelpfula device couldn't support (software) EQ on one output and not another unless there were two DACs
03:08:32TheSevensaratoga: aren't we also using hardware eqs on some targets at least?
03:08:47saratogamaybe, but i think thats not even available from the EQ screen
03:09:27notByanI am very tempted to put my old build on here and see what happens
03:09:27TheSeveniirc (and i have only very briefly looked at that kind of stuff) on devices that support it, the more trivial eq settings are mapped to the hw eq
03:10:22saratogaif the hardware adjustments don't impact the line out I doubt thats possible to adjust in software
03:10:23 Join S_a_i_n_t [0] (S_a_i_n_t@
03:11:06notByanbtw, I guess I was talking about the treble and bass settings in sound settings
03:11:07notByannot the EQ
03:11:16notByanI didn't even realize there was a seperate EQ menu
03:11:19saratogayeah those are hardware
03:11:31notByanshould they be effected the line out?
03:11:32notByancause they are
03:12:36*TheSeven doesn't see how this could change, as it is done in hardware... or is there a separate set of regs for this stuff on lineout which we are accidentally writing? can't really imagine that though...
03:12:43*Unhelpful imagines that may depend on the DAC etc
03:12:45 Quit phanboy4 (Read error: Connection reset by peer)
03:13:37TheSevenwell, he didn't change his dac, obviously...
03:14:13 Quit GeekShadow (Read error: Connection reset by peer)
03:14:17TheSevennotByan: if possible, you should try the 3.5 build. this would help narrowing down where that issue is coming from.
03:14:33notByanyeah, sure
03:14:45UnhelpfulTheSeven: i mean that whether it "should" effect line out would depend on the DAC and possibly its configuration.
03:15:25notByanso, help me understand something, where in hardware is the treble and bass stuff being EQ'ed?
03:16:25notByanI am gonna try my old build real quick and see if it happens there
03:16:48saratogatry pre 3.5
03:16:52saratogathat has the DAC changes in it already
03:17:09notByanold build is pre 3.5
03:17:51saratogaapps\plugins\fft needs to go away
03:18:08saratogacan you just do ../codecs/lib from inside a plugin or will that not work?
03:18:48notByanI don't think I even updated the apps/plugins/fft last time
03:18:50notByanhrm >_>
03:19:09saratoganotByan: I wasn't referring to you
03:19:52notByanhrm, I guess I was crazy
03:20:06notByanbut.. this release sounds much better..
03:20:35notByanI thought it was in my head
03:20:37notByanbut it definately does
03:21:00 Quit ecio1 (Read error: Connection reset by peer)
03:21:06 Quit DerPapst (Quit: Leaving.)
03:21:25*notByan switches again to makes ure
03:22:02 Quit TheSphinX^ (Ping timeout: 240 seconds)
03:22:10 Join ecio [0] (
03:24:25 Quit n17ikh (Remote host closed the connection)
03:25:37 Join TheSphinX^ [0] (
03:26:23 Join n17ikh [0] (
03:32:33 Quit AsaelReiter (Quit: CGI:IRC)
03:48:01JdGordonteru: pong
03:52:36teruJdGordon: it seems like that file descriptor for font file doesn't seems to be closed when reloading skin file.
03:53:17JdGordonnuts, and I see why
03:53:52CIA-8New commit by 03jdgordon (r24803): make sure font fd's get closed when a new skin is loaded
03:56:12 Join robin0800 [0] (
03:59:48 Quit efyx_ (Remote host closed the connection)
04:09:47 Quit TheSeven (Disconnected by services)
04:09:59 Join The_Seven [0] (~theseven@rockbox/developer/TheSeven)
04:10:09 Nick The_Seven is now known as TheSeven (~theseven@rockbox/developer/TheSeven)
04:11:48 Quit TheSphinX^ (Quit: XChat)
04:15:44 Quit Strife89 (Read error: Connection reset by peer)
04:20:43 Quit Res1 (Ping timeout: 240 seconds)
04:32:32 Quit Adnyxo (Remote host closed the connection)
04:33:19 Quit SeismicMike (Quit: Hwyl)
04:44:52 Join Barahir_ [0] (
04:47:30 Quit Barahir (Ping timeout: 240 seconds)
04:57:37***Saving seen data "./dancer.seen"
05:01:46 Quit xavieran (Ping timeout: 256 seconds)
05:26:04 Join qcjn [0] (
05:29:53qcjnhi, i 've got a sansa e280 V2, i like it, but one thing i don't like is the way the music is handled. I'd prefer to be able to see it, just as folder, would be easier to find. So i was wondering if rockbox does that, and if i'll lose the features like the radio, and recording that's on my sansa ?
05:31:09S_a_i_n_tqcjn: you can give Rockbox a 'trial run' by downloading the simulator
05:31:29qcjnoh, yeah, thats great
05:31:44qcjni m on ubuntu
05:31:53qcjndoes it work on ubuntu
05:32:15S_a_i_n_tAh...not sure about that actually...
05:32:33qcjnanyway, i still have xp, but it s so slow
05:32:36notByanit likely does.. but I don't think there are binaries available for linux
05:32:51 Quit karashata (Quit: The fluffy dragon has left completely!)
05:33:02S_a_i_n_tIf you have XP too...than I don;t see a problem :D
05:33:24qcjnit s just that i don't use it anymore, it s to slow
05:34:08notByananyway, rockbox has a files menu that lets you browse folders and manage music that way
05:34:21moosqcjn: if it helps
05:34:33S_a_i_n_tand no, you won't lose recording or FM radio IIRC
05:34:48qcjnok, great
05:35:14moosyou can even records the radio ;)
05:35:45qcjnwhere is the simulator"> here it is
05:37:05moosnot up to date, but enough to see
05:37:10robin0800qcjn: in extras on the front page
05:37:17S_a_i_n_tit's at 3.5 IIRC
05:37:35notByanqcjn: if you have wine, you should able to run it that way
05:37:45qcjnoh, yeah
05:38:23notByanit uses sdl, which works natively on linux, so it should work flawlessly in wine
05:39:15qcjni don't know what is sdl, cause i almost never uses wine, cause usually people uses it mostly for games, and i don't play games
05:41:06 Join xavieran [0] (
05:41:49notByansdl is a cross platform graphics library
05:41:53moosqcjn: just try on the device! Nothing to lose, and you have dualboot, and in last case yiu can remove it and goes back to evil original firmware ;)
05:42:30qcjnit s dualboot ! incredible, wow
05:42:54qcjnthere s some sort of grub at start up ?
05:43:07moosyup check the manual, just to hold a key on boot to load sansa crap
05:43:13 Join froggyman [0] (
05:43:40qcjnwow, men it s great,
05:44:09moosRockbox rules the world!!! ;)
05:45:02*S_a_i_n_t thinks if RB *wasn't* any better than the OF...noone would use it, and we wouldn;t be here :D
05:45:18qcjnwell i ve installed it, about 2 years ago on my daughter in law's sansa, and she liked it
05:46:44qcjncause her firmware was doomed...but since i wasn't using it, and in fact wasn't using my sansa...i didn't try it myself
05:47:30moosS_a_i_n_t: ...since years and years now
05:48:47S_a_i_n_tI cant imaging *not* using RB on my iPods turns them from "fashion crap" into an actual, functional DAP :D
05:50:04qcjnwell since it s dual boot... i think you just found a new user :)
05:50:09FarthenS_a_i_n_t: so now it is a fashionable AND functional player ;-)
05:50:18moosS_a_i_n_t: next step, you will buy your futur DAPs because rockbox suport.... :)
05:50:49S_a_i_n_tnext step?...Ha! that's the *current* step!
05:51:08moosyou already had your ipod, no?
05:51:21S_a_i_n_ta few of them..
05:51:28S_a_i_n_tbut I bought *more* :P
05:52:08moosyou will no more buy an DAP if not Rcbx, or at least hope to ;)
05:52:36moosweee then !!! \o/ you are on the matrix dude
05:52:56 Quit robin0800 (Remote host closed the connection)
05:54:09 Join robin0800 [0] (
05:57:35 Quit robin0800 (Remote host closed the connection)
05:57:55 Join robin0800 [0] (
06:02:05 Join phanboy4 [0] (
06:02:13 Quit Barahir_ (Ping timeout: 264 seconds)
06:07:29 Join Llorean [0] (
06:07:29 Quit Llorean (Changing host)
06:07:30 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
06:09:10 Quit Sajber^ (Quit: Leaving.)
06:10:36 Quit linuxguy3 (Ping timeout: 264 seconds)
06:12:28 Join linuxguy3 [0] (
06:19:56 Join S_a_i_n_t_ [0] (S_a_i_n_t@
06:20:35 Quit S_a_i_n_t (Ping timeout: 245 seconds)
06:21:19 Join ucchan [0] (
06:21:57 Nick S_a_i_n_t_ is now known as S_a_i_n_t (S_a_i_n_t@
06:23:40 Quit Farthen (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
06:25:00 Quit planetbeing (Ping timeout: 264 seconds)
06:27:38qcjnwaht is DAP ?
06:27:55qcjndigital audio player ?
06:33:43 Quit CGL (Quit: Ya está, ahora solo a configurar XD)
06:35:55ucchanThere is asking. Is there developper who can change my (username: Uchida) right of Flyspray ?
06:36:15ucchanI want you to change right of "close task" to "yes". (see FS #10424)
06:39:07S_a_i_n_tqcjn: yes, DAP = Digital Audio Player
06:40:57 Quit ucchan (Quit: Leaving...)
06:42:58 Quit phanboy4 (Read error: Connection reset by peer)
06:43:42 Join phanboy4 [0] (
06:55:55 Part toffe82
06:56:23 Quit anewuser (Quit: =ooo ϢINTER ϾHIP 5iVE is OOON!!)
06:56:25 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
06:56:59 Join ucchan [0] (
06:57:40***Saving seen data "./dancer.seen"
07:02:50 Quit ucchan (Quit: Leaving...)
07:03:48saratogais there someway in a plugin to get the SVN version number?
07:03:57saratogatest_codec should really print that in the logs
07:07:20 Quit robin0800 (Remote host closed the connection)
07:16:51 Join karashata [0] (
07:17:10mc2739saratoga: I think so, battery_bench does it.
07:18:19 Quit panni_ (Read error: Connection reset by peer)
07:19:40*S_a_i_n_t finds FS #11042 and decides this is indeed a *damn* good idea...
07:21:47 Quit kramer3d (Quit: Leaving)
07:32:56 Quit froggyman (Quit: kthnxbai)
07:34:23 Join Horschti [0] (~Horscht2@xbmc/user/horscht)
07:37:15 Quit Horscht (Ping timeout: 245 seconds)
07:44:53 Join Barahir [0] (
07:53:26 Join mt [0] (~mtee@rockbox/developer/mt)
08:00:16qcjnthanks guy's . i'll beback surely
08:01:09 Quit qcjn (Quit: Ex-Chat)
08:12:57 Quit Barahir (Ping timeout: 252 seconds)
08:26:14amiconnsaratoga: Close and reopen is definitely overkill. Calling fsync() is enough
08:28:05amiconnAnd unrelated to that, you can make perl chomp() handle \r\n if you set $/ to \r\n before. But it's probably better to use a regular expression here
08:29:40amiconnUnhelpful: dvd vob support wouldn't require any other audio codec than ac3, 'cause an ac3 track is mandatory afaik
08:42:49saratogaamiconn: why does that help?
08:42:57saratogai don't really understand the problem
08:44:25amiconnThe fat, directory entry is only updated on file close, or when calling fsync
08:44:44 Join stripwax [0] (
08:45:05amiconnIf test_codec crashes or otherwise ends without closing its log file, you get a zero-byte file now, because test_codec does neither
08:46:16amiconnFor some reason test_codec may cause (low bat?) shutdown on PP devices using the PCF5060x PMU, even if there is no actual low bat condition
08:47:34amiconnHmm, actually not only this PMU is affected
08:48:09amiconnI observe this effect on my Sansa c200v1, iPod Photo and Mini G2
08:48:18amiconnThe H10 (small) is unaffected for some reason
08:49:16saratogaamiconn: i had test_codec quite normally and still got a 0 byte file
08:49:26saratogaso i assume theres some bug in test_codec itself
08:49:33amiconnIt's reproducible by testing ape -c5000. That's why there are no values for PP5022
08:49:46saratogayes I saw that as well, but i got it even with APE removed
08:50:01amiconn(cause neither c200 nor Mini G2 work - the PP5020 value is tested on H10)
08:50:41amiconnYou don't need to remove all ape, just the -c5000 file
08:50:59saratogawhen I dumped the disk to get the lost file, the entire test codec log was there
08:51:07saratogaso it finished everything, it just somehow never closed the file
08:51:08stripwaxI get a zero-byte file if I connect USB without selecting 'quit' from the test_codec menu. I guess test_codec doesn't regard usb connect as a normal way to exit.
