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 2009-07-13

00:00:08gevaertskugel: that e200v2 failed install on the forums looks like it could be using mtp
00:01:36 Quit krazykit ("Connection reset by beer")
00:02:05CIA-71New commit by 03jdgordon (r21825): Fix the shipped themes statusbar config line
00:02:12CIA-71New commit by 03unhelpful (r21826): Rework to also fix prefix generated-file depedencies with the target directory, and simplify the sed expression for mkdepfile.
00:06:09CIA-71New commit by 03peter (r21827): Update Dutch langfile
00:06:23 Join thegeek_ [0] (
00:08:16 Join krazykit [0] (
00:14:48 Quit tvelocity (Read error: 110 (Connection timed out))
00:15:39 Quit dys (Connection timed out)
00:15:51 Join dys` [0] (
00:16:30 Nick dys` is now known as dys (
00:16:54CIA-71New commit by 03dave (r21828): Basic changes to add nano 2g to the build system, based on the Meizu M3 port. The bootloader builds, but does nothing due to lack of any drivers.
00:17:14 Join tvelocity [0] (
00:18:28kugelJdGordon: the recording screen bug is a blocker :(
00:18:49JdGordonwhats the issue there?
00:18:59amiconnZagor: "Duplicate name" on stalled client connection still exists
00:19:29Zagor22:17:36 Joined: client nereid-amiconn
00:19:32amiconnNot always, but it just happened when fixing the system time in my VM
00:19:36Zagor-22:18:25 HELLO dupe name: nereid-amiconn
00:20:26*gevaerts also fixed the time on his VM and saw the same symptom
00:20:33kugelJdGordon: when you leave it, the main menu is in the viewport of the recscreen list
00:20:51kugelthe peakmeter shows above (not moving)
00:23:10gevaertsZagor: Unrecognized command '_UPLOADING
00:23:29kugelgevaerts: indeed, he was in mtp
00:23:35Zagorgevaerts: a single time?
00:23:45gevaertsno. For every upload
00:23:48kugelI have that too
00:24:04Bagdercommand '_UPLOADING
00:24:04 Quit thegeek (Read error: 110 (Connection timed out))
00:24:12Bagdernotice the newline
00:24:16Zagoroh right, I committed that but didn't bump the required version
00:28:03 Quit dmb (No route to host)
00:29:07 Quit bubsy (Read error: 60 (Operation timed out))
00:29:13 Join kugel_ [0] (
00:29:23 Quit kugel (Nick collision from services.)
00:29:29 Nick kugel_ is now known as kugel (
00:30:00 Join JdGordon| [0] (i=43a02c5a@gateway/web/freenode/x-c60bd0e27785ea80)
00:30:05CIA-71New commit by 03zagor (r21829): Allow param-less commands.
00:30:39 Join funman [0] (n=fun@rockbox/developer/funman)
00:33:14JdGordon|kugel: the rec screen isnt juts missing some really obvious "fix viewports" call on exit is it?
00:33:14 Quit Zagor ("Leaving")
00:33:30kugelno, it's a bug in my list code
00:35:24 Join Thundercloud [0] (
00:37:05r0b-kugel might i ask why bindings on the sansa players are odd
00:37:28r0b-atleast for Doom
00:37:47 Join GreatBeaver [0] (
00:37:51funmanwhich sansa players ? (there are quite a few)
00:38:09r0b-well for the E250's
00:38:17r0b-errr E200's
00:39:58funmanare meizus big endian ARM ?
00:40:09 Quit petur ("Zzzzz")
00:42:20 Join AndyIL [0] (i=AndyI@
00:42:25linuxstbfunman: Yes
00:43:26CIA-71New commit by 03amiconn (r21830): * Fix overlooked r12 usage possibility in mpegplayer ARM idct ...
00:47:23 Quit dash32 (Remote closed the connection)
00:48:20 Quit notlistening ("Leaving")
00:49:25kugelJdGordon|: found it
00:51:30funmanI can modify/delete post of other users on some forum sections, but not on "Official Test Builds", is that normal ?
00:52:22 Quit ej0rge ("leaving")
00:52:48JdGordon|funman: whack scorche|sh
00:53:18 Quit AndyI (Read error: 110 (Connection timed out))
00:56:51***Saving seen data "./dancer.seen"
00:57:22funmanscorche|sh: do developers have permission to remove/edit forum posts in "Official Test Builds" section?
00:57:39scorche|shlet me check
00:57:55JdGordon|they do, unless i have special permission.. which i doubt :p
00:58:48kugelyea, that's hardly imaginable :)
00:58:54kugelthe new patch is up, btw
00:59:17JdGordon|ok, ill have a look later... any obvious reasons to not commit it?
00:59:25JdGordon|want to get some test builds going for the forums?
00:59:28kugelI added some comments to viewport.h, to discribe a bit what the new functions do
00:59:59kugeluhm, not really. for me, the best test build is the current build :)
01:02:47kugel800K binsize only
01:03:17scorche|shfunman: well, i guess if JdGordon| can....can you not?
01:03:49funmanata-sd-pp.c was moved 3 times since it was created ..
01:04:06funmanscorche|sh: no
01:04:49JdGordon|kugel: 800K?
01:04:49kugelfunman: I seem to be able
01:05:04kugeldelete at least, I don't see an edit button
01:05:18scorche|shfunman: you should be able to now...
01:05:19funmankugel: the name is "modify"
01:05:34funmanscorche|sh: thanks, now I can ;)
01:05:41kugelI know, I don't see this button though
01:05:46kugelJdGordon|: yes
01:05:50 Join robin0800 [0] (
01:06:08funmanscorche|sh: was I blacklisted or something ? o:
01:06:12kugelno, 800bytes of course :D
01:06:13 Join mt [0] (n=mt@
01:06:24scorche|shfunman: should be the same as all other devs
01:07:42kugelJdGordon|: have a look, I think I'm going to use the GUI_EVENT_REFRESH event for doing cleanup and the likes. It can take a redraw function as parameter
01:08:30JdGordon|I dont really agre that that event is needed, but ok
01:09:00kugelWhy? it does a wasteful clear, update and possibly another redraw function
01:09:07kugelthat all isn't needed when no viewport is needed
01:09:25kugelis used*
01:11:06 Join robin0800_ [0] (
01:11:42 Quit robin0800_ (Client Quit)
01:11:50 Quit robin0800 (Remote closed the connection)
01:12:09 Join robin0800 [0] (
01:12:25kugelfeel free to come up with something better. I think the event is good, but I'm not overly obsessed of it
01:17:51JdGordon|my thinking is once background wps is enblaed it wont be needed
01:22:19kugelwe can just remove_event() then
01:22:56mtlinuxstb: ping
01:23:01linuxstbmt: pong
01:23:30kugelr0b-: feel free to fix, but at least for doom you can actually change the keymap
01:24:01mtlinuxstb: Any comments on the last commit ?
01:24:38 Quit bertrik (Read error: 113 (No route to host))
01:25:19linuxstbmt: I havent' really looked at it... The main commit was r21806?;revision=21806
01:26:26funmanby the way why doesn't wincent have SVN commit access for his work on puredata ?
01:27:10mtI had that same question too ^
01:27:35linuxstbfunman: There's a mailing list for that question...
01:27:35*kugel isn't sure whether this is a topic to be discussed in public irc
01:28:43 Quit DarkDefender ("Leaving")
01:29:09mtlinuxstb: btw, here's a brief description on what I did :
01:29:21funmanI remember all gsoc students would be given access but perhaps i'm wrong here, sorry
01:29:24mt(last point)
01:30:42linuxstbfunman: Students are treated the same as other patch authors. But as kugel said, this isn't the place to discuss individuals.
01:30:49 Quit martian67_ (Remote closed the connection)
01:31:07r0b-kugel if i had the build enviroment i would
01:31:16 Join martian67_ [0] (n=martian6@about/linux/regular/martian67)
01:31:29r0b-as i cant get the tools for CYGWIN
01:31:52 Quit Thundercloud (Remote closed the connection)
01:36:52 Quit robin0800 (Remote closed the connection)
01:37:01 Join DarkSpectrum [0] (
01:38:37 Quit kugel (Remote closed the connection)
01:39:52 Join robin0800 [0] (
01:41:11 Quit GreatBeaver (Read error: 104 (Connection reset by peer))
01:43:55 Quit robin0800 (Remote closed the connection)
01:44:14 Join robin0800 [0] (
01:44:33 Quit gregzx ("ChatZilla 0.9.85 [Firefox 3.5/20090624025744]")
01:45:25stripwaxHm, clix keymappings seem messed up. I can't go 'up' on ipod target.
01:48:02linuxstbmt: I don't really understand your code, but it looks nice and short ;) Have you tested it with large files?
01:48:02stripwaxhm, and online ipod video html manual only goes up to section 4? The Manual builds seems borked.
01:48:23stripwaxpdf looks fine, just ipod video html online manual
01:48:51 Quit robin0800 (Remote closed the connection)
01:49:21linuxstbstripwax: Yes, I think fml broke the manuals last night - pixelma pointed it out to him.
01:49:21mtlinuxstb: yes, tested it on all the samples I have (2 seconds to 30 minutes long)
01:49:34linuxstbHow big (in MB) is your 30 minute file?
01:49:40stripwaxlinuxstb - ah.. :(
01:49:54mtlinuxstb: about 13 MB
01:50:05 Quit stripwax ("")
01:50:19linuxstbmt: Hmm, so it might still be completely buffered. I was curious about a file larger than the audio buffer.
01:51:00mtHow large is the audio buffer ?
01:53:14linuxstbabout 26-28MB on most targets.
01:53:42funmanabout 60kB on c200v2
01:53:45r0b-i think mine is 30MB
01:53:48mtlinuxstb: Also, is it possible to have seek_time > duration so that this case should be explicitly handled ?
01:53:52r0b-on the e200c1
01:54:13linuxstbmt: I don't know. I would hope not (the core playback code should prevent that).
01:54:38linuxstbfunman: ;)
01:54:55mtfunman : KB ? and is it around that for the c200v1 ?
01:55:35funmanmt: no, the c200v1 has 32MB of SDRAM, while the c200v2 only has 2MB
01:55:52r0b-why would they cut the RAM?
01:56:38r0b-this is why i use older players :)
01:56:42mtlinuxstb: if I kept seeking beyond the end of the file I get a codec failure (control reaches the seeking code) unless I make it ( goto done; ) .
01:57:37mtThis happens in cook. I don't think it happens for other codecs ?
01:57:58linuxstbmt: I'm not sure. You should look at other codecs, and see how they behave.
01:59:57 Part wincent ("Kopete 0.12.7 :")
02:00:01 Quit JdGordon| (Ping timeout: 180 seconds)
02:00:05DarkSpectrumwhat does the M after the version number on the C200 mean?
02:00:05*FlynDice wonders if "SD Expert" funman could shed some light on sdhc determination in ata_sd_as3525.c if he's got some time....
02:02:24mtlinuxstb: ok, fixed. ;)
02:02:26*funman was just notified by his secretary that he has got some free time right now
02:02:50FlynDiceit seems right now it gets checked for with SD_SEND_IF_COND and I thought this just checks voltages if I'm reading correctly?
02:03:03 Join dmb [0] (n=Dmb@unaffiliated/dmb)
02:03:47FlynDicecmd8that would be..
02:04:54FlynDicearound line 257
02:07:00funmanFlynDice: I understand that CMD8 (SEND_IF_COND) was added in SD Specification 2.00 which adds SDHC support, and that non SDHC cards would just ignore this command
02:07:45funmanwith a quick look I can't find this explicited in the specification itself, but I find this explanation in ata-sd-pp.c and various webpages/forums after a google search
02:08:29FlynDicefunman: The way I'm reading it is that if the card supports the voltage range it will echo back the voltage bits and the check pattern
02:09:22FlynDicefunman: and it says that bit 30 will be set to 1 if you have a hc card after it has inited
02:09:45funman It is
02:09:49funmanmandatory to issue CMD8 prior to first ACMD41 for initialization of High Capacity SD Memory Card
02:10:24mtlinuxstb: Do you know of any place where I could find such large samples ?
02:10:26funmanFlynDice: bit 30 of the argument must be set to 1 (by you, not by the card), when sending ACMD41
02:10:43FlynDiceyes, but you can issue it to a non hc cardI think
02:11:03 Quit martian67_ (Connection timed out)
02:11:29linuxstbmt: Possibly on the BBC's website - - they archive a large number of radio shows in realaudio (or at least, used to...) I'm looking now.
02:11:52funmanFlynDice: ACMD41 yes, but not CMD8 afaiu
02:12:12FlynDicebit 30 , card capacity status, has this note : 1) This bit is valid only when the card power up status bit is set.