08:51:27saratogahmm that should be easy to fix
08:51:41amiconntest_codec doesn't have proper usb handling. It's a test plugin after all
08:52:33amiconnI didn't observe zero-byte test_codec log files on normal exit so far
08:52:34 Quit saratoga (Quit: Page closed)
08:52:44amiconnBut then I don't use the log feature that often
08:53:43amiconnCalling fsync() is proably still a good idea, because of potential codec crashes
08:54:10*amiconn has no real idea how to track this weird bug down
08:56:30stripwaxany chance there's some built-in protection to prevent running at full boost for long periods of time?
08:57:15stripwaxor is it demonstrably test
08:57:24stripwaxor is it demonstrable test_codec only
08:57:43***Saving seen data "./dancer.seen"
09:00:51amiconnTheSeven: I've updated SoundCodecMonkeysAudio from your pasted test results
09:07:40 Quit mt (Ping timeout: 245 seconds)
09:24:03 Join TopyMobile__ [0] (
09:28:17 Quit TopyMobile_ (Ping timeout: 260 seconds)
09:35:59 Join stoffel [0] (
09:39:19 Quit CaptainKwel (Ping timeout: 272 seconds)
10:06:41 Quit phanboy4 (Ping timeout: 248 seconds)
10:17:05 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
10:18:03*S_a_i_n_t has some BIG issues with alternating sublines...
10:21:07pixelmaJdGordon: there is something very fishy about the FMSs. I get the same problem with an own .fms on my M5 by the way (can't stop the radio anymore, other controls work), I set the .fms through the Theme settings menu now if that makes a difference - maybe related: there is something weird about applying the setting
10:21:12S_a_i_n_t*issue being, when a viewport is loaded, but not displayed (ie. if it's conditional is not met). when it *is* displayed any alternating sublines that it contains are displayed 'en-mass'. That is to say, if the vieport is not visible for say 1 minute, and it has alternating sublines which display every 10 seconds...when the viewport *is* displayed, it will display all 6 alternations before it resumes normal function.
10:24:23S_a_i_n_tI can provide a WPS that proves this quite effectively.
10:30:01pixelmadidn't you tell about this earlier? And is there a tracker entry for this?
10:31:12pixelmahmmm... and while my theme uses sublines and alternating sublines, there are none of the latter in conditional viewports so I haven't seen this
10:32:28S_a_i_n_tpixelma: at the time I first brought this up, I wasn;t sertain that it wasn't my code at fault...but not I am (kinda...)
10:32:42S_a_i_n_tI'm surprised noone else has noticed it though...
10:34:37pixelmaI think I read a report in the forums (but maybe that was you, don't know). I usually test those things with WPSs etc. that only contain the few lines which are needed to demonstrate a bug or test a new tag.
10:34:49 Join flydutch [0] (
10:35:37S_a_i_n_t <- if anyone wants to see it for themself, or if anyone can see an obvious flaw in my code.
10:36:03S_a_i_n_tpixelma: don't think it was me...pretty sure it wasn't actually.
10:36:45S_a_i_n_tif you use the WPS I posted (Nano 1/2g), put hold on for a minute or so, then take it off.
10:36:53pixelmareduces potential other causes of error and makes it easier to read for someone who wants to work with it. E.g. my current test .fms only contains %Tn (current preset name)...
10:44:55 Join Barahir [0] (
10:47:22 Join tuma [0] (
10:50:13tumaif I have understood correctly, time stretching should be functioning from 3.4 or 3.5, right? I have enabled this feature, but it seems not to work (tried mp3 and spx files)
10:51:16tumaI tried with default configuration, with only difference that this feature was enabled
10:54:44LloreanWhat are you expecting it to do if you don't set the rate?
10:55:28 Join pamaury [0] (
10:56:13tumaof course I set also pitch rate, right
10:57:46***Saving seen data "./dancer.seen"
10:58:03tumaand as a effect I get pitch going up/down (fast playback with higher pitch)
11:00:46tumaalthough I would assume fast playback with normal pitch, if I have understood this feature corretly.
11:02:36pamaurymc2739: are you there ?
11:09:14tumais there anyone that has been using time stretching? I would like to know if it works for someone.
11:10:06S_a_i_n_ttuma: it works perfectly for me...
11:10:14stripwaxtuma - have you check the relevant manual section? rather than assume how it works.
11:10:38tumaI actually once read that, but maybe I should double-check
11:12:26tumayes, i re-checked section 6.10 in manual.
11:12:49tumashould be straight-forward
11:13:04tumaand I believe I have tried to use it correctly.
11:13:28stripwaxAnd did you reboot after enabling the Timestretch option?
11:14:16 Join mt [0] (~mtee@rockbox/developer/mt)
11:14:24S_a_i_n_ttuma: which target?
11:14:34tumasansa e200 (e280)
11:16:35stripwaxtuma - have you read section 4.3.3 too? and put it into separate speed/pitch mode?
11:17:30*stripwax observes that the manual section 4.3.3 doesn't actually tell you HOW to put it into separate speed/pitch mode. sigh.
11:18:06stripwaxtuma - I don't know which button you need for e280, on ipod video I press Down. then, scroll affects the pitch while left/right affect the speed
11:19:32stripwaxoh wait yeah it does. First row in the table! for e280 it's the Rec button
11:19:42*stripwax needs some caffeine. for his eyes.
11:21:38stripwaxtuma - works?
11:24:11tumayes it does
11:24:19*stripwax \o/
11:25:36S_a_i_n_tpixelma: I cut it back as far as I can (and still reproduce the problem), HOLD toggles the viewports.
11:26:32linuxstbamiconn: My recollection is that NTSC DVD players *must* support LPCM and AC-3, and PAL players *must* support LPCM, AC-3 and MP2.
11:27:17amiconnAh yes, lpcm needs to be supported as well. I don't know about mp2 (but then mpegplayer already does support mp1/2/3)
11:27:31tumastripwax: really, not very intuitive this feature... have to read the manual very carefully to use it
11:27:50stripwaxtuma - feel free to submit a patch for the manual! :-p
11:28:05tumamaybe I should=)
11:28:14linuxstbamiconn: But in practice, I don't think you'll find many DVDs with only MP2 (or only LPCM). Even music DVDs seem to have AC-3 in addition to any LPCM track.
11:28:37tumamaybe pitch screen could also hint about mode switching
11:28:54tumathere's plenty of space on screen
11:29:10 Join Highlander [0] (
11:29:12pixelmadepends on the screen...
11:29:13amiconnWell my dvd recorder can (optionally) produce dvds with just lpcm at the highest quality setting (1 hour on single layer)
11:29:39pixelmaand the font you uses
11:29:40amiconnFor all other quality levels it always uses ac3 2.0
11:30:48linuxstbLPCM is pretty easy to decode and play. I have some code somewhere to do it.
11:31:13 Join dantje [0] (
11:35:18 Quit karashata (Quit: The fluffy dragon has left completely!)
11:39:01 Join dfkt [0] (dfkt@unaffiliated/dfkt)
11:40:10TheSevenstripwax: did you see my results from yesterday?
11:40:14tumastripwax: actually I have submitted two small (but for me, important) patches to the code and they have got no attention at all. so I am not sure if small and rather small-significance patches get applied easily
11:41:50stripwaxtuma - which FS#s?
11:42:10stripwaxTheSeven - your pdf? yeah, I saw that, really cool. Decent speed up from new MDCT on nano2g too, which is reassuring
11:42:47TheSevenyes - the wma speedup is a bit suspicious though ;-)
11:43:06stripwaxWould be great it we could systemise that even more and just have it viewable from wiki - pick a target, pick a date range ...
11:43:21tumaFS #10482 and FS #10483
11:43:22stripwaxTheSeven - suspicious? not at all!
11:43:23TheSeventhat's what i suggested
11:43:33stripwaxTheSeven - ah! heh
11:43:52TheSeven[22:36]* TheSevenwonders if it would be worth the effort to 1) add the device, build type and revision to the test-codec output, 2) make a website where one can upload the logs, 3) make an analyzer tool for the gathered results :-)
11:44:28stripwaxyeah, it's a great idea. and even if we only get results for all targets once per release, it's still great data
11:44:54TheSeventhe wma speedup suggests that you accidentally fixed something else at the same time :-)
11:45:18TheSeveneverything else gained like 4 or 5 MHz, while WMA gained 11
11:46:23 Join DerPapst [0] (
11:47:15stripwaxTheSeven - WMA uses smaller block sizes than Vorbis, iirc. the new ifft uses cache optimally (in the sense that recurses depth-first rather than breadth-first). maybe WMA blocksizes mostly fit in cache and get bigger benefits as a result? old mdct was breadth first
11:47:41stripwaxI certainly didn't explicitly fix anything in WMA :)
11:51:07TheSevenactually this kind of behavior seems to be nano2g-specific
11:51:44TheSevenwma used to be slower than it should be in comparison to other targets and other codecs, and now it has seen this weird speedup, while wma on other targets also got one in the ~5MHz range.
11:51:58 Join stripwax_ [0] (
11:52:22TheSeven[11:49]<TheSeven>actually this kind of behavior seems to be nano2g-specific
11:52:23TheSeven[11:50]<TheSeven>wma used to be slower than it should be in comparison to other targets and other codecs, and now it has seen this weird speedup, while wma on other targets also got one in the ~5MHz range.
11:52:25TheSevenso you say this might be related to caches? that could be an explanation, as nano2g has only 2x4K of cache
11:54:24 Quit stripwax (Ping timeout: 264 seconds)
11:55:01stripwax_yep, could be.
11:55:05 Nick stripwax_ is now known as stripwax (
12:00:02tumabye for now! and thanks again, stripwax, S_a_i_n_t
12:00:03 Quit tuma (Quit: leaving)
12:02:55pixelmahmm, got a few IMA-ADPCM files which play in Rockbox now but the length is displayed wrong (just 0:01)
12:15:10 Join MethoS- [0] (~clemens@
12:16:40pixelmaAlexP: in your FMS test, did you use the menu entry to set the .fms or did you still edit the .cfg?
12:23:29AlexPpixelma: both I think
12:24:29AlexPJdGordon: Could your latest commit fix an issue where the wrong font is used on the remote?
12:25:32AlexPI haven't tried yet, but as you may guess my testing on H140 has hit an issue :)
12:26:37pixelmado you just have one fms to test or could you try setting another (through the menu)?
12:27:37 Nick fxb__ is now known as fxb (
12:30:43AlexPpixelma: I just have one, but I could write a quick other to test
12:30:51AlexPI take it you are having issues with it?
12:33:33AlexPIt'll have to be a bit later today though, I'll let you know
12:34:39 Quit leavittx (Remote host closed the connection)
12:35:02pixelmathere are some very weird effects - on my OndioFM as well as on my M5 - first the setting seems to apply correctly but then trouble start, it's a bit longish to explain here, I really need to put it on the tracker
12:35:40pixelmaon both I can't stop the radio anymore
12:38:36pixelmaand since there are also problems with the setting being kept or not (or being able to chose a new .fms), I wonder if it is related
12:40:39 Join Highlander_ [0] (
12:43:53 Quit Highlander (Ping timeout: 240 seconds)
12:50:43 Quit avacore (*.net *.split)
12:50:43 Quit kadoban (*.net *.split)
12:50:43 Quit Guest82304 (*.net *.split)
12:50:44 Quit rasher (*.net *.split)
12:52:19 Join kadoban [0] (
12:54:43 Join feisar [0] (
12:55:12 Nick feisar is now known as Guest34960 (
12:56:04 Join avacore [0] (
12:57:48***Saving seen data "./dancer.seen"
12:59:07 Join rasher [0] (
12:59:07 Quit rasher (Changing host)
12:59:08 Join rasher [0] (~rasher@rockbox/developer/rasher)
13:01:17CIA-8New commit by 03teru (r24804): FS #11017: add Playlist Catalog and Properties item to database's context menu.