02:13:09FlynDiceI'll go read it again and see if I can make sense of it....
02:14:13funmanI remember 0x1AA is a documented value, but don't remember where I saw that ..
02:14:27TheSevenlinuxstb, markun: I just realized that the S5L8701 (nano2) seems to only have 176 instead of 256 KB of SRAM!
02:14:36TheSevenyou'll need to adjust that.
02:15:08pixelmaDarkSpectrum: the M means modified build
02:15:12linuxstbTheSeven: OK, thanks. But it has 32MB RAM?
02:15:19funmanFlynDice: anyway, this matches the OF code as far as i remember
02:16:35*TheSeven is gonna go to bed now...
02:17:05 Part TheSeven
02:17:31FlynDicefunman: yes, I can make sense of the 0x1AA, it suggests using AA as the check pattern
02:18:02funmanFlynDice: right, bits 7:0 are the check pattern, and bits 11:8 are the supply voltage (lsb surely means lowest voltage)
02:18:42funmanFlynDice: where is this suggestion in the spec?
02:19:37FlynDicefunman: lemme look
02:20:47FlynDicep 40 on my sheet last line in a box below 4.3.13 Send Interface Condition Command (CMD8)
02:20:51funman4.3.13 : "it is recommended to use '10101010b' for the 'check pattern'" (10101010b = 0xAA)
02:21:33funmanhum it's page 51 here (out of 129) : "Part 1 Physical Layer Simplified Specification Ver2.00 060925.doc"
02:22:54FlynDicepg 52 of the pdf but page 40 at the bottom of the page
02:23:11linuxstbmt: Here are some "listen again" links to radio programmes from the BBC - they should all be at least an hour long.
02:23:32linuxstbmt: I use "mplayer" to download them - "mplayer -dumpstream -dumpfile myfile.ra rtsp:///...."
02:23:34funmanFlynDice: Ok, it's page 40 here as well
02:26:12mtlinuxstb: thanks a lot !
02:26:23linuxstbmt: Lots of nice samples ;)
02:27:26mtYeah. :)
02:29:27 Quit funman ("free(random());")
02:30:12 Quit QncOPFGl ("Ex-Chat")
02:36:17Unhelpfulit seems to me that we can remove the @echo, and also move the bmpdepfile call up one and have it write to the temp file?
02:40:40CIA-71New commit by 03unhelpful (r21831): Add new asmdefs mechanism for exporting information only available to the C compiler for use in asm files, and use it in arm jpeg idct. See ...
02:42:06 Quit mt (Remote closed the connection)
02:53:36 Join mc2739 [0] (
02:54:11mc2739Unhelpful: I think you have red
02:55:06JdGordonnew even!
02:55:42JdGordoncompletly destorying the old record by 400K!
02:56:52***Saving seen data "./dancer.seen"
02:57:07 Quit bzed (Remote closed the connection)
02:57:10 Join bzed [0] (
02:57:10linuxstbThe first million-point commit!
02:57:40linuxstbmc2739: You "think" ? ;)
03:00:11mc2739well, it is not completely red - player boot is green
03:02:47 Join cr08 [0] (
03:04:40 Join timc [0] (
03:13:43 Quit BHSPitLappy (Remote closed the connection)
03:14:33 Join saratoga [0] (i=9803c6dd@rockbox/developer/saratoga)
03:15:42CIA-71New commit by 03unhelpful (r21832): Fix red: broken tools/
03:19:31saratogamt: (for the logs) 48.24MHz needed for 64kbps cook on my Sansa
03:20:39linuxstbsaratoga: That doesn't seem too bad, considering there's been no real optimisation.
03:20:54saratogalinuxstb: yeah just looking at it now
03:21:54saratogahmm mt told me he was using the IMDCT library today, but looking at the code it doesn't seem so to me
03:25:16saratoganevermind was just confused by the various wrapper functions
03:34:49 Quit efyx_ (Remote closed the connection)
03:38:54saratogahmm putting COOKContext's buffers into IRAM has a surprisingly small impact on performance
03:40:02saratogait only gives me 42.46MHz
03:45:49 Join karma [0] (
03:46:10karmaGuys! I have shred the whole sansa drive
03:46:22karmaConnecting it to the usb will not mount anything
03:46:24saratogamt: (for the logs) the gain_table variable in COOKContext looks unused to me
03:46:26karmaWhat can I do?
03:46:46JdGordonwhich sansa?
03:46:46saratogakarma: have you seen the e200UnbrickGuide in the wiki?
03:46:56karmasaratoga, no.. let me see
03:47:04karmaye, sansa e200 something
03:47:11saratogaassuming you have a V1 sansa, thats what you need to do
03:47:40karmaand i thin i do have a v1
03:49:40Unhelpfulsaratoga: the codeclib mdct... it's all ints and arrays of ints?
03:50:03saratogaUnhelpful: its all integer, but I'd have to double check to see if theres any structs
03:50:10saratogai didn't write it, just borrowed from Tremor
03:51:06Unhelpfulyes, structs passed to asm are precisely what i'm hunting. :)
03:51:22saratogathe c version of the fixed point maths use unions
03:51:56saratogabut that shouldn't ever interact with any ASM code
03:52:23Unhelpfulright... it's only structured data passed to asm code that i'm looking for, to add more asmdefs files :)
03:52:35saratogaUnhelpful: no it looks clean
03:52:55saratogaUnhelpful: do WMA files crash with your changes?
03:54:37Unhelpfulso far i've only tested things that i happened to have on my e200... curiously, the long-call stubs without the rest of EABI break vorbis.
03:55:52Unhelpfuli built a toolchain that ought to be generating the same asm, as it's gcc-4.0.3 with the same patches we use except for passing a flag to the assembler to tell it to generate EABI objects. since the compiler doesn't know it's doing anything differently, structs and such ought to be exactly the same... but vorbis files end instantly :/
03:57:27Unhelpfulvorbis stalls at the start of playback with gcc-4.3.3 and full EABI. MP3 crashes with an undefined instruction at an address beyond its icode section if i use gcc-4.3.3, even without EABI... but there might still be some oddball ABI change there.
03:57:49saratogaUnhelpful: vorbis has a lot of ASM
03:58:02saratogaI suggested WMA since it has very little ASM beyond the IMDCT library
03:58:06Unhelpfulum... the IDCT lib uses either 8-bit or 31-bit constants?
03:58:19saratogashould be all 32 bit
03:58:28saratogawhat are you looking at?
03:58:47saratogaanyway I would try out WMA and/or AAC since they're not as well optimized but use IMDCT, so if they play you know the library is fine
03:59:35CIA-71New commit by 03saratoga (r21833): Put COOKContext struct into IRAM. Speeds up decoding by 6MHz on PP5024 at the cost of 30.5 kbytes of IRAM.
03:59:39Unhelpfuli'm looking at the constants in the mdct2.h
04:00:38saratogawe don't use LOW ACCURACY so they should be 32 bits
04:02:03 Join alienseer23 [0] (
04:03:19saratogamt: (for the logs) I put COOKContext into IRAM for a nice little speed up, but that puts quite a few buffers into IRAM
04:03:38saratogaall look important but its possible one or two of them should be removed in favor of more important buffers elsewhere in the codec
04:03:43alienseer23on ubuntu 9.04 i downloaded the ipodpatcher, chmod to allow execution, and execute as root, but only gives me permission denied. what did i miss?
04:05:03saratogalinuxstb: I just realized the fixed point operations are done in c
04:05:10saratogano wonder its slower then I expected!
04:05:31saratogaalienseer23: have you tried rbutil
04:06:20alienseer23same issue
04:06:25alienseer23permission denied
04:08:39Unhelpfulthe "high-precision" ones are used in the asm version as well... armv5 has some very nice multiply goodies if you're doing 16x16 or 16x32. but i guess that we can't use 16-bit constants for 16-bit audio...
04:09:16saratogaUnhelpful: we can in a lot of places since rounding error gets distributed through out all the time domain samples when you do it in the frequency domain
04:09:34saratogabut we've never had ARMv5 targets before so it hasn't been done
04:10:42saratogai am certain the MDCT (and also MP3 synthesis filter) are much slower then they ought to be on ARM5/6
04:12:48saratogathese cook fixed point ops seem silly, who puts a branch in a fixed point multiply function
04:12:53Unhelpfulsaratoga: i'd rather keep rounding errors below the range of the final output if possible :)
04:13:12saratogaUnhelpful: we've got a lot of accuracy to play with right now
04:13:39saratogawhen i was testing WMA we'd do better then the Windows floating point decoder
04:13:41Unhelpfulright, but you can't go to 16x16 multiplies without getting rounding errors into the intermediates.
04:14:05saratogais there a savings for 32x16 multiplies?
04:14:25saratogayou could probably truncate a lot of the trig constants without meaningful loss of accuracy
04:15:27Unhelpfulfor the armv5+ 16x16 and 16x32 multiplies, 1) they all come in a multiply and a multiply-accumulate flavor 2) the 16-bit inputs can be selected from either the low or high half-word
04:16:32saratogabut they're probably not any faster then just doing a 32x32->64 mul
04:16:50saratogahmm cook seems to use mostly 16 bit fixed point multiplies
04:16:54saratogai wonder why they did it this way
04:16:56Unhelpfulthe jpeg IDCT gets a rather large boost from smulxy/smlaxy, as it can load two inputs per register, and two constants per register, and if a constant is never needed in an add, you will never need to sign-extned it.
04:21:49Unhelpfulalso 5e lets you load two register at once with pc-relative addressing (though limited in range) and timing the same as a 2-register ldm
04:22:41 Quit daurnimator (Read error: 104 (Connection reset by peer))
04:23:47 Quit Zarggg ()
04:24:11 Join Zarggg [0] (
04:31:33 Quit tvelocity (Remote closed the connection)
04:34:45Unhelpfulsaratoga: i've found the tables for the smulxy timing on arm9e... i'm not really understanding what they say at all. :/
04:35:17saratogaUnhelpful: the multiplier is fully pipelined right? so unless you can save register loads, I don't think theres any savings
04:36:36 Quit mc2739 ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060215]")
04:36:49Unhelpfulsaratoga: if i'm understanding correctly, these multiplies are all one-cycle operations in terms of the pipeline, with an extra one-cycle stall if the next instruction uses the result. if that's the case, they're really making it rather confusing. :)
04:37:16saratogathat would make sense
04:37:38saratogaso basically you can save IRAM (which doesn't really help much for ARM5/6) and you can perhaps save some register loads
04:37:40Unhelpfulbut yes, you can save register loads with the 16b parts of the 16x16 and 16x32 multiplies.
04:38:29saratogawell the trig tables just sample points along a sin/cos function, so in theory you could take the larger amplitude parts, shrink them to 16 bit and probably not have any more rounding error then you would the lower amplitude parts anyway
04:38:43saratogabut I'm not sure how much time you'd really save
04:39:05saratogaactually I think we should probably look more carefully at the ffmpeg IMDCT again before putting too much work into the Tremor one
04:39:13Unhelpfulalso, if you have 16-bit input data, you can save sign extension... whether it's on load via ldrsh, or later with sxth for armv6 or shift pairs for earlier.
04:39:45Unhelpfulyou're helped rather a lot here compared to the image/video domain... the MDCT seems to be a bit more standard in its implementation?
04:40:28saratogayes its basically entirely standard
04:40:50saratogaaside from MP3, all MDCT codecs we have and even some we don't (wma pro) use exactly the same IMDCT formula
04:41:18Unhelpful"the" DCT and IDCT seem to have various tweaks to normalize values.
04:42:39saratogaanyway cook seems to use rather slow fixed point math, which is surprisingly to me that its still as fast as AAC, so with proper fixed point routines I expect it will be extremely fast
04:42:53saratogaprobably wouldn't take much effort to make it much faster then all the other MDCT codecs
04:44:44 Part alienseer23
04:46:15Unhelpfulsaratoga: um, the arm9e manual shows mul as taking 3 cycles, 4 with interlock, on arm9e. so, on our armv5 targets (thought i think none of those are supported), it might save a bit?
04:47:16saratogaUnhelpful: I'm not sure what the break down is for actual arm targets, but I think all ARM9E and above have 1 mul/clock throughput
04:47:37saratogaperhaps there are ARM5 targets without the fast multiplier? or maybe thats latency and not throughput
04:48:11Unhelpfuli think it might be. these tables are really quite confusing to me... but the "table" seems to actually be listing what's happening each cycle.
04:52:06Unhelpfuli see smull with an output register used in the next instruction, though... so perhaps this could be made to go a bit faster. :)
04:53:39Unhelpfulalso, on arm7 targets i think you may see a speed gain if you put the constant in the first input operand, since the constants are full-precision.
04:54:06saratogaUnhelpful: I was looking at the old ARM docs, and they refer to the ARM7 multiplier as the "fast" version, apparently they used to have a 2 bit multiplier unit, so the 8 bit one is "fast"!
04:55:16Unhelpfulhah! the "fast" multiplier is 32x8... the idea of a 32x2 multiplier just sounds funny to me :P
04:55:31saratogaalso funny, on the ARM922, the MMU that we don't use is almsot as large as the entire CPU
04:55:45Unhelpfulit's the *tdmi targets that have the 32x8 multiplier, right?
04:55:59saratogaand I think some StrongARM
04:56:45Unhelpfuli don't *think* we have any arm9tdmi? but we have many arm7tdmi targets, don't we?
04:56:54***Saving seen data "./dancer.seen"
04:57:01saratogaUnhelpful: I think AMS is basically ARM9TDMI
04:58:09saratogaah yes Wikipedia says all non-E ARM9 cores are TDMI
04:58:10Unhelpfulif i recall how this works properly, you should see a gain in speed from putting those 32-bit constants in the first input operand, and the variable input, which at least *might* be 16 bits, in the second.
04:58:24saratogayes because it can do early termination
04:58:48saratogaI never really looked into it though
04:59:10saratogatheres probably some savings possible
04:59:47saratogaplease feel free to look into our MDCT if you need a new project
05:00:05saratogai can send you my half completed splilt radix IMDCT too
05:00:23Unhelpfulwell, it looks to me like mdct_arm.S has a couple of spots for small optimizations of this sort... it's putting the constants second everywhere, and i also see smull instructions followed immediately by shifting the high part of the output
05:00:45Unhelpfulthose occur in pairs, so you could save some interlock pain by changing mul shift mul shift to mul mul shift shift
05:01:08saratogai'd be very interested to know if that gives any significant speed up
05:01:20 Quit n00b81 ("Leaving")
05:02:07Unhelpfulwell, how hard would it be for me to find a chunk of "typical" input data to these functions? would we maybe be able to rig one of the codecs to dump MDCT inputs with DEBUGF?
05:02:53Unhelpfulthe TDMI multiplier's early termination makes me think you ought to use real-world input to bench it, and not just zeroes, or whatever was on stack before you allocated an array :)
05:02:56saratogaUnhelpful: i always just tested by doing test_codec on a short WMA file
05:03:35saratogaWMA uses almost exactly 50% of its time in the IMDCT so its easy to figure out how much it speeds up
05:04:22Unhelpfulwill i find a set of short sample files on the wiki?
05:04:27saratogabut otherwise, yes you can just dump the contents, or make up any input thats doesn't have too many zeros
05:04:58Unhelpfulyes, but too few zeroes can cause your benchmarks to look much *worse* than they should.
05:05:09saratogathat is my standard file
05:05:15 Join daurnimator [0] (n=daurnima@unaffiliated/daurnimator)
05:06:01saratogai actually don't think i ever checked if its common to have a lot of zeros in the input
05:06:06saratogai can do that right now though
05:08:29Unhelpfulthey don't even need to be zeroes, though, TDMI can terminate early on each whole byte worth of zero/sign bits
05:10:38Unhelpfulwhat about values that have been multiplied by one of the smaller constants in a prior stage, since those are hi32 multiplies?
05:13:21saratogaUnhelpful: quite blocks are almost all zeros, so they should be fast, let me check a louder one now . . .
05:14:08saratogaUnhelpful: I don't follow that question
05:16:06Unhelpfulsaratoga: a non-zero input to a multiply can still save cycles on TDMI when the top 8, 16, or 24 bits are all 0 or all 1
05:16:22saratogabut about hi32?
05:16:38saratogaUnhelpful: I've got a couple MB worth of MDCT inputs, you want them?
05:16:59Unhelpfuland i see smull that only keep the top 32 bits - so if outputs of those with small constants are fed to later multiplies, they might also be "small" inputs.
05:17:10saratogathe composition seems to very considerably though, with lots of values that are less then 24 bits, a few that are nearly all zero blocks, and some that are all 32 bit blocks
05:17:49Unhelpfulby "hi32" i mean 32x32->64 multiplies that discard the bottom half
05:19:01saratogaheres the MDCT for a typical WMA file if your'e curious:
05:19:11saratogatheres actually a lot more variety then I expected
05:19:39saratogawait that can't be right
05:20:26Unhelpfulwell, i've got the first round of wma/vorbis/mp3 speed tests done - i happen to have ~1min mp3 and vorbis tracks on my e200 anyway. :)
05:20:33saratogacrap had the pointer wrong
05:20:57saratogaMP3 uses its own MDCT by the way, not the library one
05:22:56saratogabut anyway, the input the the IMDCT have lots of values that will early terminate, so I'm not sure it'd be much better to flip the constants around
05:23:15saratogathe actual coefficients may be zero more often
05:23:31 Join BHSPitMonkey [0] (n=stephen@unaffiliated/bhspitmonkey)
05:24:12 Quit Horscht ("Verlassend")
05:24:28 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
05:25:05Unhelpfulsaratoga: but if i recall the order correctly, mdct_arm.S has the constants in the slot that can early-terminate.
05:25:25saratogai misunderstood you before then
05:25:45saratogaat least for 128k files, theres a lot of zero coefficients
05:26:33Unhelpfullooking at the jpeg IDCT source that i wrote after i benched the multiplies both ways on e200, yes, the first input is the 32-bit input, the second is the one that is chunked.
05:29:28 Join martian67_ [0] (n=martian6@about/linux/regular/martian67)
05:29:31saratogaUnhelpful: would you like me to send you my profiling data for the MDCT functions?
05:30:55 Quit cr08 ()
05:30:58 Quit martian67 (SendQ exceeded)
05:31:00 Quit martian67_ (SendQ exceeded)
05:31:59saratogaactually looking at it i can summarize easily: mdctbutterflies uses 50% of the MDCT time, post rotation, pre rotation and bitreverse each use about 1/6th
05:33:51Unhelpfuli also think that some constant loads could move up to somewhere much earlier than their use
05:33:59 Join dash32 [0] (
05:34:46Unhelpfuli see a couple of places where constants are loaded immediately before use, but where that register is untouched for a very long time before
05:35:55Unhelpfulare x1/x2 pointers to constants in mdct_butterfly_generic_loop?
05:37:29saratogaUnhelpful: looks like they're the mdct coefficients
05:39:49Unhelpfullooking at the C version T0 is declared const, and sincos_lookup is passed in that parameter... so those are the inputs i should probably assume to be large
05:40:03 Join martian67_ [0] (n=martian6@about/linux/regular/martian67)
05:41:10saratogaok i have to get back to work tonight, good luck
05:42:04 Quit saratoga ("Page closed")
05:46:21 Join kinetic [0] (
06:27:10 Quit dash32 (Remote closed the connection)
06:33:14Unhelpfulhrm... ~2% speedup on vorbis and mp3...
06:35:30 Nick karma is now known as ayahuasca (
06:39:55 Nick fxb__ is now known as fxb (
06:40:45 Quit daurnimator (Read error: 113 (No route to host))
06:50:04CIA-71New commit by 03unhelpful (r21834): Reorder some operands to increase frequency of multiply early termination on TDMI targets, reorder some operations to try to reduce stalls.
06:56:57***Saving seen data "./dancer.seen"
06:59:33r0b-if i got a MicroSD card can i also load games for Rockboy or no?
07:14:08 Nick fxb is now known as fxb__ (
07:15:55krazykitit doesn't matter where the files are located
07:16:30 Join saratoga [0] (n=41becb3b@gateway/web/cgi-irc/
07:17:17saratogaUnhelpful: you were looking at synth_full?
07:18:00 Join einhirn [0] (
07:19:57 Quit DarkSpectrum (Read error: 104 (Connection reset by peer))
07:33:10r0b-ok i think i figured out how to modify the doom buttons
07:33:13r0b-but i have no way to test it
07:33:33 Quit kinetic (Read error: 113 (No route to host))
07:57:55 Quit safetydan ("Leaving.")
07:59:30 Part toffe82
08:11:33 Quit r0b- (Read error: 104 (Connection reset by peer))
08:14:57 Join petur [50] (n=petur@rockbox/developer/petur)
08:29:04 Quit saratoga ("CGI:IRC (Ping timeout)")
08:30:01 Join n1s [0] (n=n1s@rockbox/developer/n1s)
08:34:44 Quit dmb (Read error: 113 (No route to host))
08:45:05 Join flydutch [0] (
08:48:39CIA-71New commit by 03zagor (r21835): Slow clients (avg build score <= 5000) get easier builds.
08:49:26peturyay \o/
08:49:44peturis that revision 23?
08:49:48 Join Rob2223 [0] (
08:50:50petur(Zagor ^^^^)
08:51:23*petur sees he isn't here and shuts up
08:51:45 Join bertrik [0] (
08:53:14JdGordonthe 5000 value is too low though... looks like under 9000 would be better
08:53:43JdGordon5000 would mean next round petur's ts439 will get a full build which it cant handle
08:54:05DhraakellianI'm reliably getting my Fuze to freeze by pressing Prev while on the FM Radio screen with the radio muted
08:54:49Dhraakelliantried in both preset and scan modes
08:55:18Dhraakellianforward too
08:55:18 Join __lifeless [0] (n=lifeless@
08:56:14Dhraakellianrecent svn
08:56:58***Saving seen data "./dancer.seen"
08:57:23*Dhraakellian guesses that there are probably other venues that would be better for reporting such information, but it's almost 3am, and he should've been in bed long ago
08:59:39peturJdGordon: it did get a lot of builds last time...
09:00:27peturI hope that didn't influence the score... could make it an oscillating system :)
09:01:46bertrikDhraakellian, that's a known bug (at least known to me ...)
09:02:05bertrikif you want, you can open a bug report on it
09:02:16Dhraakelliannot right at the moment
09:02:32Dhraakelliansomeone can feel free to poke me to look into bug reporting tomorrow though
09:07:50 Quit Rob2222 (Read error: 110 (Connection timed out))
09:10:20 Quit _lifeless (Read error: 101 (Network is unreachable))
09:12:14 Quit BHSPitMonkey ("Ex-Chat")
09:17:49amiconnUnhelpful: Did you test whether this asmdefs mechanism also works for other archs than arm?
09:19:00amiconnThis would be nice, even if we don't switch abi versions or similar on these in the near future
09:26:10 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
09:31:44 Join mt [0] (n=mt@
09:33:10 Join Thundercloud [0] (
09:34:21mtCan I give negative values to advance_buffer(), so that it returns the buffer to a previous location ?
09:36:43mtlinuxstb: I have tested a 2-hour long (38 MB) sample, everything works fine .. seeking becomes slow though when around the end of the file, so I'll try to speed it up.
09:43:44 Quit bertrik (Read error: 113 (No route to host))
09:44:17 Quit petur (Read error: 104 (Connection reset by peer))
09:44:23 Join petur [50] (n=petur@rockbox/developer/petur)
09:45:18 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
09:48:49amiconnmt: Are cook files constant or variable bitrate, or can they be both?
09:50:03 Quit Thundercloud (Remote closed the connection)
09:53:39 Quit martian67_ (Connection timed out)
09:55:10 Join robin0800 [0] (
09:55:11mtamiconn: constant
09:56:03mtamiconn: It's said on ffmpeg's wiki that all codecs in RM are CBR.
10:03:50mtUnhelpful: ping
10:04:55 Join stoffel [0] (
10:11:59amiconnmt: Does each packet have a timestamp?
10:13:20amiconnIf so, you should be able to speed up seeking by first calculating a rough position in the file from the time to seek to and the bitrate, subtract a few KB, lseek there, sync to the stream, and only go packet-to-packet for the last few seconds of precision
10:13:41mtamiconn: No, each group of packets which form a block/scrambling unit have the same timestamp
10:14:38amiconnThose should also work iiuc
10:14:42mtamiconn: but using the same idea, I could seek to a rough keyframe (the packet determining the beginning of a new block).
10:15:46amiconnCBR should make seeking easy. In fact if all scrambling blocks in a file are equally sized, you wouldn't need to fine tune at all
10:16:19amiconnJust calculate the number of scrambling blocks to skip, multiply by scambling block size, and lseek
10:18:29 Quit DataGhost (Nick collision from services.)
10:18:37 Join DataGhost [0] (i=dataghos@unaffiliated/dataghost)
10:19:00mtamiconn : They should all be the same size yes .. except for the last one, it could be only partially filled .. but that shouldn't be a problem. Thanks for yout help :)
10:21:46linuxstbLlorean: ping
10:28:41 Join DarkDefender [0] (
10:57:00***Saving seen data "./dancer.seen"
11:17:51 Join scorche [50] (n=scorche@rockbox/administrator/scorche)
11:19:11 Join mcuelenaere [0] (
11:37:27 Join Ubuntuxer [0] (
11:37:45Lloreanlinuxstb: You rang earlier?