13:03:52pixelmaAlexP: including the weirdness that the patch's hardcoded fms works and doesn't give me trouble with the controls
13:14:24 Join AsaelReiter [0] (
13:18:12AsaelReiterIs there a way to create a diff with the differences between two diff files?
13:21:34TheSevenapply them and diff the results
13:23:15AsaelReitertheseven: thanks
13:23:36AsaelReiterI hoped that there is a simpler way
13:24:42 Join Farthen [0] (
13:29:16CIA-8New commit by 03b0hoon (r24805): Packard Bell Vibe 500: Enable recording.
13:34:19pixelmaAlexP: and which targets did you test on?
13:34:20TheSevenis there a way to change the name of the .rockbox dir?
13:34:37 Nick fxb is now known as fxb__ (
13:35:05TheSeveni would like to have different installations for a "production rockbox" and a testing one
13:36:11stripwaxTheSeven - you could keep a copy of both (prod.rockbox, testing.rockbox) and copy the one you want to .rockbox? There's no way to select at boot time which to point to
13:36:18TheSevena quick grep shows that there seem to be quite a few places where ".rockbox" is hardcoded
13:36:33AlexPpixelma: S60, and I'm trying H120, but there are some font issues so I haven't got to the fms yet
13:37:02TheSevenstripwax: if i could patch the main binary to use one or the other directory, this should work, as i would like to have 2 main binaries anyways
13:37:05 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
13:37:27TheSeveni.e. i would have 2 complete installations that don't need to know anything about each other
13:37:39 Quit AsaelReiter (Quit: CGI:IRC)
13:37:40 Quit piotrekm (Read error: Connection reset by peer)
13:37:57 Join piotrekm [0] (
13:37:57 Quit piotrekm (Changing host)
13:37:57 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
13:38:13stripwaxTheSeven - and then use ROLO to switch between them?
13:38:24stripwaxi.e. without having to patch bootloader too
13:38:24TheSevenno, that would be done at bootloader level
13:39:01TheSeveni'm not using the rockbox bootloader, so this won't be an issue
13:39:17TheSeveniloader can do that out of the box
13:40:09stripwaxgotcha. Well, I don't know of a built-in way to do that, you may just need to put in a #define for it, set it as a config option, and replace the hardcoded occurrences with the symbol.
13:40:45TheSeventhere *is* a define for that (in the makefile, even), but it doesn't seem to be used consistently :-/
13:42:29stripwaxoh, ugh.
13:42:48TheSevengah, i'm going to fix that configure bug now, this is really annoying
13:43:24TheSevenaha, another rbdir=".rockbox" in .configure... :-/
13:43:35TheSevenseems to be only a default though
13:47:23CIA-8New commit by 03theseven (r24806): Make configure remember that it is supposed to use the EABI toolchain when make reconf is run
13:57:39 Quit thegeek (Read error: Connection reset by peer)
14:02:21 Join efyx_ [0] (
14:09:36 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
14:11:58 Quit shai (Quit: Leaving)
14:25:47 Join robin0800 [0] (
14:32:05 Quit n17ikh (Ping timeout: 276 seconds)
14:32:54CIA-8New commit by 03teru (r24807): remove unused fields from struct gui_synclist.
14:34:43 Quit dantje (Quit: Ex-Chat)
14:37:20 Join n17ikh [0] (
14:38:35pamauryI'm investigating "FS #10913 - File Browser not updated properly upon microSD insert/removal with dircache enabled" and I think I found out the issue. It seems that there is a synchronization problem with SYS_FS_CHANGED: depending on the order in which the thread treat the message is important. Perhaps we need something like this: a first message to inform threads that need internal structure update, then a ack, and then the final t
14:38:35pamauryrigger message ?
14:48:13pamaurynobody interested ?
14:57:45 Join CGL1 [0] (~CGL@
14:57:49***Saving seen data "./dancer.seen"
14:58:28 Quit Highlander_ (Quit: Quitte)
14:58:49 Part CGL1
14:59:00 Join CGL [0] (~CGL@
15:07:46 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:10:13*amiconn just tested his theory that IRAM on PP5020 (not 5022) is not strictly single cycle
15:10:46amiconnIt is correct - and the extra latency even depends on the cpu clock
15:12:36kugelpamaury: I don't see an obvious other way to do it
15:13:17amiconnAccessing the same address over and over is single cycle. Striding across iram isn't
15:13:41 Quit CGL (Quit: Saliendo)
15:13:47*amiconn will test various patterns to find out what's going on exactly
15:13:51pamaurykugel: the problem I see with what I propose is that now, all threads have to monitor a new message and ack, even the one which are not "interested" in hotswap
15:13:53 Join CGL [0] (~CGL@
15:14:44stripwaxamiconn - meaning ldm even more important on pp5020 than on pp5022?
15:15:12kugelwhat threads are involved? maybe the initial post could go to the dircache thread directly which then does SYS_FS_CHANGED?
15:15:18 Quit teru (Quit: Quit)
15:15:20amiconnMeaning that PP5020 needs very careful decisions what should go into iram and what shouldn't
15:15:52Bagdermaybe stable and unstable should rather be in two columns on the front page, to make it less tall
15:16:01amiconnIt means iram is *not* always equally fast or faster as dram, taking the cache into consideration
15:18:01pamaurykugel: to me, that's the most logical way. Currently, the ata/mmc/sd/... code emits the SYS_FS_CHANGED message whereas it has nothing to do with FS. The SYS_FS_CHANGED should be sent by the FS code.
15:18:44 Join likemindead [0] (~633a72d3@gateway/web/freenode/x-okjazhaqvgknbjru)
15:19:54likemindeadMy iPod (4th Gen/Color) won't mount in Xubuntu 9.10. How can I do that manually? I want to upgrade to Rockbox 3.5 :D
15:21:27 Part likemindead
15:21:28stripwaxlikemindead - if it doesn't mount even after booting to apple firmware and/or apple diskmode, then that sounds like a Xubuntu question not a rockbox question..
15:21:37stripwaxoh. well. seeya.
15:22:51 Quit kugel (Ping timeout: 265 seconds)
15:26:09 Join Omlet [0] (
15:26:40mt stripwax: Is the todo on fasterMDCT up to date ?
15:26:44mtOh and .. \o/ .. for merging :)
15:26:56stripwaxmt - yep, I believe so, I updated it yesterday
15:27:05stripwaxand \o/ thanks :_)
15:27:21mtgreat !
15:28:09 Nick YPSY is now known as Ypsy (
15:30:02gevaertsBagder: personally I'd like the table to be less tall, but I'm a bit worried that it would be more confusing that way
15:30:22Bagderyeah, quite possibly
15:31:32 Join Schmogel [0] (
15:34:14 Join Sajber^ [0] (~Sajber^
15:36:41 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:37:04 Quit mc2739 (Ping timeout: 246 seconds)
15:39:08 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
15:39:28 Join tuma [0] (
15:40:14kugelpamaury: hm, that would make it harder to close/not even start the dircache thread if it's disabled though (what I would like to see in the future)
15:40:54kugelwho can fix the sidebar in the forums?
15:41:06pamaurykugel: true but that is not logical currently
15:41:58 Quit stoffel (Ping timeout: 246 seconds)
15:42:56pamaurythe ata code has nothing to do with the file system code so why should the ata code trigger a FS changed ?
15:48:57kugelI agree
15:54:55pamauryI will think about it, perhaps there is a simple solution to this problem
15:56:04 Join TheSphinX^ [0] (
16:00:51kugelTorne: would it be useful to align skin allocations to 16bytes for dma?
16:01:15kugelfor images for example, but I can imagine it's not worth it
16:03:44kugelJdGordon: this bidirectional skin buffer makes resizing on boot hard
16:04:03 Part fleebailey33
16:11:32TheSevengevaerts: is it normal that a pebble-build doesn't have a backdrop with default settings?
16:12:07gevaertsit shoudln't really change anything there
16:12:11 Quit TheSphinX^ (Ping timeout: 252 seconds)
16:14:03TheSevenskin buffer too small or something?
16:14:16TheSeveni saw something in the patch that sets it to 16K
16:14:50kugelhow can I get a file listing a the directory, and if I get it search for a file ".skin_buffer_XXXX" where X is a number, and if I have that file how would I parse the XXXX number?
16:15:04*kugel is obviously a fs operation noob
16:15:18TheSevenoh, on the wps most things are also missing
16:15:36Unhelpfulkugel: as in dma transfer from disk?
16:15:45kugelUnhelpful: yes
16:16:09Unhelpfulthat's not the right buffer to resize for that. ;)
16:16:25Unhelpfulerm, align, rather.
16:16:45TheSevenhm, stkov usb when trying to do a screendump :-/
16:17:15TheSevendamn, ftl dead again
16:17:45kugelUnhelpful: ?
16:18:02TheSeveni really need to work on a more stable ftl recovery method...
16:18:24Unhelpfulkugel: the bmp loader doesn't read from disk directly to the address at which the bmp data will be stored, it uses an on-stack read buffer.
16:18:32kugelwe clearly we need a pet name for the nano2g that involves stkov!
16:18:58kugelUnhelpful: ah got it
16:20:07Unhelpfulthis is done, and is not really a great loss, because from-file bmp data must go through various transformations before use.
16:21:04 Join ender` [0] (
16:22:46Unhelpfulso what you'd want to do is make the read buffer cacheline-aligned and disk-block-sized (i believe these are the requirements?)
16:22:55Unhelpfuland then you'll have even more stkov :D
16:28:00 Join TheSphinX^ [0] (
16:30:15TheSevensomebody wants to write an ftl recovery tool?
16:31:32pamauryWhat does it consists in ?
16:33:32TheSevenscanning the whole flash, generating a block map from the decentral set of control data (in the spare bits of the individual pages), trying to forcibly mount the ftl and check to which degree this is still possible, then comparing the central block map tables to the decentral data we gathered and fix the central table.
16:33:32 Quit ecio (Read error: Connection reset by peer)
16:33:47 Join ecio [0] (
16:34:30TheSevenwhat might be a bit more tricky is recovering from a state where it was interrupted in a middle of some control data write
16:35:27pamauryWow, I can't understand anything ;) Anyway, I haven't got a flash target around so that would be difficult to write :)
16:35:53TheSevenit an fsck-like task ;-)
16:36:37CIA-8New commit by 03stripwax (r24808): Unroll and optimise channel coupling for arm gives 0.6Mhz speed for vorbis on pp(ipod video)
16:36:52stripwaxoddly, unrolling it made it fractionally worse on coldfire
16:37:30pamauryI've never written an fsck-like thing ! And I would not be able to test it ;) Although I'm 200% sure I would do it right the first time ^^
16:39:28TheSevengevaerts: how would i try your starfield pebble btw?
16:39:50gevaertsTheSeven: %P|starfield| in the wps should do the trick
16:40:07*TheSeven has never even looked at a wps before
16:41:07gevaertsif you just want to try pebbles, you can make a wps with only that line
16:41:31gevaertsit won't be a very interesting wps then, but it should work :)
16:41:54Unhelpfuldid i miss the part where we had real work happening on visplugins?
16:42:13pixelma%P once was used for something else
16:43:48gevaertsUnhelpful: depends on what you call "real work" :)
16:44:36TheSeven"*PANIC* Stkov ¤¤"
16:44:52TheSevenwhen trying to move a file
16:45:06TheSeveni really like this...
16:45:18gevaertspixelma: I picked that because it appeared to be unused. The "pebble" concept isn't anywhere near ready for commit though (if it ever will be), so there's still some time to discuss this :)
16:45:20Unhelpfulgevaerts: so the "pebbles" are TSR? or are they just stopped/started when their viewport is displayed?
16:45:42Unhelpfulit seems the latter might be best since they could just be "regular" plugins then...