11:38:08scorcheLlorean: you have a survey to do...
11:39:01Lloreanscorche: I kinda thought I wouldn't be involved with that any more. I can do so if necessary, though I've been more or less completely out of touch.
11:39:37scorcheLlorean: it would have been nice to know...i think all of us were thinking you were going to keep doing that at least
11:40:02scorcheBagder: did you want to be the main mentor then?
11:40:28LloreanWell, I tried to make it clear I was stepping out of all of my positions. I didn't mean to mislead, sorry. I can do it if necessary, as I said.
11:41:18scorchei guess it was just a bit unclear that you were stepping away from that as well
11:41:38scorchedid you want to continue to be his mentor then or just do the survey?
11:42:05LloreanWell, I can do the survey now, and we can discuss if Bagder can manage to take over later, maybe?
11:42:27LloreanI didn't want to force it on anyone, but nobody said "can you still mentor?" to me after I 'quit' so I thought it was okay.
11:43:32 Join daurnimator [0] (n=daurnima@unaffiliated/daurnimator)
11:43:47 Join wincent [0] (
11:46:42mthmm.. turns out the slow down when seeking in a large file was due to the need to rebuffer
11:49:08mtlinuxstb: Is there a way around that ? I've modified the seeking code to speed it up, but the small slow down due to rebuffering is still persistent.
11:53:12mtOr is that normal for files larger than the audio buffer ??
12:02:14 Nick fxb__ is now known as fxb (
12:06:23CIA-71New commit by 03mt (r21836): Modified the code for seeking to speed it up a bit. Instead of searching ...
12:07:21linuxstbmt: Yes, I can't see how you can avoid that - if the data isn't buffered, it isn't buffered...
12:10:50 Quit Ubuntuxer ("Leaving.")
12:10:51mtlinuxstb: I don't really know how the buffering thread works, doesn't it keep buffering more data during playback ?
12:11:38linuxstbmt: It fills up the buffer with compressed audio, and then when the buffer is almost empty, fills it again.
12:11:58linuxstbobo: ping
12:16:36Lloreanmt: Does the slowdown happen in a sort of "user holds button until the time indicator says what they want, user releases button, user waits for spinup then once buffer catches up, hears audio"?
12:18:10mtLlorean: The latter.
12:18:14mtBut now I'm not sure the buffering problem I described is the actual issue ; I keep getting "rebuffer_handle: space is needed".
12:18:59mttill it rebuffers and then playback resumes.
12:20:32mcuelenaerewincent: I hear some 'crackling' noise when playing pdpod_test.pd in the simulator, which doesn't appear when I use pd (0.41-4); is this normal?
12:21:27wincentThis seems to be because I provide data chunks of less than 32k. I do so to minimize delay.
12:21:43wincentBut on the device it works without crackling.
12:22:06 Join TheSeven [0] (
12:23:26mcuelenaerewincent: perhaps you could use #ifdef SIMULATOR to avoid this?
12:24:51wincentmcuelenaere: I did so already, but it seems that I have to change the size of the buffer.
12:30:09mtwincent: Couldn't you change the size of the buffer for the sim ?
12:30:45mcuelenaerewincent: I seem to hear the same distortion (well not exactly the same) on my target (Onda VX747)
12:30:56mcuelenaereit also reboots after a while, but that's probably due to the buggy PCM driver
12:33:17mcuelenaerewhy were the USB_ENABLE_* defines moved to config.h? Now I need to do a full rebuild whenever I want to change the usb drivers
12:33:25wincentI changed the value of AUDIOBUFSIZE for simulator in pdbox.h to 32x of the normal instead of 4x. This removed the crackling.
12:35:08wincentBut again, it multiplied the latency and the amount of calculations needed to fill a chunk of audio buffer.
12:35:31linuxstbgevaerts: Do you know why this line is needed in the Meizu crt0.S? (svn blame attributes it to you) - "mcr 15, 0, r0, c1, c0, 0 // set bigendian"
12:35:53linuxstbBy that time, your (big-endian) code is already running, so I would have thought the SoC would be in big-endian mode...
12:37:09 Join daurnimator_ [0] (
12:37:26linuxstbmt: Do you know which codec is more common in RealMedia - ac3 or mp3? mp3 could be quite interesting to do, if it's used anywhere...
12:37:34 Quit daurnimator (Read error: 60 (Operation timed out))
12:38:44gevaertslinuxstb: crt0.S was moved a few times, and I did the last move apparently. I didn't actually write any of it
12:39:59 Quit robin0800 (Remote closed the connection)
12:40:17 Join robin0800 [0] (
12:44:48mcuelenaerewincent: I think the crackling is due to pcm_apply_settings() getting called multiple times
12:45:02mcuelenaereperhaps I should just check whether the asked frequency isn't already set
12:45:22wincentmcuelenaere: Where??!
12:45:45mcuelenaerewincent: I don't know where, but I enabled all logf() calls and it seems there are calls to pcm_apply_settings()..
12:46:44mtlinuxstb: IIRC only one mp3-in-rm sample is known to exist. :)
12:46:54mt(the one in mplayer's samples)
12:47:14wincentmcuelenaere: You gave me some chill :-) pcm_apply_settings is called exactly two times: The first one when audio device is "opened" by Pure Data and the second time when the same device is "closed".
12:47:44linuxstbmt: Then I guess it's not useful at all...
12:47:49linuxstbmt: What about ac3-in-rm?
12:48:33linuxstbmt: Are you interested in ATRAC? That's a completely new codec to Rockbox, and one where you (I think) will actually need to do the fixed-point conversion yourself...
12:48:35mcuelenaereI'm seeing lots of this: (the 80* lines are in my PCM driver, they mean <ADDR> <SIZE> <UNRELATED1> <UNRELATED2>)
12:48:46*linuxstb wishes he could find his ATRAC-in-rm file...
12:51:04 Join efyx_ [0] (
12:51:24mtlinuxstb: I'm not sure which one is more common, I'll look around. But if it's a matter of what's more interesting to work on, I'd be interested to work on ATRAC yes. I want to try doing fixed point conversion, but given that I haven't actually tried it before, I'm not sure the codec could be ready by the end of the program ?
12:51:48linuxstbI hope you won't abandon us at the end of the summer... ;)
12:52:18linuxstbBut yes, ac3 is probably a good second codec - the codec itself should be easy, and it gives you the chance to make sure your RM code is generic.
12:53:14mtlinuxstb: I'm not planning on running away. :)
12:54:13linuxstbAren't there also some very old realaudio codecs as well (earlier than cook)?/
12:55:16mtI think ac3 will not take much time at all so I could do it then start working on ATRAC, this way even if the program ends and I haven't yet finished we would have something better.
12:56:42mtlinuxstb : RealAudio 1.0 & 2.0.
12:57:01***Saving seen data "./dancer.seen"
12:59:17linuxstbThe other area of work is optimising cook, but it seems saratoga is interested in doing that as well...
13:01:11linuxstbTheSeven: I added the Nano2G to the Rockbox build system last night - you can now compile a "bootloader" build for the Nano2G, which uses the core S5L8700 code wirtten for the Meizu port. I doubt it will run yet, but it's a start of something I think could be helpful for hacking...
13:01:45mtBy the way, how could I measure the speed-up gain on-target ?
13:02:45 Nick daurnimator_ is now known as daurnimator (
13:02:55linuxstbYou need to enable it in apps/plugins/SOURCES, then use "open with" on an rm file
13:04:23mtah ok. Thanks.
13:20:16mtimdct library is > 4 MHz faster than cook's imdct (the one from tremor) !
13:22:13 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
13:28:59 Join dmb [0] (n=Dmb@unaffiliated/dmb)
13:31:53amiconnmt: What target do you test on?
13:32:05mtamiconn: e200 (arm)
13:32:57amiconnOk, so PP502x (ARM7TDMI with a properly working cache)
13:33:07amiconnIs there a cook test file somewhere?
13:34:22*amiconn would be interested in performance figures on coldfire (no data cache at all, i.e. IRAM is very important) and PP5002 (broken cache, so iram is important too)
13:35:49 Quit __lifeless ("suck my blood")
13:38:18mtamiconn: - I was testing the vetenskap one.
13:38:36mtI was trying to get you the mplayer samples website but it seems to be down.
13:40:24mtlinuxstb: Seems like aac is the next most common codec after cook. (I asked in ffmpeg's channel)
13:44:38mtamiconn: If you want to do the same tests I did on the imdct library, you'll have to modify two #ifdef's
13:45:04mtYep, that's the one :) .. it isn't working for me currently though, I don' know why.
13:45:36n1s seems to have rm cook samples
13:45:48*n1s slow
13:46:30linuxstbmt: Ah, OK. AAC would be nice to do then.
13:47:49*amiconn first builds r21830, i.e. without the IRAM change, and will test on iPod G2 and iriver H180
13:52:46 Join graey [0] (
13:53:09graeyhi all
13:53:22graeyanyone inhere knows anything about the beatbox plugin/app?
13:53:52 Quit flydutch (Remote closed the connection)
14:01:11 Join flydutch [0] (
14:07:54 Join mt_ [0] (n=mt@
14:09:02 Quit mt (Read error: 113 (No route to host))
14:09:33 Quit jfc^2 (Read error: 104 (Connection reset by peer))
14:10:57 Quit mrkiko ("Lost terminal")
14:16:09CIA-71New commit by 03mcuelenaere (r21837): Make clix more usable on touchscreen targets
14:22:15 Quit dys (Connection timed out)
14:29:40 Join LambdaCalculus37 [0] (i=44a0430d@gateway/web/freenode/x-17963c3e75eddc50)
14:33:14mcuelenaeregraey: adding it to SUBDIRS doesn't work?
14:33:23linuxstbmcuelenaere: Let him ask the question! ;)
14:33:35*mcuelenaere couldn't wait ;)
14:33:39*linuxstb predicts some top-posting in IRC...
14:33:56linuxstbOr no-posting...
14:34:13graeyso the question is: how can I get rockbox to compile beatbox? adding it to the makefile like normal plugins doesn't work since it's in a subfolder. and yes, I've googled :P
14:35:30 Quit DarkDefender ("Leaving")
14:35:32graeygonna try adding it to SUBDIRS I guess :)
14:36:01graeyalso, a completely different question, but is it impossible to roll a cross-compiler for windows without doing 'linux on windows'?
14:36:23gevaertsgraey: cygwin is not linux :)
14:36:40graeyyeah, but it's still not fully native either, if you catch my drift :P
14:37:13linuxstbgraey: Most of the tools are from the Unix world, so the answer is no...
14:37:20graeyok :)
14:37:48linuxstbAny Unix will work though, even OS X...
14:37:56graeyyeah, I know
14:38:03Unhelpfulamiconn: i didn't test it on other targets, but shouldn't code like that produce label: .word value on pretty much everything?
14:38:04mcuelenaeregraey: well there is something like Interix, but it's not a recommended compiling environment
14:38:14graeybut it's for my crappy exotic laptop, it doesn't run UNIX decently
14:38:28linuxstbmcuelenaere: Isn't that a Unix-like thing for Windows as well though?
14:38:30Unhelpfulmt: you rang? :)
14:38:48amiconnUnhelpful: I have to check that. One detail is that e.g. on SH1 all C symbols are prefixed with _
14:38:56mcuelenaerelinuxstb: yeah, but it's more 'integrated' into the kernel then the other solutions :)
14:39:00graeymcuelenaere: I'll give vbox a go, and if that won't work I'll reinstall windows and go for cygwin instead :) thanks
14:39:14*amiconn would use it in libdemac though, where SH1 is not an issue
14:39:19mcuelenaeregraey: I would strongly recommend any VM over cygwin, cygwin is sloow..
14:40:56Unhelpfulamiconn: is that done by the compiler or the linker? even if it's like that in the asm, we could just add a _? in the perl to handle it.
14:41:14mt_Unhelpful: Yep :) .. I can't find apps/core_asmdefs.h did you commit it ?
14:41:29 Nick mt_ is now known as mt (n=mt@
14:42:08 Quit daurnimator (Read error: 104 (Connection reset by peer))
14:42:14 Join daurnimator [0] (n=daurnima@unaffiliated/daurnimator)
14:42:29Unhelpfulmt: apps/core_asmdefs.h is generated in the build directory from apps/core_asmdefs.c
14:43:04graeymcuelenaere: is there anything faster than either of those? it's an old celeron system, from 2003-2004, so it's not too fast :P
14:43:20mcuelenaeregraey: fastest is native Linux/Unix of course ;)
14:43:32graeyyeah, but if I do that I can't get a graphical desktop...
14:43:36graeyor wifi.
14:43:44graeyI could live without X, but not without wifi...
14:43:45 Join dys [0] (
14:44:04mtUnhelpful : ah ok. I didn't compile a clean build so that's why.. :)
14:45:19Unhelpfulmt: did it fail to generate it?
14:47:03Unhelpfulit might have done if your make.dep was out of date...
14:47:15mtUnhelpful: Yes, but let me recheck.
14:48:46amiconnmt: Umm, elapsed time display is totally borked on coldfire
14:49:58mtamiconn: Is it viewable in a sim ?
14:50:02amiconnHuge numbers (in the order of thousands of hours) jumping erratically etween positive and negative...
14:50:25amiconnI doubt it, as I do think it's an endianess problem
14:50:40graeyI'm setting up a xubuntu VM in vbox, for rockbox development, maybe it'd be nice to host that somewhere as a click-n-go solution?
14:50:47amiconnI would expect it to show up in a sim e.g. on a PowerMac (PPC is big endian too iirc)
14:52:06mtUnhelpful: Yeah, just needed to update the dependencies.
14:54:11*Unhelpful thinks that we could perhaps have a second, and somewhat more complicated, deps scanner that could be run to generate deps for files in SRC that don't appear in make.dep, or when make.dep is older than source files...
14:54:24mcuelenaeregraey: there's one;
14:55:47mtamiconn: elapsed time update is corrupted even with no seeking ?
14:55:58amiconnYes, just when playing the file
14:56:00 Join Tomcat_ha [0] (
14:56:26amiconnTotal playtime display is correct, elapsed time jumps around (far outside the possible range)
14:56:30 Join teru [0] (
14:56:46Unhelpfuleven that would fail if a dependency of an unchanged source file changes
14:57:02***Saving seen data "./dancer.seen"
14:57:11Tomcat_hacan someone help me with setting up the ipod 240 gb monster build?
14:57:38LambdaCalculus37Tomcat_ha: Ask on the forum thread for support for that build.
14:57:46LambdaCalculus37That's an unsupported build.
14:57:59Tomcat_hai know but i thought i could get support quicker if i joined the chat :B
14:58:23linuxstbTomcat_ha: Do you have the standard Rockbox development environment installed, and can you compile an unmodified Rockbox build?
14:58:44Tomcat_hai have done nothing except installing standard rockbox in the past
14:58:57Tomcat_haim looking at this:
14:59:12graeymcuelenaere: yes, but that's for vmware, and vmware is not fully 'free' :)
14:59:18Tomcat_hathe part i dont entirely get is Then you
14:59:18Tomcat_hawill need to run ipodpatcher -a bootloader-ipodvideo.ipod with your ipod
14:59:41mcuelenaeregraey: ah, and I suppose vbox isn't able to load vmware images?
14:59:41Tomcat_hai dont understand where to put the -a bootloader-ipodvideo.ipod with the ipodpatcher
15:00:07graeymcuelenaere: it just might, never thought of that!
15:00:28linuxstbTomcat_ha: Ah, I misunderstood... You need to open a command prompt, and then type that command (after "cd"-ing into the directory containing ipodpatcher.exe and that bootloader).
15:00:34*mcuelenaere remembers some free alternatives to VMWare being able to load VMWare images
15:02:18graeymcuelenaere: the vbox released last week can indeed read vmware images, my bad.
15:02:39Tomcat_hasays that the command isnt recognized as an internal or external command
15:03:00Tomcat_haim typing in -add-bootloader bootloader-ioidvideo.ipod
15:03:52graeymcuelenaere: ofcourse, the vmware-image on hasn't got the guest additions necessary for ease of use, such as mouse pointer integration and/or swapping around files with the host os easily...
15:03:56linuxstbTomcat_ha: You type "ipodpatcher -a bootloader-ipodvideo.ipod"
15:04:20mtamiconn: Negative values too ?
15:04:26mcuelenaeregraey: yes, I can imagine that. Also, that image was going to get an update IIRC
15:04:27Tomcat_hayeah i just figured that out myself
15:04:35Tomcat_haim slow today
15:04:40amiconnmt: yes
15:04:59amiconnmt: Btw, saratoga's IRAM change, despite the rather moderate speedup on PP502x, works wonders on coldfire
15:05:23amiconn244% realtime -> 575% realtime on my H180, using the file you first linked to. That's a factor of 2.35
15:05:25mtamiconn: Could you try changing the type of 'i' to uint8 and see what happens ?
15:05:28graeymcuelenaere: well, I've nearly got a xubuntu setup ready, I could just as well compile all cross-compilers instead of just the one for my sansa, and then it'd be good enough to use for everybody I guess :)
15:05:52mtamiconn: How much of MHz is it faster ?
15:06:15*amiconn only noted down the percentages, not the MHz.
15:06:20Tomcat_haanyone know how i can check if i have the 32 or 64m versions?
15:06:23amiconnThe percentages are much easier to compare imo
15:06:31Tomcat_haof the ipod video?
15:06:43AlexPTomcat_ha: What capacity§
15:06:45linuxstbTomcat_ha: What disk size was it originally?
15:07:21amiconnIt dropped from roughly 50MHz to 21.5MHz
15:07:27Tomcat_haand apparently its not cleanly spread between 32 and 64 for that
15:07:28AlexPTomcat_ha: 64 MB then
15:07:50linuxstbTomcat_ha: Or if strange things happen with the 64MB build, then 32MB...
15:07:53amiconnSpeedup on PP5002 is 15% (288% -> 331% realtime)
15:08:37linuxstbmarkun: Did you write crt0.S for the Meizus?
15:08:40amiconnmt: As mentioned earlier - the coldfires used in our coldfire targets have no data cache, and sdram is slooow compared to a cpu...
15:08:50mcuelenaeregraey: sure, you can put a link in the wiki somewhere I think
15:08:54Tomcat_hai see
15:08:58Tomcat_haill just try this then
15:09:12linuxstbmarkun: I was wondering about the line that sets the CPU to big-endian mode - isn't it already in big-endian mode (as it's running your code)?
15:09:22mtamiconn: Ah .. I see.
15:09:46graeymcuelenaere: yeah, I'll find a host first and put a link up somewhere. Any things I specifically should or shouldn't do?
15:10:07markunlinuxstb: I think it was basically a copy-paste job
15:10:29markuncan't remember setting it to big-endian
15:10:38linuxstbmarkun: No-one seems to accept responsibility for that code ;)
15:11:05markunyes, I noticed my name in the logs when you asked other people about it
15:11:16AlexPlinuxstb: It was me! :)
15:11:18mcuelenaeregraey: you could install roughly the same packages as VMwareDevelopmentPlatform does; also try adding an easy way for users to move files between host & VM (samba is installed in the VMWare image for that)
15:11:19Tomcat_hai assume i dont need to run rockbox utillity now
15:11:25markunlinuxstb: do you have a way to run code on the nano2g now?
15:12:05linuxstbmarkun: Yes, the linux4nano people have worked out an exploit in the notes viewer which lets them start some code which then listens on usb for a file upload, which it then accepts and executes.
15:12:26linuxstbSo no way to permanently install new code, but that's probably a good thing...
15:12:45 Join n00b81 [0] (n=n00b81@unaffiliated/n00b81)
15:13:28graeymcuelenaere: yes, virtualbox has it's own guest additions for that, basically you select a folder that is shared between the host and guest OS, and it's automagically configured right. I think it uses samba, too
15:13:36 Quit Sajber^ (Read error: 110 (Connection timed out))
15:15:04Tomcat_hai get:
15:15:19Tomcat_ha*panic* unsupported physical sector size: 4096
15:15:23 Join Sajber^ [0] (
15:15:30graeymcuelenaere: do you know a place where I can get it hosted? I mean, I can get it on rapidshare or something, but I find those annoying, needing to wait and a low download-speed... is it possible to get it on the rockbox-server itself?
15:15:30CIA-71New commit by 03amiconn (r21838): Fix another file for r12 being a scratch register. Overlooked earlier because this file used ...
15:15:59mtpixelma: ping
15:16:21Lloreangraey: You'd need to offer some significant improvement over the current option, probably, if you wanted yours to be hosted instead.
15:16:53mcuelenaeregraey: you'll need to ask Bagder to put it on the Rockbox server, if you want it on a place with no waiting time I can recommend
15:17:05Tomcat_hai love mediafire
15:18:17 Join Lear [0] (i=chatzill@rockbox/developer/lear)
15:18:32graeyLlorean, mcuelenaere: ok, so the way to go seems to be: put it on, and if it is enough of an improvement ask Badger :)
15:19:13graeypersonally I think any option that gets rid of proprietary products (be they 'freeware' or not) is a big improvement already :P
15:19:16Lloreangraey: Remember that the current one is rather optimized to be a small download, too.
15:19:33Tomcat_hahow can i change the physical sector size now?
15:19:35AlexPgraey: Also, make sure that it can build all of the various architectures, and sims and manuals (I think the current one does that)
15:19:39LloreanJust being "freer" but not being as good elsewhere probably isn't going to do it.
15:20:28graeyAlexP: yes, I'm going to build all cross-compilers rockbox has to offer, if that's what you mean :)
15:20:35linuxstbgraey: How big is your vm? The existing ones are designed to be relatively small...
15:20:54graeywell, my current VM is quite big, but that's ubuntu-based and also has some mono-development stuff installe
15:21:35LearHm, the improved seeking in tremor appears to have been a fairly easy merge. Works in the sim at least. :)
15:21:35graeyI'm gonna do one on xubuntu or just debian, with a dynamicly expanding disk of about 8GB.
15:21:51graeydownload size I'm targetting is 250-300MB max.
15:21:52linuxstbMaybe we could simply provide instructions for setting up Rockbox in a "standard" vm, if they exist (i.e. ones hosted elsewhere...).
15:22:35graeyLlorean: it's not just freeer, it also provides an easier way of sharing files with host os, and virtualbox is less limited than the free vmware player in many ways :)
15:22:35Lloreanlinuxstb: Well, setting up Rockbox development is pretty simplified already ( could be extended to manuals, I guess?)
15:22:43AlexPgraey: That, plus the packages needed to build the sims and manuals
15:22:54graeyAlexP: yes, ofcourse, forgot to mention that.
15:23:26Lloreangraey: Yes, but as I said, it needs to actually meet all of the needs already met by the current one, without drawbacks. Being less limited than vmware player hardly matters if none of vmware player's limitations prevent it from being used for the purpose the VM we provide is designed for.
15:24:04graeyLlorean: I'll make it so that it meets the needs the current one does, and where possible more, without making it bigger in size.
15:24:08linuxstbLlorean: Yes, I was thinking more along the lines of some step-by-step instructions (including downloading and installing the VM), designed for Linux novices. Maybe (or a new script) could do most things for the user, like making sure all required packages are installed.
15:24:27 Join mrkiko [0] (n=mrkiko@
15:24:40graeylinuxstb: I'll make a guide of how I built my VM. should be applicable to both debian and ubuntu installs.
15:24:44linuxstbOr maybe we should simply recognise that people now have more bandwidth, and increase the image size...
15:24:59Lloreanlinuxstb: Something as far as checking for the package, and if not present, offering to manually download it or providing them a list of package names different distros use (or even specific to their distro)?
15:25:18LloreanI think anything less than CD sized is "small enough" probably
15:26:30graeyTrue, lots of people download full distro's these days, not to mention the torrenting and stuff.
15:29:05LloreanIs virtualbox itself as easy, or easier, to use than VMWare? Less technical knowledge expected/necessary?
15:29:48graeyLlorean: when using prepared OS images like you do with VMWare player, yes
15:30:34graeyand you actually have the ability to create your own VMs with virtualbox, something which player hasn't. that could save people bandwidth, by letting them set up an environment with only the stuff they need :)
15:31:44graeyAnd I'll be writing a guide on how to setup a VM in virtualbox yourself, mostly it'll be the same as the existing guide for Linux setup, with some additions about virtualbox
15:31:45LloreanVMWare is more or less for the people who *can't* do things for themselves
15:31:50LloreanOr rather, any VM we host
15:32:01LloreanPeople who can do things for themselves could use the VM of their choice.
15:32:08graeyyes, I know, and virtualbox is just as good at that
15:32:31graeyand by the way, if you can develop C, surely you can start a prepared VM from virtualbox :P
15:33:10LloreanMany of these people can't develop C
15:33:18LloreanThey're just trying to apply a single patch, or a few, and compile
15:33:24graeyyeah, fair enough.