16:46:05Unhelpfulie, starfield would draw into a viewport when given a pointer to one on startup?
16:46:15pamauryTheSeven: how many stkov do you experience per day ? :)
16:46:20gevaertsUnhelpful: I think they're somewhere in between
16:46:24TheSevenusually none
16:46:41TheSevenbut today i have started digging in some areas where i haven't been before (and thus haven't fixed them yet)
16:47:18gevaertsUnhelpful: they have a function that's called on startup so they can initialise things, and they have a function that's called whenever they should redraw their viewport
16:47:19*TheSeven grumbles at gevaerts
16:47:30*gevaerts is innocent!
16:47:52*TheSeven blames gevaerts
16:48:02pamauryTheSeven: What did gevaerts do?
16:48:07TheSeventhis: + snprintf(filename,MAX_PATH,"/.rockbox/pebbles/%s.pebble",temp);
16:48:24gevaertsTheSeven: I've had data aborts and illegal instructions using that patch, but never a stack overflow!
16:48:58TheSevengevaerts: i haven't even managed to make a pebble start yet, those stkovs are all different issues on my way up to that point
16:50:04gevaertsUnhelpful: normal plugins can't work for visualisation as long as drawing isn't supposed to happen outside of the GUI thread
16:51:19gevaertsTheSeven: I'd advise to start with the "test_pebble" pebble, that one actually works on some targets
16:51:35*TheSeven first needs to fix that hardwired path anyways
16:51:36pixelmagevaerts: I just wondered why it seemed a familiar tag. It was used to preload a bitmaps for use as progressbar which is now unified to one %pb tag. There won't be problems with old WPS still using it because these will be broken due to the old %pb syntax anyway. The worst that could happen is someone gets confused who has to deal with a really old WPS
16:52:30amiconnHmm, bad news about the PP5020
16:53:18amiconnOf its 96KB of IRAM, only the first 24KB (one quarter) is single cycle, the rest has one waitstate on access
16:53:41gevaertsShould we keep the Ondio SP red? I don't think it's likely to shrink enough for rombox in the near future, and this way there's no current build for people to use
16:53:43amiconnLooks like they moved the PP5002's cache bug to IRAM in PP5020
16:54:21TheSevenhmm, "boot changed" detection also doesn't work with a nonstandard rockbox base dir
16:55:10 Join stooo [0] (
16:55:17*kugel has semi-working resizable skin buffer
16:55:27pamauryTheSeven: why do you need a nonstandard rockbox base dir ? You boot nano2g on sd ?
16:55:59TheSevenno, but i kinda have a "production build" and an "experimental (pebble) build" on it
16:56:49TheSevenbtw, the clock pebble works
16:56:56TheSeventhe starfield one locks up :-/
16:56:58kugelTheSeven: BOOTDIR is important, not ROCKBOX_DIR
16:57:53***Saving seen data "./dancer.seen"
16:59:19kugelnow I just need to make growing work, i.e. parse a skin and ask for reboot if it doesn't fit
17:01:46kugelparse all skins actually, that's the hard part :(
17:03:56 Quit piotrekm (Quit: piotrekm)
17:04:07gevaertskugel: you could of course just resize-on-fail, and then have people reboot seven times before it works
17:04:18Unhelpfulgevaerts: so a regular plugin could *be* a visplugin if its drawing code is isolated in a function, and it exports some extra header data?
17:04:38kugelgevaerts: yea, being bloody annoying is cool :)
17:04:49gevaertsUnhelpful: that might be doable. I'm not sure if it makes sense though
17:05:19kugelI need two passes, one to get the size and one to actually parse it
17:05:24Unhelpfulgevaerts: well, people already seem to like some visualization-type things as normal plugins. it seems to me it could make sense.
17:05:53Unhelpfulwhy exactly "must" all draws happen in the GUI thread?
17:05:53*kugel adds a 30k buffer which should be enough to do size+parse in the first run, and load images in the second
17:07:06 Quit mt (Ping timeout: 256 seconds)
17:09:30gevaertsUnhelpful: this needs thinking. One difference between visualisation plugins and regular plugins as they exist today is that most or all plugins assume a fixed screen==viewport size, which (in some cases at least) is a performance gain. Ideally we'd want at least some of those to allow for different viewports to work with the .sbs of course
17:09:32 Join froggyman [0] (
17:10:55gevaertsAnyway, yet again I don't feel like spending too much time on that sort of thing as long as I don't know if the entire concept of non-fixed-address plugins works
17:13:45CIA-8New commit by 03tomers (r24809): Brickmania: Use predefined macro definition when possible
17:13:49kugelthe filesize isn't a good estimate for the space an image needs inram, is it?
17:14:07 Join tomers [0] (
17:14:49kugela non-resized bmp image that is
17:16:27Unhelpfulkugel: not really, no.
17:16:52Unhelpfulbmp rows are padded in files, and bmp can be > or <16-bit.
17:17:01kugelso I really need to load them in the first pass for their size?
17:17:49kugelwell, accuracy doesn't matter too much
17:18:03kugelas long as they are never bigger than the file size
17:18:28kugelcould they be bigger?
17:18:46amiconnOf course
17:19:48kugelcan you explain that please? do you mean for instance when I load a mono bitmap on a color screen?
17:20:09amiconnBMP files can be 1/4/8/16/24/32bpp
17:20:28amiconnSo if you load a 4-bit BMP, it will become bigger in-memory on colour targets
17:21:15amiconnSame for 1-bit and 8-bit of course
17:21:45amiconnIf you're loading mono bitmaps, they will always become smaller, except on mono targets
17:22:29amiconnI mean mono bitmaps in rockbox format, and the exception is wrong
17:22:39 Quit CGL (Remote host closed the connection)
17:22:50kugelso, mono bitmaps don't get bigger?
17:23:01 Join CGL1 [0] (~CGL@
17:23:24kugeldo I really need to worry about 4/8. do we even support loading those? I've never seen them in reality I think
17:23:27amiconnThey can get bigger as well, due to pixel packing
17:23:31 Quit CGL1 (Remote host closed the connection)
17:23:37amiconnYes we do support them
17:24:44amiconnIf you load e.g. a 32x1 pixel 1-bit BMP as mono bitmap, it will become 8x on-disk size (ignoring the header) in memory
17:25:01 Join CGL1 [0] (~CGL@
17:25:18 Join iq [0] (~chatzilla@unaffiliated/iq)
17:25:49amiconnIn order to predict in-memory size you need at least the bmp header, and the destination format (mono or native)
17:27:40 Quit CGL1 (Client Quit)
17:27:59 Join stoffel [0] (
17:28:00kugelwould it be worth it reading only the header instead of full loading?
17:28:05 Join CGL [0] (~CGL@
17:28:24 Quit CGL (Read error: Connection reset by peer)
17:28:31 Join CGL [0] (~CGL@
17:34:29*kugel will worry about that later
17:35:42 Join Buschel [0] (
17:36:36 Quit iq (Quit: ChatZilla 0.9.86 [Firefox 3.5.5/20091102152451])
17:41:03 Join CaptainKewl [0] (
17:42:07 Quit tuma (Quit: leaving)
17:46:16 Join iq [0] (~iq@unaffiliated/iq)
17:49:22 Join funman [0] (~fun@rockbox/developer/funman)
17:49:49 Part stooo
17:56:10 Join ifonefox [0] (~irchon@
17:57:11 Quit ifonefox (Remote host closed the connection)
17:58:13 Quit funman (Quit: free(random());)
17:59:11 Quit dfkt (Read error: Connection reset by peer)
18:00:52 Quit GeekShadow (Quit: The cake is a lie !)
18:02:59TheSevengevaerts: around?
18:03:14TheSeveni fixed the starfield pebble :-)
18:03:18TheSevenor rather the core
18:03:52gevaertsWhat was wrong?
18:04:03TheSevenpebble.c, lines 104 to 108... do you see it?
18:04:51TheSevenpost-mortem memory dumps are very helpful to track down that kind of issues :-)
18:04:59gevaertsyes, probably
18:05:21TheSevenno, it will only patch every second GOT entry
18:05:21 Join mitk [0] (
18:05:36TheSevenyou're incrementing i both in the for statement and in the loop body :-P
18:05:46*gevaerts is an idiot
18:05:57gevaertsIt was a while loop at some point
18:06:09TheSevenyes, i figured that it must have been :-)
18:06:15amiconnPP5020 IRAM behaviour is even more weird
18:06:40amiconnThe first quarter (24KB) is zero-wait for the CPU, but one waitstate for the COP
18:06:57 Quit stripwax (Quit:
18:07:12amiconnFor the second 24KB it's the opposite - zero-wait for COP, one waitstate for the CPU
18:07:25amiconnThe rest (48KB) is one waitstate for both
18:08:14amiconnSomehow portalplayer has messed up the crossbar switch
18:09:01amiconnNow we just need to find a sane way to deal with this</sarcasm>
18:12:46gevaertsTheSeven: I guess that clears the last hurdle. These things are clearly doable, at least on ARM
18:13:23TheSevengevaerts: i looked at a disassembly of that pebble and almost fell off my chair
18:13:34gevaertsthat bad?
18:13:54TheSevenremove -fPIC, remove the GOT and the code patching it altogether, and i bet it will still work :-D
18:14:04TheSevenand faster / smaller / less code, of course
18:14:20TheSeventhe way they're calling functions is just crazy
18:14:25gevaertsthe code itself, yes, but the data section?
18:14:54TheSevenhm yes, the data might need to be relocated
18:15:20TheSevenbut why aren't they using sane branches, if they know the exact (relative) target address at compile time?
18:16:02amiconnThey don't
18:16:11CIA-8New commit by 03tomers (r24810): Brickmania: Hopefully prevent dissappearing of the ball
18:16:25gevaertshm, my clock pebble is broken apparently
18:16:31amiconnAt least not if the target is in another module (source file) or section
18:17:19pamaurytomers is on a brickmania fixing spree :)
18:17:32 Quit robin0800 (Remote host closed the connection)
18:17:41tomerspamaury: yeah :-)
18:17:48amiconnIf you put the restriction that pebbles must not use iram (but that probably already applies) you could compile without -mlong-calls
18:17:58amiconnMaybe this will improve code generation
18:18:32pamaurytomers: did you use another trick to prevent the ball from disappearing or did you eventually fix the intersection code ?
18:18:58tomerspamaury: It was not the intersection code that was buggy, it was the check of collisions
18:19:55tomersIn this commit I fixed a case where the ball hit the floor, but code flow continued to check whether it hit the left wall, then moving it a bit and modifying its speed
18:21:00tomersbut actually I'm not sure what caused it to hang in an infinite loop or alternatively to keep moving to ball to infinity ~ (-infinity, +infinity) actually
18:21:26tomersI just found one wrong thing and committed. This is why I wrote 'hopefully' :-)
18:22:37 Quit topik (Ping timeout: 260 seconds)
18:22:48*gevaerts decides to test on m68k
18:22:57tomersThere's still the bug of D2 not able to save game state well - and I don't think it's a bug in Brickmania, rather some bug with consequent multiple writes to a file descriptor...
18:24:52amiconnDoes the D2 use the SD card as primary filesystem?
18:25:02 Join stripwax [0] (
18:25:34tomersamiconn: yes, because the internal storage is read-only
18:25:37gevaertsamiconn: yes, if it boots from SD at least
18:26:33 Quit CGL (Remote host closed the connection)
18:28:45 Join CGL1 [0] (~CGL@
18:29:03 Quit CGL1 (Remote host closed the connection)
18:29:08 Join CGL1 [0] (~CGL@
18:29:15 Quit CGL1 (Remote host closed the connection)
18:29:56 Join CGL1 [0] (~CGL@
18:30:45 Quit CGL1 (Client Quit)
18:31:27 Join CGL [0] (~CGL@
18:32:31 Join toffe82 [0] (
18:32:46 Part mitk ("Ex-Chat")
18:33:53 Join funman [0] (
18:33:54 Quit funman (Changing host)
18:33:54 Join funman [0] (~fun@rockbox/developer/funman)
18:37:45TheSevengevaerts: and please fix that hardcoded path...