15:33:30graeyI think it's easy enough
15:33:34graeylet me show
15:33:39 Join daurnimator_ [0] (
15:33:40LloreanI'll take your word for it, honestly
15:34:05LloreanJust wanted to make sure it's clear that a major goal here is "ease of setup for the less-technical" rather than "offering a flexible, powerful, free tool that can be expanded to other purposes"
15:34:20LloreanThe latter's nice, but not if we have to sacrifice the first (in my opinion)
15:35:08graeyyes, I understand :)
15:35:27 Part Tomcat_ha
15:37:11 Quit gevaerts (Nick collision from services.)
15:37:20 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
15:38:11graeyLlorean: when you start virtualbox, if you only have one VM installed, it's automatically selected. so novice users would start vbox, click start and be done
15:38:47graeynow I'll stop talking about how much I like vbox and finish the VM ^^
15:39:15 Quit daurnimator (Read error: 60 (Operation timed out))
15:40:23Lloreangraey: Will installing it in windows automatically associate the extensions?
15:41:21graeyI think it does, yes
15:41:30graeylet me check
15:44:20graeyit does not!
15:44:24amiconnmt: Changing the type of 'i' doesn't change anything. Btw, it is recommended to use 'int' or 'unsigned int' for plain variables like 'i'. Using data types of specific sizes often causes gcc to emit extra code, i.e. it's larger and slower
15:44:59amiconn(sign or zero extension in various ways depending on the architecture)
15:45:13Lloreangraey: That's a little bit disappointing. It would be nice to be able to say "install virtualbox, download this file, and double click on it to use our VM"
15:45:15graeyLlorean: you can export vm settings and hard disk to a .ovf, open virtualisation format, but you have to click File -> Import appliance, then you have to browse to it, and only then you can use it.
15:45:17LloreanBut not the worst thing ever.
15:45:53graeyLlorean: I'll update to the latest version, and check virtualbox SVN. if it isn't a feature yet, I'll file a bug report/feature request and it'll probably be in the next version, as it's only a minor fix.
15:47:51graeyLlorean: also, does vmware player support USB forwarding?
15:48:05mtamiconn: Could you log the values of rmctx.audiotimestamp and i ? (Though I'm guessing the problem is with audiotimestamp)
15:48:29Lloreangraey: I'm not sure what you mean by "usb forwarding", if you mean "can it preempt USB in some manner so the devices appear to connect to the VM rather than to windows" I believe the answer is "yes"
15:48:31LloreanI'm not 100% sure though
15:48:58LloreanWell, "rather than to the host OS" I suppose. I only use "windows" there since that's what I was asking about earlier.
15:49:36graeyLlorean: ah, ok. virtualbox does that too, on all host OSes it supports.
15:50:07AlexPAs a side point, VM Server is free and lets you make VMs
15:50:19AlexPYou don't have to use the player
15:50:46AlexPAnd only the none open source Vbox does USB passthrough as far as I'm aware
15:50:51*gevaerts thinks that as long as the free virtualbox can't do USB forwarding, there's no point in preferring it over vmware
15:51:20graeyAlexP: vmware server is only partially free, but I see your point
15:51:36gevaertsgraey: same goes for virtualbox
15:51:44AlexPgraey: It is free as in price
15:52:09AlexPgraey: And if we are arguing about free as in libre, then the same is true for the version of virtual box with all the features (such as USB)
15:52:31graeyalexp: yes, I just noticed, I thought that had changed already, but it hasnt, yet.
15:52:47LloreanDo we consider USB forwarding essential?
15:52:54AlexPProbably not
15:53:23AlexPI don't care which we use, VM or VBox, I was just pointing out that neither are completely "free"
15:53:26graeyEven if it would be essential, the non-open source Virtualbox does it, and virtualbox virtual machines work equally well with both the open-source and the non-open source.
15:53:28gevaertsdepends. You need it for e200tool, tcctool, meizu_dfu, e200rpatcher,...
15:53:30LloreanThough it does make installation easier
15:53:39LloreanWith USB forwarding you don't need access to the files from the host in many cases.
15:53:48Lloreangevaerts: I thought e200rpatcher didn't actually work in the VM
15:53:57LloreanOr some e200R related tool
15:54:11gevaertsah, possibly. I never tried
15:54:14graeywith virtualbox, there is an option for open-source, and only if you need specific features you have to use the closed-source version :)
15:54:45LloreanIs virtualbox equally fast in compile times?
15:54:53*gevaerts doesn't consider this GPL-as-shareware-crippling to be proper open source
15:55:02graeyI wouldn't know.
15:55:24LloreanI mean, the main reason we recommend the VM, rather than Cygwin or similar alternatives, is that it compiles much faster on Windows
15:55:28linuxstbgevaerts: e200rpatcher and tcctool work on windows with libusb.
15:56:00graeyI'll make the VM and try.
15:56:49*Llorean would expect it to perform equally well, but you never know
16:00:10gevaertslinuxstb: yes, but you need precompiled versions of them, or cygwin or mingw, or a VM that has mingw. It gets complicated
16:00:34linuxstbWe have binaries, don't we?
16:00:45graeyLlorean: yeah, I've seen multiple times that vbox is equal to vmware in performance, or that it outperforms it but I've never tried myself
16:00:57graeyjust switching pc's now though, be back in a second
16:01:15*dz needs to get rid of vmware on account of vmware-modules not wanting to play well with 2.6.29 and above
16:01:19gevaertslinuxstb: sure, but this is about development. Maybe the binaries are stable enough not to worry about them though
16:02:09 Join r0b- [0] (
16:02:36 Join thegraey [0] (
16:03:58thegraeyas to localisation, should I stick with debian standard values like USA and stuff?
16:06:58 Quit r0b- (Read error: 104 (Connection reset by peer))
16:07:42 Quit graey (Read error: 60 (Operation timed out))
16:09:51 Join r0b- [0] (n=rob@
16:10:56 Join jgarvey [0] (
16:13:59 Join Strife89 [0] (n=Michael@
16:14:35 Quit gevaerts (Nick collision from services.)
16:14:44 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
16:15:05obolinuxstb: pong - Mornin!
16:17:20 Join AndyI [0] (i=AndyI@
16:20:41CIA-71New commit by 03alle (r21839): Correctly compute the array size regardless of the element type
16:21:08linuxstbobo: Hi. I was wondering how things were going with the View?
16:22:05oboSlowly, as ever. I seem to have hit a bit of a wall trying to communicate with the LCD
16:22:29linuxstbIs there anything else you can work on? Do you have any form of feedback (backlight, LEDs, ...) ?
16:22:45oboI have backlight and button lights
16:23:44 Quit mrkiko ("sorry")
16:24:26obothe View appears to have the same USB controller as earlier generation PP devices, so I could try to track down the GPIO used for that...
16:26:12 Join kugel [0] (n=kugel@rockbox/developer/kugel)
16:26:36oboapart from that I'm open to suggestions
16:26:42 Join evilnick [0] (i=0c140464@gateway/web/freenode/x-a34949799583decc)
16:27:33 Quit crwl (
16:27:33 Quit feisar-- (
16:27:58NJoinfeisar-- [0] (
16:27:58NJoincrwl [0] (
16:29:09 Quit AndyIL (Read error: 110 (Connection timed out))
16:30:43kugelobo: Working on USB for some days might be sensible, it's probably a good idea to take a short break from the lcd wall
16:31:12 Join schrottplatz [0] (n=max@
16:31:18kugel(also you'll have a bit more to show off [if that's needed] for the mid-term evaluation)
16:31:31schrottplatzhas anybody running rockbox on a sansa fuze?
16:31:34gevaertskugel: mid-term evaluation ends today :)
16:31:53kugeltoday already? A bit late then
16:32:06kugelschrottplatz: yes
16:32:58schrottplatzand can you listen to music etc?
16:33:10schrottplatzgreat :)
16:33:25schrottplatzthe lcd of my e200 broke and now i need a new player
16:33:25kugelI've also seen pictures of people playing pokemon on it
16:33:37linuxstbobo: What about ATA/NAND? Do you know if that's similar to the e200?
16:34:37 Nick daurnimator_ is now known as daurnimator (
16:36:15schrottplatzwhere can i see the fiffrence between fuze v1 and v2?
16:37:00obolinuxstb: it has a similar SD<->NAND controller to the e200.
16:37:37linuxstbobo: So that could be interesting to work on as well. My main suggestion was simply to be to try something else, and go back to the lcd later.
16:39:13linuxstbobo: Also, do you have anything to commit? I don't see a View in the build system...
16:39:20kugelschrottplatz: I don't think you can before turning it on and looking for the firmware version
16:39:42obolinuxstb: kugel: okay, thanks. I'll try looking into these new areas.
16:39:43 Join jfc^2 [0] (
16:39:50schrottplatzi will buy one
16:40:38obolinuxstb: no. I was thinking along the lines that since I didn't have any code that produced any output, it wasn't worth adding the target to the build system
16:40:55schrottplatzthx and bb
16:40:59linuxstbobo: That hasn't stopped me in the past...
16:40:59 Part schrottplatz ("o.O")
16:41:34 Join _zic [0] (
16:42:07 Join toffe82 [0] (n=chatzill@
16:43:50obolinuxstb: okay, I can add it in. To begin with I was thinking of adding the PP6100 like the PP5024 - it just includes the PP5020 header. But for main builds some defines will need to be changed to make sure the COP isn't used on the View
16:46:09 Quit bmbl (Connection timed out)
16:48:05CIA-71New commit by 03alle (r21840): Slightly reduce the bin size by using ushort instead of int in arrays
16:49:23 Join Strife1989 [0] (n=Michael@
16:50:12 Quit Strife89 (Nick collision from services.)
16:50:51 Nick Strife1989 is now known as Strife89 (n=Michael@
16:53:38 Join funman [0] (n=fun@rockbox/developer/funman)
16:55:02linuxstbobo: So there's no COP?
16:55:37obolinuxstb: AFAIK the COP is the Nvidia GPU/media accelerator
16:57:04***Saving seen data "./dancer.seen"
16:57:14linuxstbobo: OK, that should be easy to disable in config.h I think
16:59:40 Quit kugel (Nick collision from services.)
16:59:44 Join kugel [0] (n=kugel@
17:02:44 Join fml [0] (n=4fd3cb07@gateway/web/cgi-irc/
17:03:03 Quit teru ("Quit")
17:03:38fmlWhat do the errors for the two last commits mean? Are they caused by the commit or by the build infrastructure?
17:04:14kugelfml: I wonder if 50bytes are really worth added complexity through not using int
17:04:47kugelwhat was the binsize hit anyway? Maybe you're worrying a bit to much
17:05:14fmlkugel: what complexity? Nothing has changed code wise.
17:05:16Learfml: "No space left on device" is clearly a problem with one particular build server.
17:05:33kugelnot in the C code, yes .)
17:06:20kugelI wonder a bit if the pitchscreen needs yet another fixpoint implementation anyway
17:07:21fmlkugel: I don't think it needs
17:09:22kugelthere's some in fixedpoint.c
17:10:15kugelI don't quite understand all the changes, I thought only the UI was questionable?
17:10:46funmanobo: have you seen some code ran by the COP, or you only have informations from documents?
17:22:09funmanobo: since you can control the button light / backlight you can get back binary numbers from the sansa view.
17:22:42funmanFor example check the content of a lcd status register to see if commands are being transmitted at all
17:23:10CIA-71New commit by 03learman (r21841): Import Vorbis seeking improvements from Tremor SVN.
17:23:53obofunman: there are checks/branches in the OF testing if a COP is present. The only docs I have are the nvidia sales info, which describes an ARM1176JZ with a "hardwired media co-processor"
17:24:37funmanIf i'm not mistaken, "media" co-processors have a special instruction set dedicated for multimedia processing
17:24:41thegraeyjust got back home, and I forgot how to include beatbox in my build. its a plugin, but in a subdirectory, so the normal plugin-adding routine doesn't apply?
17:26:02obofunman: that what I assumed, some kind of DSP? There is an extra chunk/mi4 in the firmware partition titled mediaproc.mi4 I'd be happy not to touch it
17:26:30thegraeybtw, I have googled, but not found. and I've tried to add beatbox to SUBDIRS, but that yields an error about no beatbox.make being there...
17:26:57funmanobo: you better shouldn't ;) there is a reason the DSP on some ipods is not used
17:27:06 Quit n00b81 ("Leaving")
17:27:16 Quit Strife89 ("Gotta run some errands. Thanks again, Lambda. :)")
17:27:33funmanthegraey: just look at .make files in other subdirectories
17:28:26obofunman: I didn't think the ipods had a DSP? But I have a 250MHz arm to play with, that should be more than enough.