18:37:54TheSevenjust use snprintf(filename,MAX_PATH, ROCKBOX_DIR "/pebbles/%s.pebble",temp);
18:38:54 Quit iq (Remote host closed the connection)
18:44:17 Join anewuser [0] (anewuser@unaffiliated/anewuser)
18:44:24 Quit Farthen (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
18:45:43kugelhm, it almost works :/
18:46:27kugelimages are only a tad bit corrupted :(
18:49:54 Join panni_ [0] (
18:50:24kugelbackdrop, static images and volume bars seem to work
18:53:52*TheSeven would like to know what's wrong with the backdrop and some other graphics in that pebble build
18:57:27TheSevenjust lack of space...
18:57:41TheSevengevaerts, why did you shrink the skin buffer in the pebble patch?
18:57:46gevaertsdid I?
18:57:55***Saving seen data "./dancer.seen"
18:57:58TheSeven- (WPS_MAX_TOKENS * sizeof(struct wps_token))
18:58:00TheSeven+//#define SKIN_BUFFER_SIZE (MAIN_BUFFER + REMOTE_BUFFER) + \
18:58:01TheSeven+// (WPS_MAX_TOKENS * sizeof(struct wps_token))
18:58:03TheSeven+#define SKIN_BUFFER_SIZE 0x2000
18:58:06 Quit Strife89|Desktop (Quit: Booting into Windows to do maintenance.)
19:01:41 Join phanboy4 [0] (
19:03:41 Quit kugel (Read error: Operation timed out)
19:03:53Unhelpfulgevaerts: so basically what you're doing now is loading pebbles into the buffer at whatever address they can get, and assuming that all ARM code is relocatable?
19:04:46gevaertsUnhelpful: the code is compiled with -fPIC, which gives me a .got section that has all addresses I need to change in it
19:05:06gevaertsSo it's basically a very primitive relocation format
19:05:48Unhelpfulah. how messy is handling the relocs?
19:06:44gevaertsthree lines of code
19:07:02TheSeventrivial from the loader side - a huge mess for the compiler (and the code generated by it)
19:07:38TheSevenwhat about ELF / PE / whatever executables and a proper relocator?
19:08:25UnhelpfulTheSeven: elf apparently has an appalling number of reloc types for ARM :/
19:08:35gevaertsThe consensus seems to be that ELF is too complex, but that COFF might be doable
19:08:46TheSevencoff is rather trivial
19:08:59gevaertsHowever, nobody seems to be doing it, so I decided to go ahead to wake things up :)
19:09:01TheSeveni haven't seen anything but reloc type 3 being used there
19:10:10UnhelpfulTheSeven: how nasty *is* this code, exactly? so much ARM code is already PIC due to the use of relative addressing for jump, constant loads, etc.
19:10:40 Join Grahack [0] (
19:10:43TheSevenwell, the jumps *aren't* relative if you use -fPIC
19:11:03TheSeventhey load the destination address from the GOT instead of just using a relative jump -.-
19:12:58gevaertsMaybe -mno-gotplt helps? I'm not sure if I understand what it does
19:14:26 Join mt [0] (~mtee@rockbox/developer/mt)
19:15:12UnhelpfulTheSeven: ideally we'd like to see references within the plugin done by offset, with relocs only when offsets are out-of-range or to things in core (if we're doing relocs, relocs for direct jump into core would be nice-to-have. ;P)
19:16:47Unhelpfulalthough direct-bl for calls into core might not be a practical relocation scheme anyway...
19:16:54TheSevena total of 32 bytes (6 instr + 2 data) and two memory loads for something that could be a single BL instruction
19:18:04TheSeventhe core calls are still better than the internals calls in the pebble!
19:19:04UnhelpfulTheSeven: well, how would we produce a binary that has reloc data that we can use, without -fPIC?
19:19:35 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:19:53JdGordonAlexP: I wouldnt expect it to, whats the issue?
19:19:59JdGordonpixelma: arg, thats damn wierd
19:19:59TheSeveni don't undestand why the compiler is generating that crap in the first place. this simply isn't needed for anything, as long as the call is within the same module
19:20:08JdGordonkugel: why would that make it any harder to size?
19:20:21kugelbecause you need to know the end
19:20:33*DerPapst is annoyed by this feature introduced in r24696 (Clickable Sim Buttons). If the sim is active and i accidenty scroll with the mousewheel it recives some strange button presses (even with −−nobackground) and stops playback as a consequence. and after that the sim doesn't react to any button presses via recived the keyboard.
19:20:56kugelanyway, I have it almost working
19:21:20gevaertsTheSeven: have you tried without -mlong-calls?
19:22:06UnhelpfulTheSeven: per manual PIC code "accesses all constant addresses through a global offset table"... if all means *all* then apparently it is assuming that even things in the same module may move relative to each other?
19:22:43TheSevenUnhelpful: afaict the compiler emits the code in a way that inhibits this (only a single section)
19:23:10JdGordonkugel: ? the end is start+size? whats so hard about that?
19:23:21Unhelpfulwoo, found the stupid bug in this divider. :D
19:23:27kugelJdGordon: the size is unknown.. to write it in ARM :/
19:24:11JdGordonthen you are creating the buffer too early
19:24:28kugelJdGordon: you can't just use the entire audio buffer and then cap it after skin parsing
19:24:32*Unhelpful ponders doing evil things with the carry flag...
19:24:51kugelbecause giving the whole audio buffer means it will put stuff at the start and at the end of the audio buffer
19:25:32TheSevenwhy does one even need a GOT?
19:25:46TheSevenwhy can't they just pc-relative values instead?
19:25:56gevaertsTheSeven: -funit-at-a-time makes a big size difference (128 for starfield.pebble). Can you have a look at the code?
19:26:14JdGordonwhat happened to storeing the them size in the .cfg and if that size is bigger than the current theme require reboot? and if its unkown just use a really large buffer?
19:26:18TheSevengevaerts: where in that weird make file do i place that best?
19:26:23UnhelpfulTheSeven: those are range-limited. reasonably for jumps, but more tightly for loads.
19:26:40gevaertsTheSeven: at the end of PEBBLEFLAGS in apps/pebbles/pebbles.make
19:26:44TheSevenUnhelpful: ldr Rx, =difference, add Rx, Rx, PC
19:26:46kugelJdGordon: that's all much harder than just grabbing the entire audio buffer and cap it
19:26:52TheSevenstill a lot faster than the GOT crap
19:27:13JdGordonkugel: but then you have to stop playback to reload a theme?
19:27:19JdGordonimo thats the wrong way to do it
19:27:30kugelwhat's the wrong way?
19:27:55JdGordonalso, what happened to figuring out the required size during parse time and only continue loading if it fits?
19:28:04UnhelpfulTheSeven: faster still would be not using PIC... gcc would then emit code that loads the values from nearby constant pools, and as long as we retain reloc info we could fix it up.
19:28:04kugelthat's what I do right now
19:29:13JdGordonso just change it to default to a large value if its unknown, or the .cfg value + extra for known and you're done
19:29:26kugelthat's what I do right now
19:29:29TheSevenUnhelpful: the constant pools are close enough to be loaded via LDR Rx, [PC,#whatever]
19:29:42kugelI'm just saying it's harder than it needs to be
19:30:18JdGordonits exactly as hard as it should be
19:30:18kugelimage loading is broken I think. it should fail to parse if an image can't be loaded
19:30:34JdGordonusing the whole buffer means having to stop playback which is lame
19:30:46kugelyou don't understand me
19:31:14kugelusing the whole buffer at boot, then load skins into it. for changing skins afterwards use the existing one (and ask for reboot if needed)
19:31:23kugelthat would be simple and nice
19:31:25TheSevenbtw, why isn't the linker cabable of inserting thunks for out-of-range relative branches?
19:31:36TheSeventhen we could disable that long-call mess altogether
19:31:46kugelTheSeven: ...eabi is doing that :)
19:31:56 Quit Uretti (Read error: Connection reset by peer)
19:32:10TheSeveni'm using eabi, and it doesn't seem to do it
19:32:22kugelhave you removed -mlong-calls?
19:33:07TheSevennot manually, but why isn't that done on eabi by default?
19:33:29kugelit should. but if it's doing long calls it's buggy :)
19:33:46gevaertsTheSeven: -mlong-calls is enabled by CFLAGS I think
19:34:43kugelprint GCCOPTS to be sure
19:35:28kugelTheSeven: weren't you able to remap memory? you don't need long calls at all then
19:35:50 Quit stripwax (Quit:
19:35:54TheSevenno, the 940t doesn't have an mmu
19:36:13kugelbut wasn't there a trick?
19:36:17gevaertsthat also wouldn't help for all targets
19:37:02kugelconfigure should only add -mlong-calls if it's no eabi build but it appears to be broken if you still get them
19:37:07TheSevenkugel: we had a trick to do uncached reads/writes, but not for remapping
19:37:18 Quit anewuser (Quit: =ooo ϢINTER ϾHIP 5iVE is OOON!! omg 7 days left =oo)
19:37:27TheSevenyes, i can't find where it sets that flag, but it is set in my makefile
19:38:37*JdGordon accuses AlexP of sabotaging his weekend!
19:38:50JdGordonmultifonts on the remote worked when it was commited!
19:39:03kugelconfigure, line 263+
19:40:40kugelJdGordon: image loading is broken, parsing should fail if it can't be loaded
19:41:22kugelload_skin_bitmaps always returns true
19:41:35CIA-8New commit by 03theseven (r24811): Fix configure to not use -mlong-calls when configuring for EABI
19:42:01JdGordonkugel: so fix it?
19:42:14TheSevenwow, that has been a stupid bug
19:42:23TheSevenweird that nobody had noticed this before
19:42:45*kugel hides
19:42:56 Join iq [0] (~iq@unaffiliated/iq)
19:42:57*TheSeven wonders how much his core binary will shrink now
19:43:04kugelabout 50k
19:43:28*TheSeven was already a little un-impressed by seeing a 2K binsize reduction when enabling eabi
19:48:28 Join karashata [0] (
19:49:51kugelJdGordon: %X|-| is valid?
19:50:01JdGordonshouldnt be
19:50:05JdGordonnot anymore
19:50:10JdGordondid that get left in somewhere?
19:50:15 Join Blue_Dude [0] (
19:50:30JdGordon- is set as default, but that line shouldnt be valid
19:50:40 Join Frakenstein [0] (
19:50:49kugelI'm a bit confused by "if (backdrop[0] == '-')" in skin_backdrop_load
19:50:57kugelah ok
19:51:43TheSevenkugel: yes, 56K binsize decrease
19:55:21Blue_Dudekugel: Thanks for the playlist viewer screen fix yesterday.
19:56:42UnhelpfulTheSeven: funny-i-could-swear-i-did-that-ages-ago. :/
19:57:01Blue_DudeBTW, still looking for feedback on making a new delete hotkey for the tree view. Some discussion a day or two on that. It's something I'd use a lot but I'm not 100% sure how much anyone else would get from it.
19:57:09TheSevenUnhelpful: yes, you did it, but it was broken
19:57:29TheSevenlook at the diff i just commited ;-)
19:57:30UnhelpfulTheSeven: thanks for fixing then. :)
19:57:58 Quit stoffel (Remote host closed the connection)
19:58:13*Unhelpful wonders how tired he must've been when he wrote that
19:58:40TheSeventhe nasty thing about that is that it just fails silently
19:59:21JdGordonBlue_Dude: one click delete?
19:59:59UnhelpfulTheSeven: test pretty much always does ;)
20:00:34*gevaerts is confused
20:00:38Blue_DudeJdGordon: more like one click to confirm-delete.
20:03:52kugelperl needs { } for single line ifs?
20:04:21kugelBlue_Dude: I wouldn't use it a lot and would rather see insert on the rec button
20:04:46kugelBlue_Dude: I thought I was the only one annoyed by that :)
20:04:51JdGordonkugel: perl needs {} for ALL ifs
20:05:45Unhelpfulkugel: yah, it's one of its more annoying deviations from "just copy C's syntax" that for/if/while require braces.
20:06:27 Join darkham [0] (
20:07:15funmanI have put my hands on the patches to linux kernel which add as352x and as353x targets
20:07:53JdGordondoes this mean I shold go buy another clip+?