17:29:24thegraeyfunman: I will, thanks
17:30:53 Join daurnimator_ [0] (
17:35:26 Nick daurnimator_ is now known as daurn (
17:39:36TheSevenfyi, it looks like ipod classic (6g) is extremely similar to the nano 2g
17:39:47TheSevensomebody just reported that ibuggerloader would run on it
17:42:13 Quit daurnimator (Read error: 101 (Network is unreachable))
17:42:26linuxstbTheSeven: You mean it runs fully? i.e. enumerates as a USB device running your code?
17:42:49TheSeventhat's what someone reported
17:43:34linuxstbThat would be good news. The Classic is far more interesting (IMO) than the Nanos - very few companies make hard-disk DAPs any more...
17:44:12TheSevenand it's a currently-sold model
17:44:37TheSevenwe really need to get usb fully working
17:46:23linuxstbIsn't there two revisions of the "6g" though?
17:46:28thegraeyfunman: it's building now, thank you :)
17:49:01 Nick thegraey is now known as graey (
17:50:43Lloreanlinuxstb: Are there really? I thought the 120gb one was just a larger single-platter drive?
17:51:15linuxstbLlorean: Well, lists "Classic" and "Classic 2G". That's my only source for that info.
17:51:23 Quit fml ("CGI:IRC 0.5.9 (2006/06/06)")
17:51:24graeyfunman: now it's complaining about missing, did I do something wrong again? ^^
17:51:30kugelI don't think the hw is different though
17:51:54LearJdGordon: Better check your build server...
17:51:58 Quit robin0800 (Remote closed the connection)
17:52:11JdGordonyeap, done, fixed
17:54:14Lloreanlinuxstb: Ah, interesting, because I don't think apple distinguishes them on the "identify your iPod" page like they do the 5G and the later 5.5G
17:54:49graeyor anyone else for that matter ^^
17:56:00 Quit petur ("work->home")
17:56:05funmangraey: paste the log on some pastebin
17:56:18 Join robin0800 [0] (
17:59:08 Join saratogalab [0] (n=9803c264@gateway/web/cgi-irc/
17:59:37saratogalabamiconn: you cook test file is way faster then mine, I only got ~190% realtime with my 64k file on PP5024
17:59:48graeyfunman: it's all hapiness, until this:
17:59:54graeyfunman: all out of the blue, too
18:01:17linuxstbgraey: ld is (I think) writing that file - do you perhaps have another file already there with that name, but with odd permissions/ownership?
18:01:45graeylinuxstb: in the build-directory?
18:01:48*linuxstb realises that probably doesn't make sense, re-reading the error message...
18:02:09linuxstbDo you even have apps/plugins/beatbox/ in your build dir?
18:02:31graeynope, I clean up my build dir every time, so it's empty to start with. is that wrong?
18:02:43graeyI mean, it builds perfectly without beatbox
18:03:47graeyjust noticed a typo
18:04:03graey$(SBEATBOXBUILDDIR)/beatbox.rock: $(BEATBOX_OBJ) should be $(BEATBOXBUILDDIR)/beatbox.rock: $(BEATBOX_OBJ)
18:04:13graeysorry ^^
18:05:32 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
18:06:41amiconnsaratogalab: It's a 32kbps file, that's why it's faster. But it doesn't change the fact that your iram change works wonders on coldfire
18:09:08saratogalabhow nice of them to put all buffers in one struct that could be easily moved to IRAM
18:09:10amiconnPutting some code into iram would help on PP5002
18:09:33saratogalabASM'ed fixed point functions would help a lot
18:09:47saratogalabthe fixed multiply function they use looks like a mess
18:09:47amiconnThere are also some arrays of constants of which I don't know how often they're used
18:10:19amiconnIf they're used heavily, they should go into iram as well
18:10:38 Quit AlexP ("No Ping reply in 90 seconds.")
18:10:54*amiconn still wonders about the jumping elapsed time on coldfire
18:10:55 Join AlexP [0] (n=alex@rockbox/staff/AlexP)
18:11:09*amiconn has no real idea how he could help debugging that
18:11:20amiconnI tried some things which didn't work
18:12:16 Quit dz (Remote closed the connection)
18:12:20 Join dz [0] (
18:14:02linuxstbamiconn: So it plays fine, but the elapsed time is wrong? What about seeking?
18:15:24 Join Ubuntuxer [0] (
18:23:26 Join bertrik [0] (
18:31:05 Join aaron424 [0] (
18:31:21 Join dash32 [0] (
18:32:03funmanaaron424: i can use mkamsboot fine on my m200v4 firmware "4.01.08-A", md5sum fc9dd6116001b3e6a150b898f1b091f0
18:32:03graeyhas midi changed a lot lately? I get far less buffering issues now on my sansa e250.
18:32:18graeywith latest revision that is, btw
18:32:33 Join JdGordon| [0] (i=ad804a17@gateway/web/freenode/x-ee4d069c3d8f625d)
18:32:56aaron424funman: the .bin I am using was originally named m200p.bin. (I got it from badgers website)
18:34:47funmanaaron424: I used this same file, please check that it downloaded correctly (using md5sum for example)
18:35:24linuxstbaaron424: Also are you using the most recent mkamsboot?
18:35:42aaron424just downloaded it today
18:35:53aaron424from svn I believe
18:36:24funmanaaron424: can you paste the full output of mkamsboot on a pastebin, with a md5sum of your m200p.bin if possible
18:36:39 Quit graey ()
18:37:02aaron424what is a pastebin
18:37:03 Join DarkSpectrum [0] (
18:37:16 Quit DarkSpectrum (Client Quit)
18:37:29 Join DarkSpectrum [0] (
18:37:33funmanaaron424: for example
18:38:01funmanpaste the output of mkamsboot in it, click paste, and give us the url you obtained so we can see what you pasted
18:38:55linuxstbaaron424: I've just tried (downloading m200p.bin from Bagder's site), and it works fine for me as well...
18:39:30funman18:29 < aaron424> [ERR] Whole file checksum failed
18:39:43funmanI suspect an incorrect download?
18:39:51kugelI just had a "sd bank error: -1"
18:40:04aaron424<script src=''></script> ?
18:40:21aaron424dont know how to use it
18:40:41kugelafter clicking paste just post the url from the addressbar here
18:41:03funmanthe error is explicit (and it is not the one you had mentioned), you have to build a m200v4 bootloader first
18:41:05linuxstbaaron424: That's a different error to the one you pasted before...
18:41:09 Nick thegeek_ is now known as thegeek (
18:41:21 Quit Lear ("ChatZilla 0.9.85 [Firefox 3.5/20090624025744]")
18:41:25 Quit thegeek ("( :: NoNameScript 4.2 :: )")
18:41:27mtamiconn: ping
18:41:31aaron424linuxstb: I ran the command again
18:41:37 Join thegeek [0] (
18:41:38aaron424with a fresh .bin
18:42:18linuxstbYou now need a bootloader-m200v4.sansa file in the same directory as mkamsboot and m200a.bin
18:42:54aaron424where/how do I get that
18:43:32linuxstbYou need to compile it yourself.
18:43:54*linuxstb doesn't even know the status of Rockbox on the m200v4...
18:44:06*aaron424 is trying to find out
18:45:36funmanlinuxstb: sporadic (volume related) shutdowns, and playback stopping (like on clip and c200v2)
18:48:21pixelmamt: guess you wanted a coldfire tester, so problem solved?
18:51:08 Join graey [0] (
18:52:02mtpixelma: umm .. I'm not sure it's going to make any difference, but it's a very small modification so we might as well try. :)
18:52:15 Quit dmb (No route to host)
18:52:17bertrikI also remember LCD problems (probably due to too fast delay loops) resulting in the display data to be shifted down
18:53:26bertrikthe m200v4 runs on an AA battery (or was it AAA) so it is different from the other ams sansas (that uses a lithium battery) with respect to power
18:53:32mtpixelma: in apps/codecs/librm/rm.c - line ~553 (pkt->flags & 2) .. change the 2 to 2000 and see if that makes a difference in the elapsed time update ?
18:54:04bertrikthere are some settings that determine how the power is regulated that we haven't explored yet as far as I know
18:54:45bertrikI also noticed that the audio volume on the m200v4 was quite low compared to the other ams sansas
18:57:05***Saving seen data "./dancer.seen"
19:02:42 Quit Ubuntuxer ("Leaving.")
19:03:34 Quit r0b- (Read error: 104 (Connection reset by peer))
19:04:41 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
19:09:35 Quit saratogalab ("CGI:IRC (EOF)")
19:13:36 Quit JdGordon| (Ping timeout: 180 seconds)
19:18:43 Join JdGordon| [0] (n=Miranda@nat/microsoft/x-a9011f40c0660390)
19:20:43 Join _lifeless [0] (n=lifeless@
19:22:41 Join Ubuntuxer [0] (
19:24:55graeyI'm trying to make the metronome tick every first beat of the measure, basically meaning you'll have a 'tick' and a 'tock' sound. However I don't fully understand the sound array, I suppose it's raw sound data like in a wav file, but then as an array of chars? if so, what's the second sound array doing?
19:28:45n1sgraey: there's a patch for that in the tracker somewhere IIRC
19:28:56 Nick efyx_ is now known as efyx (
19:29:06linuxstbgraey: One is wav (for swcodec targets), one is mp3 (for hwcodec targets).
19:29:15graeylinuxstb: thanks :)
19:29:41graeyn1s: does it also allow you to set different types of measures too? like 3/4, 4/4, 6/8 and so on?
19:29:59 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
19:30:02n1sgraey: i don't know
19:32:06graeythanks :)
19:33:12 Quit DarkSpectrum ()
19:34:41 Join benanne [0] (
19:34:48aaron424I cannot make my m250 update the firmware.
19:34:56*n1s wonders why fml reinvented the ARRAYLEN macro in r21839...
19:35:15kugelI wondered that also
19:36:31aaron424I was finally able to make the bootloader with mkamsboot but copying the file over and renaming it and then removing the m200 from the computer. It just refreshes database and goes back to the of
19:38:04 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
19:39:48benanneI know this isn't an iPod support channel, but I've been running rockbox on my iPod nano 1st gen all its life and now it seems to have died... it's not responding to the reset button combination, and when I attach it to my PC it seems to go into a reboot loop (showing the apple logo for a few seconds, then powering down, etc. ad nauseam)
19:40:02benanneI'm having a hard time identifying the problem, but I'm affraid the battery might have died... anybody else ever seen this?
19:41:30evilnickbenanne: Leave it charging for 24hrs or so, and try to hard reset it again
19:42:25benanneevilnick: the problem is that when I attach it to my PC for charging, it goes into this reboot loop... it doesn't seem very healthy to let it do that for 24 hours on end, I even doubt that it actually charges at all in that process. Or should I just go ahead and try that :)
19:42:28mtpixelma: Any luck ?
19:42:43evilnickbenanne: Don't you have a wall-charger?
19:43:33benanneif I could find an externally powered USB hub somewhere, would that work?
19:44:04linuxstbbenanne: You could also try turning on the hold switch, to try and get it to start the Apple firmware. Or force it into disk mode (hold SELECT+PLAY immediately when it comes on)
19:44:28evilnickbenanne: I'm not sure. I was thinking of the kind of adaptor that you plug into the wall that you then plug the USB end of the iPod cable into.
19:44:29benannethose sound like viable ideas, thanks :)
19:44:42benanneevilnick: yeah, I know what you're talking about, don't have one of those unfortunately :(
19:45:05evilnickbenanne: Get an iPhone and you get one for free (!)
19:45:14benannehehe :p
19:45:16*gevaerts slaps evilnick
19:45:50benanneentering disk mode seems to have worked :)
19:46:06benannecouldn't boot into the apple firmware either though, I guess rockbox isn't to blame
19:46:06 Quit antil33t ()
19:46:11benannealthough I didn't expect it would be
19:46:13 Join gkahla [0] (
19:48:38pixelmadid JdGordon already fix his server?
19:48:59JdGordon|its still complaining?
19:49:31pixelmaprobably not, just catching up with what happened today :)
19:50:29pixelmamt: sorry, phone call. Going to try soonish, build is already complete
19:50:48JdGordon|Bagder: something to make the new system more complicated... errors like out of disk and missing compilers and stuff will be hidden in the new system.. any chance of having it email the client owner if it happens?
19:51:38 Join Thundercloud [0] (
19:52:09mtpixelma: Great.. Thanks :)
19:52:10 Part gkahla ("Leaving")
19:52:18FlynDicefunman: Got any remarks on this: I think the 1<< 23 is covered by 0x00FF8000
19:54:37funmanFlynDice: but for non sdv2 cards, only the bit 23 is set, not all the bits of supported voltage
19:55:17FlynDicearen't we just telling it what our hardware supports?
19:57:52FlynDiceI'll go find a version 1 sd spec...