20:07:54funmanthey both add a bunch of new files, some of them have the gplv2 licence header, some others have a more restrictive header which prohibits reproduction without the written consent of AMS
20:08:28funmanwell the MMC/SD driver for as353x is clearly suited for clip+, but i must check the licence of this file
20:09:26funmanit's gplv2, so everything is fine \o/
20:09:51 Join kimi-sharamin [0] (
20:10:11CIA-8New commit by 03jdgordon (r24812): remote multifont fixes. it shoud be working again (blame AlexP! he made me break it last week!)
20:10:31Unhelpfulfunman: that seems too easy ;)
20:10:45 Quit karashata (Ping timeout: 268 seconds)
20:11:00CIA-8New commit by 03kugel (r24813): Fix wpsbuild to properly generate "-" for theme related settings (to get the default values/no file for filename settings).
20:11:08funmani think sharing the whole patch file with devs is easier than removing the bits which say "reproduction prohibited", but then i must ask the devs to not look at these parts until we sort out if we can use them or not
20:11:42JdGordondo we still have the ams contact?
20:11:54funmani have the mail of that marketing person
20:11:56BagderJdGordon: unfortunately not
20:12:07 Join saratoga [0] (~463f90ed@gateway/web/freenode/x-hszyhocvfybjjxkx)
20:12:07Bagderthey seem to have abandoned us
20:12:27funmanthis guy had sent me the as353x datasheet (after several mails/delay though)
20:12:42Bagderaha, that's good
20:13:20Blue_Dudekugel: I wonder how easy it would be to have a user-assignable hotkey. Insert or delete (or anything else in a context menu) depending on user preference?
20:13:27kugeldoes anyone know why read_bmp_file multiplies error return values with fd?
20:13:53JdGordonBlue_Dude: thats the way I'd like to see it done, but apparently it falls into "no configurable keys/menus" which means never going to happen
20:14:18*JdGordon stares daggers at the people who know who they are about this
20:14:34Blue_DudeWhat does this: "no configurable keys/menus" mean?
20:14:37kugelI thought we changed our minds about it?
20:14:51Blue_DudeWhat's a quickscreen then?
20:15:08funmannote, those files which have the restrictive header also have a "MODULE_LICENSE("GPL"); line at the bottom
20:15:37kugelso they cheat?
20:16:03funmanI checked the patches sent by Thomas Luo 4 years ago to linux-arm-kernel mailing list for as352x
20:16:16funmansome were exactly the files I have, but with a correct GPLv2 header
20:16:36funmanI would think it's probably an old habit of using that kind of headers, on the advice of their lawyers
20:16:43saratogaif they say GPL anywhere on them I would assume we can at least rewrite them for rockbox, if not use them verbatium
20:16:57gevaertsWe can't use them verbatim
20:17:01kugelUnhelpful: do you know?
20:17:05iqHi, I am compiling for Clip+. I am able to build the bootloader and simulator but can not compile the RockBox - getting error:
20:17:20saratogai don't think rockbox builds for the clip+
20:17:22funmaniq: rockbox doesn't build for clip+ yet, since the bootloader isn't finished
20:18:19iqok - I guess I'll have to wait then.
20:18:26saratogaso how hard will it be to get an SD driver using these sources?
20:18:44funmani would expect quite easy
20:19:05saratogado they cover other parts of the chip like clocks and power management?
20:19:21funmani guess everything is in here
20:19:33funmanusb, i2c, sound, clocks ...
20:20:17Unhelpfulkugel: do i know?
20:20:23kugel"does anyone know why read_bmp_file multiplies error return values with fd?"
20:20:42funmansince it's not clear to me how i can redistribute it, i'll send it to interested devs if they don't put it online, at least until we sort how that restrictive header works together with gpl
20:21:01saratogayes that seems reasonable
20:21:18saratogaif they didn't put it out in the open we probably shouldn't redistribute it anyway
20:21:24saratogawell publically redistribute
20:21:33Unhelpfulkugel: i'm not looking at it right now, but doesn't it do something like -(fd *10 + const)?
20:21:35saratogaif only so they help us when we ask about the Clip++
20:21:42funmanparts of as352x at least were sent to a public ML so at least those are ok
20:21:56 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
20:21:57 Quit pixelma (Disconnected by services)
20:22:11saratogamight as well just treat it like the other AMS docs for now
20:22:13 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
20:22:13 Quit amiconn (Disconnected by services)
20:22:15 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
20:22:26Unhelpfuli think the idea is to return both some constant indicating where the error happened, and also the fd.
20:22:39 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
20:24:50kugelso I can see it by the constant?
20:24:54kugelit=the error
20:25:13kugelif (error % 10) ... ?
20:25:23 Join Lear [0] (chatzilla@rockbox/developer/lear)
20:25:32funmani'll start deciphering the linux code and fill the missing holes in the Clipv2/Clip+ SD driver, unless someone else wants to do it
20:29:34CIA-8New commit by 03kugel (r24814): Fix the mechanism to fail to parse skins if images fail to load and fix having no backdrop at all (neither %X nor backdrop setting, DancePuffDuo was ...
20:30:10kugelfunman: I can't imagine anyone would want to (except FlynDice maybe) :)
20:31:18 Join liar [0] (
20:31:46funmanah: there was some errors in reverse engineering :/
20:33:15 Join stooo [0] (
20:39:29kugelJdGordon: I think I found the reason for the increased boot time, the sbs is parsed multiple times even if the theme has none
20:40:27JdGordonit should be parsed max once
20:41:03JdGordonskin_data_load() will happen max twice, the first will be an instant recall
20:42:47 Quit mc2739 (Ping timeout: 252 seconds)
20:43:09kugeldoes the wps need to be parsed before the rsbs?
20:43:23*gevaerts finally finds what he was doing wrong
20:43:30kugelor could I swap them (and put the skin loading within FOR_NB_SCREENS())
20:44:22JdGordonthe sbs MUST be parsed first for each screen
20:44:29JdGordonotherwise order doesnt matter
20:44:39 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
20:45:26JdGordonhmm, I meant the screens sbs must be parsed before any other skins on that screen
20:45:31JdGordonso yes, you could but it in a loop
20:46:21saratogawhat does the linux source refer to the as3525v2 as ?
20:47:06CIA-8New commit by 03Buschel (r24815): Optimization for cook codec. Rework sample output to be able to use highly optimized dsp routines. Moved some functions to iram. Speeds up codec by ...
20:47:17TheSevenbtw, what's holding back the switch to eabi? is it still buggy on some targets?
20:47:45 Join stoffel [0] (
20:48:19funmansaratoga: not sure if there is linux code for this one
20:48:47saratogaso its just the as3525 and the as353x?
20:48:58funmani believe AMS just updated the as3525 with some components present in the as353x (SDMMC, VIC, as3514)
20:49:11funmanas352x, there is also as3527 code
20:49:17funmando you want them ?
20:50:10TheSevengevaerts: on eabi without -mlong-calls the pebble code is sane
20:51:15funmanbtw original submission of (parts of) as352x linux patches are on (without follow up to changes/informations requested by the arm linux developers)
20:52:30gevaertsTheSeven: nice!
20:52:51gevaertsIn the meantime, I've (finally!) managed to put the rb pointer back automatically
20:53:22saratogafunman: sure
20:53:43moosJdGordon kugel: maybe an opportunity to look at disabling the copy of rsbs for tagets that doesn't have one
20:54:09 Quit darkham (Quit: Sto andando via)
20:55:29moosI meant that .rsb files are still there for tagets that don't have remote, sure that need an easy fix
20:56:03kugelif (setting[0] && setting[0] != 0xff && setting[0] != '-') #-> "warning: comparison is always true due to limited range of data type"
20:56:28JdGordonwtf is the check for 0xff?
20:56:42JdGordonthats there on a few settings but I dont think is still needed
20:57:16kugelno idea, it's there since I looked at it for the first time
20:57:19 Quit iq (Quit: Leaving)
20:57:29saratogafunman: if you have a link handy, would you gtalk it to me? i have to head out
20:57:43saratogai'll probably be useless for this port but i'd still like to look it over
20:57:44mooskugel: nice that you noticed a slowdown on boot, since JdGordon doesn't/don't want :p
20:57:48funmani'll send it by mail
20:57:55JdGordonkugel: I'm pretty sure we can remove that.. I tihnk its from when the config was on disk
20:57:58***Saving seen data "./dancer.seen"
20:58:17 Join stripwax [0] (
20:58:49 Join ansuz [0] (
20:58:56FlynDicefunman: no time right now but I see your comments above with great interest ;-)
20:59:12funmanFlynDice: you're a recipient of the mail as well ;)
20:59:39FlynDicefunman: You Da man!!
20:59:47 Quit stooo (Quit: Leaving.)
20:59:53kugelI'm not :(
20:59:56saratogamight as well add kugel, mt
21:00:06 Quit saratoga (Quit: Page closed)
21:00:08funmankugel: you were already ;)
21:00:54JdGordonBagder: oh, can you add uchida to the devs group on flysrpay?
21:03:31 Quit BlakeJohnson86 (Quit: Leaving.)
21:06:41 Join iq [0] (~iq@unaffiliated/iq)
21:07:40 Join BlakeJohnson86 [0] (~bjohnson@2002:1876:a27b:0:227:13ff:fe65:1262)
21:08:02*gevaerts enjoys watching his starfield wps :)
21:08:45kugelfunman: the mail came later. thanks!
21:10:07 Join perfectdrug [0] (
21:10:22funmaniq: did you run the rockbox bootloader on your clip+ ?
21:12:14 Quit BlakeJohnson86 (Remote host closed the connection)
21:12:47gevaertspebbles don't work on m68k :\
21:13:03gevaertsor at least, they need more effort
21:14:36ThomasAHfunman: were you be able to decode the hex numbers (csd?) now?
21:19:07iqfunman, no, I have not. I thought it is incomplete?
21:20:27funmanThomasAH: yes, my app was buggy, they describe a 100% normal SD card
21:20:57ThomasAHfunman: so SD vs. SDHC?
21:20:58funmaniq: yes it's incomplete, and there have been 2 clip+ bricked already, so I was wondering if you had tried it successfully
21:21:11funmanThomasAH: 2GB is SD and 4/8GB are SDHC
21:21:51ThomasAHfunman: so as expected, right?
21:22:01iqfunman, we can't unbrick a Clip+, right?
21:22:10 Join BlakeJohnson86 [0] (~bjohnson@2002:1876:a27b:0:227:13ff:fe65:1262)
21:22:31funmaniq: i don't think we can, perhaps if you manage to solder a jtag adaptor, but definitely not in a simple way
21:23:28ThomasAHiq: my clip+ 8gb survived the bootloader ... so I think it is 2 bricked vs. 3 non-bricked now
21:23:57funmani think it's even 4 non-bricked
21:24:05iqThomasAH, what procedure you followed to load it ?
21:24:28funmaniq: i think it's detailed on SansaAMS wiki page (mkamsboot)
21:25:37iqfunman, so everyone followed the same procedure and we have a little over 50% success rate.
21:25:38ThomasAHiq: I built the boot loader with the old arm compiler (not -eabi or how it is called), built mkamsboot, then ./mkamsboot clppa-original.bin bootloader-clipplus.sansa clppa-patched.bin
21:25:40TheSevenis it just bricking devices randomly, or did you manage to track down what had happened?
21:25:40 Quit stoffel (Remote host closed the connection)
21:26:08funmaniq: yes, although the code in mkamsboot changed, and the current code worked 100% of the times it was tried
21:26:09ThomasAHiq: then copied clppa-patched.bin as clppa.bin to the internal memory root dir
21:26:10 Quit Grahack (Quit: Tu m'as vu ?)
21:26:23 Nick kimi-sharamin is now known as karashata (
21:26:24funmanTheSeven: we don't know what happened
21:26:39 Quit karashata (Quit: The fluffy dragon has left completely!)
21:26:39TheSevenand they just don't power on any more, or what do they do?
21:26:46funmanfor JdGordon's clip, the code might have used r5 or r3 but i don't remember
21:26:46TheSevenno DFU on those things?
21:26:57funmanafaik no
21:27:04TheSevenwhere is the bootloader stored? nor? nand?
21:27:18TheSevenis there an internal bootrom or is it booting directly from nand?
21:27:27TheSeven(via steppingstone)
21:27:29funmanon e200v2 you can access the first sectors of the SD card where it's stored, but only on this one
21:27:30*ThomasAH was able to flash two versions of the boot loader without problems and back to plain OF
21:28:29funmanTheSeven: i don't know how it works in the details, only that the OF is stored on the first sectors of the SD (i believe it's loaded to memory by the ROM bootloader)
21:28:56TheSevenso they're using those embedded microsd chips?
21:29:07 Join Farthen [0] (
21:29:17funmanthere is a nand chip which is read/written by a MMC/SD controller
21:29:40funmanprobably to avoid wear leveling in software
21:29:54funmanand since they are sandisk they don't have to pay royalties for the controller
21:29:59JdGordonI'm happy to post my clip to anyone for an autopsy
21:29:59TheSevenheh, hardware ftl. funny. why don't they integrate that into one chip, like those microsd things do?
21:30:26funmanTheSeven: aren't the microsd just a nand chip + a controller?
21:30:35TheSevenyes, but in one package
21:31:13 Join Kitr88 [0] (
21:31:28funmanon e200v2 you can access the OF as a usb mass storage if you connect 2 pads together
21:31:44JdGordonI tried that and didnt have any lucjk
21:31:51iqThomasAH, where you got clppa-patched.bin ?
21:31:56funmanI believe you can achieve the same thing by connecting 2 pins of the nand chip (I read this method on
21:32:12funmanif i do that on my fuzev1 or clivp1 i have an usb mass storage but with 0 sectors
21:32:19ThomasAHiq: this is generated by mkamsboot
21:32:49ThomasAH(the third argument is the target file name)
21:33:02TheSevenfunman: in theory yes, but the sd controller will probably not like it
21:33:40TheSevenif anything, i would jam the communication between the CPU and the SD controller
21:33:46funmanif this method (shorting directly the nand pins) doesn't work on e200v2 then it could mean we didn't find the correct pads on other models
21:34:03funmanperhaps the OF is stored on the nand, and also somewhere else
21:34:19TheSevenfunman: all of this stuff will only help if the bootrom has a recovery mode
21:34:30CIA-8New commit by 03kugel (r24816): Properly generate (with "Save Theme Settings") and handle filename settings with "-".
21:34:34CIA-8New commit by 03kugel (r24817): Shorten settings_skins_apply() and correct some checks.
21:34:34TheSevencould it be possible to tap all the sd signals?
21:34:36CIA-8New commit by 03kugel (r24818): Don't be too wasteful and only reparse skins when actually needed.
21:34:46 Quit Kitar|st (Ping timeout: 246 seconds)
21:35:10funmanthe boot procedure is detailed in the as3525 datasheet, you want to have a look?
21:35:44 Quit Kitr88 (Ping timeout: 260 seconds)
21:35:53funmanpage 22 7.1.6 boot concept
21:37:29funmanXPC[3:1] is all 0 on my clipv1, that would mean boot device is "SPI master ST M25Pxx serial Nor Flash"
21:38:07funmane200v2 probably trigger the "USB boot promer"
21:38:37funman"if there is no bootable device" is probably the case when the NAND can't be read
21:39:19TheSevenwell, the jtag of that thing is even documented!
21:40:37 Join Kitar|st [0] (
21:42:13 Quit CaptainKewl (Quit: ( :: NoNameScript 4.22 :: ))
21:43:05TheSevenfunman: are there any pcb scans of clip+?
21:43:36funmanthere is ABI pictures but i could scan mine if you need bigger resolution
21:44:07funman and 13.jpg
21:44:38funmanoh, sorry I misread, it's only clipv1
21:44:56funmani haven't seen a clip+ opened
21:45:27ThomasAHI think I have somewhere on ABI
21:45:47 Join saratoga [0] (~9803c6dd@gateway/web/freenode/x-nclbbcioflrutssu)
21:46:15*ThomasAH checks his notes
21:46:35saratogaheh the patches show CCU_GPIOA through CCU_GPIOP
21:46:39saratogai wonder how many are on the clip
21:46:41iqThomasAH, clppa-original.bin is the clppa.bin from ?
21:46:48funmansaratoga: for as353x ?
21:47:00 Join petur [0] (
21:47:00 Quit petur (Changing host)
21:47:00 Join petur [0] (~peter@rockbox/developer/petur)
21:47:07funmani only saw GPIOA through D
21:47:09funmaniq: yes
21:47:46iqBTW, I checked out the code this morning (CST GMT-6) - so hopefully I will have the good bootloader, right :) ?
21:48:19ThomasAHiq: yes, manually renamed to avoid mistakes :)
21:48:19funmanthe bootloader isn't important at this time, mkamsboot is the part which can brick. And if you checked it out today it's the code which worked on my Clip+
21:49:10iqOkay, so now I have "clppa-patched.bin"
21:49:44funmaniq: be sure you are ready to risk bricking your Clip+ before copying it!
21:50:06iqfunman, so you won't send me a new Clip+ :) ?
21:50:52iqit is $40... so if it will help in the testing then i can risk.
21:50:54funmanno, only my gratitude :)
21:51:31iqNow i need to figure out how to access the internal memory.
21:52:02ThomasAHTheSeven: did you notice the URL?
21:52:05funmanyou need to set USB mode to MSC first i think
21:52:18saratogain the CGU code: "0 /* removed in V2 */ "
21:52:20iqThomasAH, Hopefully I won't be needing this - LOL
21:52:25funman < i wonder if the small chip is the FM
21:52:42funmansaratoga: they removed a device ?
21:54:12saratogathey removed some registers apparently
21:54:20iqI have changed the USB mode. But I still see my Music and other folders when I connect it to my machine.
21:54:34funmaniq: then just copy clppa.bin there and unplug
21:55:09ThomasAHiq: (next to music, not into :))
21:55:18iqmv clppa-patched.bin clppa.bin
21:55:27ThomasAHiq: noooooooo! :)
21:55:34*ThomasAH suggests cp
21:55:44funmanit looks like the controller supports several cards, perhaps there isn't 2 separate controllers for internal storage and µSD slot like on as3525
21:55:56ThomasAH(for forensic if it fails)
21:56:04iqThomasAH, start over ?
21:56:06AlexPI'm not sure I accept responsibility, but ta for fixing :) I'll try it out shortly
21:56:21saratogathere are registers for setting the SD card communication voltages
21:56:25saratogadidn't we have an issue with that
21:56:27ThomasAHiq: no, just cp back to your PC
21:57:19iqThomasAH, it is 15MB
21:57:20funmansaratoga: iirc it was the arm core voltage
21:58:10iqThomasAH, Copied, now unplug from the machine and power cycle ?
21:58:26iqof course unmount first
21:58:28funmaniq: no need to power cycle, update will begin immediately
21:58:44 Quit TopyMobile__ (Quit: TopyMobile__)
21:58:48iqhere goes my $40
21:58:49ThomasAHiq: umount, unplug, wait :)
21:59:05iqFirmware upgrade in progress
21:59:07 Join TopyMobile [0] (
21:59:14iqUpdate completed
21:59:22iqBlack screen :)
21:59:31ThomasAHiq: turn it on
21:59:45iqit says RockBox and dmg ->
21:59:59iqdma ->
22:00:05ThomasAHiq: dma ... though I saw the same :)
22:00:16iqso now what LOL
22:00:47iqseems like it worked okay, right ?
22:00:58ThomasAHiq: ok, to boot into OF again, turn off by holding the power button for >10sec, then turn on with Home button pressed
22:01:06ThomasAHiq: yes
22:01:11funmanhome or left button, any of the 2 should work
22:01:21ThomasAHso 2 to 4 now :)
22:01:27funmanto 5 !
22:01:33ThomasAHfunman: 5?
22:01:41iqNow I am back to the sansa firmware :)
22:01:54funmanyou, iq, FlynDice, me, and someone else who was present on irc
22:02:02saratogacomments look great, theres all sorts of stuff about power use, recommended clock dividers, etc
22:02:15Schmogel<- was number 4
22:02:22iqfunman, ThomasAH, thanks - it was fun ... what next :)
22:02:28ThomasAHfunman: ah, how could I forget someone else :)
22:02:37funmaniq: copy the original clppa.bin to remove the bootloader
22:03:10iqfunman, Is it optional ?
22:03:11ThomasAHiq: first show it around to impress people :)
22:03:23iqThomasAH, yeah :)
22:04:50 Quit cjcopi (Ping timeout: 268 seconds)
22:08:00 Quit Lear (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
22:08:08iqNow I need to write a Hello World
22:09:39 Join Lear [0] (chatzilla@rockbox/developer/lear)
22:10:17ThomasAHiq: adjust :)
22:11:05LearJust enabled dircache on my e200 after having it disabled for a while. Now I get stack overflow on boot... Anyone else seen that?
22:11:14LearUsing gcc 4.4.3 btw.
22:12:01TheSevenLear: what is the maximum directory depth you have?
22:12:23 Join cjcopi [0] (
22:12:57 Quit TheSphinX^ (Quit: XChat@Linux)
22:13:09LearWell, I do have a test folder for a total depth of 12. :) Otherwise not much.
22:13:53TheSevenup to ~20 should be supported with that stack size
22:14:01iqThomasAH, just apply this patch ?
22:14:29 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.5.8/20100202165920])
22:14:43CIA-8New commit by 03stripwax (r24819): Get a few more % speedup on ARM (measured on ipod video) - improve imdct full final symmetries using ldm/stm and simple register swapping. Also, add ...
22:14:47ThomasAHiq: no, look at it to see where printf("Hello World!"); would fit :)
22:15:06iqThomasAH, oh okay :)
22:16:16AlexPJdGordon: If you have 2 minutes could you do a quick sync of the fms patch?
22:17:55CIA-8New commit by 03stripwax (r24820): Eep, didn't mean to commit those other files
22:18:07CIA-8New commit by 03Buschel (r24821): Further work on cook codec. Rounding is not needed when using a large fract part in the internal sample representation. Move quantization array to ...
22:18:44saratoga"The ratio between bus clock and mpmc clock can either be 1:1 or 1:2. However + * this value cannot be changed dynamically because this would cause MPMC failure."
22:19:02saratogawell that would explain why we can't seem to switch memory clock dividers
22:19:20 Quit tomers (Ping timeout: 252 seconds)
22:20:13 Join bertrik [0] (
22:20:25saratogaso i guess if we want to clock the DRAM higher then 64MHz on AMS we need to do it early on before we've actually loaded anything important into DRAM
22:20:33saratogais the bootloader in DRAM or IRAM?
22:21:47stripwaxIs anything still using use tremor mdct_backward() ?
22:22:19funmansaratoga: dram for clipv2 (and fuzev2 i guess), iram for others
22:22:31saratogastripwax: no it can go
22:22:46saratogafunman: ok then it should be possible to increase the clock
22:22:53 Quit knine (Ping timeout: 252 seconds)
22:23:14funmansaratoga: before memory_init() would be the right place
22:23:37funmansdram_init()*, memory_init() isn't in the bootloader
22:24:55S_a_i_n_tAnyone feel like doing a quick check of a WPS to see if they get the same issues with sublines that I'm getting?
22:26:27 Join stooo [0] (
22:26:34S_a_i_n_t...apply hold, let it play for 1 min or so, take hold off. Sublines *should* go crazy.
22:26:43S_a_i_n_t*unless I've gone insane.
22:30:36 Quit petur (Quit: router reboot)
22:30:45LearTheSeven: I upped the stack to DEFAULT_STACK_SIZE + 0x400, debug menu says 80% usage after boot.
22:32:17 Join petur [0] (
22:32:17 Quit petur (Changing host)
22:32:17 Join petur [0] (~peter@rockbox/developer/petur)
22:35:30TheSevenS_a_i_n_t: heh, that's indeed waird
22:35:31mooswow the build table is full of colors! :)
22:35:32*stripwax is going to fix red just as soon as he figures out what happened with his last commit
22:36:35Buschelstripwax: shall I commit the fix?
22:36:52 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
22:37:50TheSevenkugel: did you notice the yellow btw?
22:38:40 Quit Omlet (Quit: ( :: NoNameScript 4.22 :: ))
22:38:50S_a_i_n_tTheSeven: thanks so much! I can add it to the tracker now :p
22:39:01 Join sudoman [0] (
22:39:21S_a_i_n_tI'm surprised noone's noticed it prior to now...
22:39:24TheSeveni don't quite get what's happening, but something is weird, yes
22:39:40CIA-8New commit by 03Buschel (r24822): Make mdct compilable again for non-ARM targets.
22:39:46 Quit stooo (Ping timeout: 265 seconds)
22:39:49TheSevenscrolling text going mad isn't really something new, saw that in some other places, too
22:40:08stripwaxBuschel - does that fix the ARM targets? sorry, mine is still (re)building
22:40:12kugelTheSeven: no, thanks. will have a look
22:40:30S_a_i_n_tdespite the viewport not being 'displayed''s still 'running in the background'? that's my guess...
22:40:40S_a_i_n_tallbeit wrong ;p
22:41:00Buschelstripwax_: it fixes the build for non-ARM (including PC-Sim)
22:41:20Buschelstripwax: you have deleted a brace too much
22:41:47 Join CaptainKewl [0] (
22:41:51stripwaxS_a_i_n_t - perhaps it just records a timestamp for when it last updated. so when the viewport is enabled, it has an old timestamp, and it (rapidly) catches up but along the way updating the text?
22:41:52 Join anewuser [0] (anewuser@unaffiliated/anewuser)
22:42:09S_a_i_n_tTheSeven: it only happens if the VP is conditional...otherwise, its fine.
22:42:16Buschelstripwax_: btw, CLIP_TO_15 is not used by any codec anymre
22:43:17S_a_i_n_tstripwax: that seems to fit, have you seen it happen? it alternates at pprox 0.1sec intervals.
22:44:19stripwaxBuschel - ah, I'd forgotten that clean builds don't show up in the compact build table (was freaking out and thought everything had broken!). Thanks for the fix, yep, you're right. Feel free to remove CLIP_TO_15 too (arm and cf, I guess)
22:44:46stripwaxS_a_i_n_t - I haven't, but I think something similar was discussed a while back, possibly in Bugs tracker?
22:45:43TheSevenS_a_i_n_t: it switches after scrolling *twice* for me
22:46:05S_a_i_n_ti'd looked, but couldn't find it if it is there...i had a forum thread on the topic going for a while, but it died.
22:46:53S_a_i_n_tTheSeven: it's at 10 second intervals...2mins on hold=12 switches.
22:47:32S_a_i_n_twell, for me anyway.
22:48:49CIA-8New commit by 03gevaerts (r24823): Disable rombox on ondio SP. It can be reenabled if the binsize goes down again.
22:49:14S_a_i_n_tthats a big *if*
22:51:20CIA-8New commit by 03gevaerts (r24824): remove incorrect cast
22:52:07 Quit tmzt (Ping timeout: 246 seconds)
22:52:13 Quit linuxguy3 (Read error: Operation timed out)
22:52:35 Quit Lear (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
22:52:59*amiconn slaps kugel with his 13 yellow points
22:53:13kugelamiconn: I was notified already
22:53:23gevaertsI just committed the un-yellower
22:53:30 Join tmzt [0] (
22:53:44pamauryTheSeven: yes ?
22:53:59TheSevensomething's still wrong with the dircache
22:54:09TheSeven12 levels => stkov
22:54:20pamauryhum, stkov ?
22:54:25 Join linuxguy3 [0] (
22:54:34kugelgevaerts: heh, I think it's incorrect
22:54:34TheSevensee the lines above my ping
22:54:43kugelbecause I had warnings on my machine without this cast
22:55:00 Quit iq (Remote host closed the connection)
22:55:17gevaertskugel: hm, wait
22:55:53kugelstatusbar_position is an int, not an enum. we changed that in global_settings due to since the enum caused failure with eabi builds
22:56:15pamauryIt's a shame that Lear just quit the chan
22:56:22gevaertsit's still an enum if HAVE_REMOTE_LCD
22:56:51pamauryTheSeven: aparently, mc2739 also reported a problem with a stkov in dircache at boot time, I will investigate this
22:57:35UnhelpfulTheSeven: funny, hasn't pamaury been optimizing dircache stack/memory usage?
22:57:41kugelgevaerts: oh, that's the bug I think
22:57:48gevaertsI'm on it
22:58:01***Saving seen data "./dancer.seen"
22:58:23kugelI mean, it should probably be an enum. maybe the setting should be casted in statusbar_position()
22:58:38 Quit flyback (Quit: Leaving)
22:58:42mc2739pamaury: Hi, you had questions about FS #11043?
22:58:50gevaertswell, it should be the same everywhere
22:59:04amiconngevaerts just negated the yellows...
22:59:19gevaertson average it's now correct :)
23:00:06S_a_i_n_tTheSeven: i have charge monitoring on nano2g now...was that you? i swear it didn't work for me before...or, i 'un-broke' something.
23:00:26TheSevencharge monitoring in terms of what?
23:00:47TheSeveni only calibrated the battery gauge a bit
23:00:54S_a_i_n_twps %cp etc. tags.
23:01:06CIA-8New commit by 03funman (r24825): as3525v2: full list of registers + description
23:01:10CIA-8New commit by 03funman (r24826): sd-as3525v2: detail MCI_CTRL bits
23:01:16CIA-8New commit by 03funman (r24827): sd-as3525v2: describe CTYPE bit
23:01:19CIA-8New commit by 03funman (r24828): sd-as3525v2: name interrupt bits (no description yet)
23:01:24CIA-8New commit by 03funman (r24829): sd-as3525v2: command register bits
23:01:31CIA-8New commit by 03funman (r24830): sd-as3525v2: MCI_FIFOTH bits
23:01:33TheSevenS_a_i_n_t: don't think i changed something regarding those
23:01:35CIA-8New commit by 03funman (r24831): sd-as3525v2: MCI_HCON bits
23:01:39CIA-8New commit by 03funman (r24832): sd-as3525v2: no limits on the amount of sectors to transfer
23:01:56funman(no functional changes)
23:01:56TheSevenerm, funman, want to hit 24500? :-P
23:02:14funmannah, 123456 ;)
23:02:53 Quit stripwax (Quit:
23:04:27S_a_i_n_thmmmm, i swear someone did *something*, as %bp & %bc work on my 2g suddenly. weird, but awesome ;D
23:05:34CIA-8New commit by 03gevaerts (r24833): statusbar_position should also return enum statusbar_values if it's a macro
23:05:50CIA-8New commit by 03Buschel (r24834): Remove CLIP_TO_15 from codeclib. Remove tabs.
23:06:22 Join flyback [0] (
23:10:57 Join iq [0] (~iq@unaffiliated/iq)
23:11:19 Quit petur (Remote host closed the connection)
23:13:06 Quit bluebrother (Ping timeout: 246 seconds)
23:14:57 Join bluebrother [0] (
23:14:57 Quit bluebrother (Changing host)
23:14:57 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
23:19:54 Join robin0800 [0] (
23:22:59saratogaFlynDice: looking at the datasheet it seem to me I can adjust the DRAM clock either by:
23:23:04saratoga1) moving PCLK to a different sources and then setting pclk div1 to reduce it back to 64MHZ for everything else (but leave it higher for mem_clck)
23:23:49funmanisn't mem_clk limited to 90MHz ?
23:23:50saratoga2) using the built in Clock ratio, CLK register to set the ARM memory controller to double the clock it gets
23:23:55saratogayeah it is
23:24:17saratogaFlynDice: i think you mentioned trying one of these, which was it?
23:24:27pamaurymc2739: yes
23:24:50pamaurymc2739: what is the maximum directory depth of your file sysytem tree ?
23:25:14mc2739pamaury: no more than 6 deep
23:25:53pamauryhum, the stkov is in dircache thread ? Are your sure of that ? Because a stkov in dircache at boot time is rather strange
23:25:57saratogahmm though the patches i have make it sound like only #1 is used
23:27:15mc2739pamaury: yes, when it does boot, the dircache stack is at 97% or 99%. The stkov will happen right after disconnect and then again on reboot.
23:27:34pamauryon which device ?
23:28:28mc2739pamaury: e280v1
23:29:13pamauryThat's incredible, I also have a e280 and have a directory depth of 20. And I can't see anything in my code which would depend on anything else :(
23:29:16 Join Strife89|PalmTX [0] (
23:30:11funmanare there recursive functions ?
23:30:37 Quit Strife89|PalmTX (Client Quit)
23:30:40 Join JdGordon_ [0] (
23:31:03pamauryThere is one recursive function (sab_process_dir). The number of recursive calls is the directory depth
23:32:25 Join Strife89|PalmTX [0] (
23:34:11S_a_i_n_tHmmmm, bugger....Plugin/pictureflow gives "*PANIC* Stkov main" on Nano2g
23:34:16 Quit Strife89|PalmTX (Client Quit)
23:34:25saratogaright now we run the PCLK off a 248MHz PLL with a 1/4 multiplier, we could instead run the DRAM off it with a 1/3 multiplier (83MHz) and the PCLK off that with a 1/2 multiplier (41MHz)
23:34:40saratogaunfortunately that slows down all the peripherals, and i'm not sure if their drivers will be ok with it
23:34:55saratogato keep the pclk the same i think we'd need to change the PLL
23:35:09kugelgevaerts: thanks for looking into it
23:35:17S_a_i_n_tr24803 that is...
23:35:40bertriksaratoga, it has two PLLs so that provides some extra freedom of choice
23:36:10saratogabertrik: i think the pclk and memory clock must use the same PLL
23:36:26saratogathey're both muxed by PCLK_SELECT
23:36:27 Quit sudoman (Quit: sudoman)
23:36:44 Nick Ypsy is now known as YPSY (
23:37:08saratogafor now i'm just curious if the system would even boot at a lower pclk, ignoring performance
23:38:57*pamaury just noticed something really strange in his code
23:39:05bertriksaratoga, which clock is the memory clock, mpmc_clk?
23:39:23saratogabertrik: correct
23:39:34saratogaMultiPortMemoryController iirc
23:40:36 Join Strife89|PalmTX [0] (
23:41:16saratogai actually dont' understand how you're expected to get a 64MHz PCLK and a 90MHz MEM_CLK when all they give you is a 1/1, 1/2, 1/3 ... divider
23:42:03pamauryTheSeven, mc2739: I could have an explaination ! I happily dereferencing a NULL pointer :/
23:42:24TheSeventhat shouldn't cause a stkov though, should it?
23:42:56kugelUnhelpful: is it safe to load a bitmap twice using the same bitmap struct?
23:43:04kugelwhatever :p
23:43:23pamauryTheSeven: on my device, it runs like a charm ! so I perhaps the result is more or less undefined
23:43:43kugelthis second pass is still buggy
23:43:46TheSevenkugel: safe was right
23:43:56kugelyea, I figured :)
23:45:14funmanthey didn't implement DMA in as353x SD/MMC driver
23:45:28 Join stripwax [0] (
23:45:51 Quit Farthen (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
23:46:19 Quit FOAD (Ping timeout: 246 seconds)
23:47:44 Join FOAD [0] (
23:48:51 Quit S_a_i_n_t (Quit: PC Restart...BBS)
23:48:56CIA-8New commit by 03pamaury (r24835): Fix a dircache NULL-pointer dereference.
23:49:24pamauryTheSeven, mc2739: could you try with this new version ? Perhaps it helps things
23:49:57CIA-8New commit by 03funman (r24836): sd-as3525v2: describe interrupt mask bits ...
23:50:01CIA-8New commit by 03funman (r24837): sd-as3525v2: panic in case of error in the isr ...
23:50:02TheSevenpamaury: i didn't get any stkovs...
23:50:20pamaurytrue, I mixed up things :)
23:52:32 Join S_a_i_n_t [0] (S_a_i_n_t@
23:52:40kugelfunman: sd transfers don't work yet, do they?
23:56:41 Quit Schmogel (Quit: Miranda IM! Smaller, Faster, Easier.
23:58:27funmanthere are no errors or data transfer over interrupts, but there are interrupts anyway
23:58:35funmani must check the status of each interrupt

Previous day | Next day