19:57:56pixelmamt: hmm, the time looks sensible now (though I think it's wrong, need to compare with vlc). But playback is broken - starts playing ok for a bit less than a second, I guess and the it's stuck playing one sample, sometimes outburst into more noise, or so and current playing time doesn't advance (stays at 0:00)
19:58:12pixelmahaven't tried plain svn though
19:58:14funmanFlynDice: i think version 2 covers all cases
19:59:04funmani think 1<<23 is what the of uses, lemme check
19:59:38FlynDicefunman: so do we only support 3.5-3.6 on a v1 sd card, is that the issue?
20:00:39benanneright, I'll leave it charging for a while and then see if I can get it to boot rockbox again
20:00:40mtpixelma: Does your source copy have any modifications related to RM ?
20:00:45benanneevilnick, linuxstb, thanks for your help :) bye
20:00:54 Part benanne ("kbai")
20:01:32funmanFlynDice: i don't know
20:02:02funmanI just get the supported voltages from the OF since we don't have the details of how the SD controler is powered
20:02:20*linuxstb wants to create a subdir in utils/ for tools for hacking new ipods, and wonders what to call that directory...
20:02:44 Join antil33t [0] (n=Mudkips@
20:03:21pixelmamt: it had the one change you told me, kugel's pluginlib actions patch and a few changes to the manual by me
20:03:25 Join DarkDefender [0] (
20:03:29 Quit DarkDefender (Read error: 104 (Connection reset by peer))
20:04:16funman0x5806 in my fuze firmware (v1.1.11 i think)
20:04:35 Part Ubuntuxer
20:04:40mtpixelma: Fine. Nothing of those could be affecting playback .. I'll go back and think. :/
20:04:41 Quit flydutch ("/* empty */")
20:05:31amiconnlinuxstb: I didn't try seeking. test_codec has the same display problem - the number of decoded milliseconds is jumping around wildly. But apart from that it's working properly
20:05:35amiconnmt: pong
20:05:48mtlinuxstb: new_ipods ? :)
20:05:52pixelmamt: I just tried without that one line change in apps/codecs/librm/rm.c and it plays fine but playing times are junpy
20:06:21linuxstbmt: I may end up just using that... Or maybe simply "ipod"..
20:06:58mtamiconn: nevermind, I just had a stupid idea that doesn't work !
20:07:47linuxstbamiconn: Do you have a WPS that shows the bitrate?
20:08:07pixelmaI do
20:08:20amiconnNo - I used the track info screen
20:08:23mtpixelma: You mean jumpy like giving out-of-range values ?
20:08:34linuxstbamiconn: OK, so that's being read correctly?
20:09:33pixelmayes, really bogus times jumping from -345464:-23 to 567:34 etc
20:10:30 Join mitk [0] (n=594df737@gateway/web/cgi-irc/
20:10:50pixelmaonly seems to be the current playing time. This thing makes the progressbar jump from being completely full to completely empty and vice versa
20:10:51mtlinuxstb: I was going to suggest just 'ipods', but thought 'new_ipods' is better since it clearly states that those tools are just for the new ipods.
20:11:34pixelmathe track length seems to be displayed correctly
20:12:05mtThe problem has to be with rmctx.audiotimestamp. It can't be with rmctx.bit_rate since it's displayed correctly and if there were a problem with 'i' there would have been seg faults.. :/
20:13:25 Join flydutch [0] (
20:14:22pixelmathe mentioned times were just examples, it's completely different all the time
20:14:33 Quit funman ("free(random());")
20:14:50 Join itcheg [0] (i=4117734b@gateway/web/freenode/x-6f8dd8e38c172de8)
20:14:56aaron424I am having trouble installing rockbox on my m250. I have put the rockbox dir in the root and compiled the bootloader (m200a.bin)
20:15:19aaron424 but when I put the .bin on my device, it does not update
20:15:32mtpixelma: Yeah I know. Thanks for the clarification.
20:16:55mitkHi. Is there anyone who have some time to take a look and enough power to commit FS #10235? It is pacbox patch for Fuze.
20:17:51linuxstbmt: I've remembered I also have one utility for old ipods, which I could commit there, so I'll just use "ipods" as the top-level dirname...
20:18:21 Nick JdGordon| is now known as JdGordon|| (n=Miranda@nat/microsoft/x-a9011f40c0660390)
20:18:44 Nick JdGordon|| is now known as JdGordon| (n=Miranda@nat/microsoft/x-a9011f40c0660390)
20:19:09kugelmitk: what's the HOME button change?
20:19:18mtamiconn: Is it possible to log the values of some parameters on-target ?
20:19:30 Quit aaron424 ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060308]")
20:20:01mitkIt changes menu access method.
20:20:32kugelyea, but why? long home is used in all other plugins
20:23:05mitkIf I understand correctly originally it was combo BUTTON_HOME|BUTTON_REPEAT. I can change t to long HOME but pacbox do not need many buttons so simple HOME can be used.
20:24:53amiconnmt: Probably, using logf(). But I didn't use that for ages, and I'm not sure whether I could use it from codecs
20:25:02 Quit stoffel (Read error: 104 (Connection reset by peer))
20:25:22amiconnAlso, what would I find? A jumping value... but why does it jump
20:25:59mtamiconn: It could be used from codecs, I was just asking because I wasn't sure it worked with your targets.
20:26:18amiconnlogf() works for all targets
20:27:16mtamiconn: No, I'm almost pretty sure the problem is with rmctx.audiotimestamp, which makes the possibility of an error in reading the value of pkt->flags high.
20:28:39mtamiconn: So if you have the time, could you help me by displaying this value ?
20:31:44CIA-71New commit by 03dave (r21842): First commit of "bin2note" utility for exploiting the Notes buffer overflow on the 2nd generation Nano.
20:34:13CIA-71New commit by 03kugel (r21843): FS #10235 - "(fuze) pacbox keymap change" by Ralph Soto. ...
20:34:59kugelmitk: alright, done
20:35:02mitkkugel: Thank you. Bye
20:36:37 Quit mitk ("CGI:IRC")
20:37:13amiconnmt: There is a difference in the test function rm_get_packet_fd() and rm_get_packet(). The former only gets packet_group and flags if version == 0, the latter always does
20:37:24amiconnIs this correct?
20:39:09mtamiconn: Yes .. it's not exactly the same but that is no problem. (I should modify _fd later though)
20:39:13CIA-71New commit by 03dave (r21844): Add flashsplit utility, previously hosted on the wiki at
20:39:49amiconnAlso, I wonder why you're reading 'unknown' at all if it's not used
20:40:03 Join einhirn [0] (
20:40:18amiconnIn the fd case you need to advance the file pointer of course, but in the in-memory case this is unnecessary
20:46:19mtThis should be removed yes.
20:47:35 Quit einhirn (Read error: 104 (Connection reset by peer))
20:48:54 Quit flydutch (Read error: 60 (Operation timed out))
20:48:56graeyI'm considering using midi sounds for the metronome, is this OK or doesn't MIDI work on all targets the way PCM/MP3 playback does?
20:49:02*kugel thinks statusbar should be a theme setting
20:50:01linuxstbgraey: midi takes a lot of CPU time, plus a lot of RAM. It's not an efficient choice...
20:50:19linuxstbPlus it's not easy to use codecs from plugins
20:50:27JdGordon|kugel: it is
20:50:54JdGordon|actually... they are
20:50:56graeylinuxstb: I thought it'd be nicely doable since it's also what beatbox seems to do
20:51:23graeybut if it's not efficient it's not the right choice indeed
20:51:33linuxstbgraey: beatbox actually uses the midi code, or does it just have it's own samples that it plays back?
20:51:43evilnickgraey: How does the current Metronome do it?
20:52:00linuxstbevilnick: Just a single PCM sample (or mp3 for Archos)
20:52:05graeylinuxstb: it does reference them in it's sources-file, so I assume it uses midi yes
20:52:22kugelJdGordon|: they're not under the theme settings menu though
20:52:33JdGordon|so move them
20:52:39JdGordon|doesnt mean they arnt theme settings :)
20:52:44 Join bubsy [0] (i=Bubsy@
20:54:18 Quit bmbl ("Bye!")
20:54:22graeylinuxstb: how exactly do I 'convert' a wave/mp3 file to char/short array like used in the metronome?
20:54:52kugelslick_letitblue looks nice with customlist
20:55:59linuxstbgraey: You would need to write a program to do it
20:56:28graeylinuxstb: a hex-editor wouldn't do?
20:56:58linuxstbgraey: Maybe. But basically you'll need to use your initiative - there are no existing tools in Rockbox to do it.
20:57:10***Saving seen data "./dancer.seen"
20:57:34gevaertsgraey: have a look at tools/bin2c
20:57:36graeylinuxstb: yeah, I thought so. I was hoping the original writer was on irc and reading along :P
20:57:45graeygevaerts: thanks, I will :)
20:58:11gevaertsYou'll need a raw audio file of course, not a wav file
20:58:32mtamiconn: Could you change line 160 in apps/codecs/cook.c to : ci->set_elapsed((rmctx.frame_number-2)*(1000*8*sps/rmctx.bit_rate)); and try again ?
20:58:33bertrikwhat targets have a built-in speak for keyclicks?
21:00:20kugelJdGordon|: do you have an idea why the list in the playlist viewer is so slow?
21:00:53bertrikI wonder how to drive the piezo speaker in the meizu, should i just put a DC voltage on it to produce a click? or maybe some kind of square wave? should I try to drive it at its resonant frequency?
21:01:19graeygevaerts: raw audio file, I think sox converts waves to raw audio files. then I use bin2c and start coding?
21:01:21bertrikunfortunately it's not connected to a timer PWM output
21:01:28gevaertsgraey: yes
21:01:49graeyok, thanks!
21:03:19amiconnmt: The flags are obviously read correctly (mostly 0 and sometimes 2), but the timestamp is jumping erratically
21:03:39amiconnThis is in rm.c: rm_get_packet()
21:05:22JdGordon|kugel: slow how?
21:05:29mtamiconn: I see, thanks a lot. Is this also with the last modification in cook.c ?
21:05:42amiconnI didn't try that
21:05:51*amiconn has a suspicion
21:06:02kugelJdGordon|: scrolling it is noticeably slower than a normal list (possibly only really noticeably with a scrollwheel)
21:06:25JdGordon|just had a quick look on the clip and it seems fine
21:06:44mtamiconn: of what ?
21:06:45 Quit martian67 (Connection timed out)
21:08:30kugeli haven't actually noticed it before i used the fuze
21:08:44 Join flydutch [0] (
21:09:53amiconnmt: I found the bug. It was indeed an endianess problem, in rm.c
21:10:06 Join saratoga [0] (i=9803c6dd@rockbox/developer/saratoga)
21:10:41amiconnget_uint16be() and get_uint32be() are inherently endian agnostic due to the way they work. But there are two versions - the one used for big endian was broken
21:11:37*mt slaps forehead :/
21:11:48*linuxstb slaps mt ;)
21:12:01amiconnmt: I'll commit the fix in a few minutes
21:12:17linuxstbDon't we have those functions in a central place?
21:12:22mtamiconn: Thanks a lot !
21:12:39mtlinuxstb: They are a bit different iirc .. I'll check.
21:12:47pixelmabeware of forehead
21:13:05*linuxstb just remembers implementing them 100s of times...
21:13:22amiconnOn coldfire we could in fact use an optimisation. Since coldfire is able to do unaligned accesses, we could simply cast pointers and read 16 or 32 bits at once. The fact that RM is obviously big endian helps here
21:14:12saratogamt: are you interested in further optimization for cook?
21:14:22saratogai have some ideas
21:14:45saratogaalthough i understand if you have other priorities
21:17:24mtsaratoga: If it's more important than implementing aac-in-rm for now then I have no problem.
21:17:44CIA-71New commit by 03dave (r21845): Add some Makefile rules to demonstrate assembling, linking, converting to binary file and finally converting to a notes .htm file.
21:17:54saratogamt: both are worth having, I would work on whatever is most interesting to you first
21:19:09CIA-71New commit by 03amiconn (r21846): Fix cook on big endian targets. get_uint*be() is already endian agnostic due to reading ...
21:19:53mtlinuxstb: alac also defines those functions.
21:20:49mtsaratoga: Optimising cook would probably take less time than adding support for aac, right ?
21:21:31mtCan't we have multiline commit messages here ?
21:23:23 Join Zagor [242] (n=bjst@rockbox/developer/Zagor)
21:24:12gevaertsZagor: there's a slight bug in the build script : if it kills some gcc processes, the gcc tempfiles stay. Probably best solved by using our own $TMPDIR
21:24:50saratogamt: yes I think it would be fairly easy, I'll probably have a go at it one of these days once I have some more tiem
21:25:01gevaertsAlso, I think the scores are unbalanced : bootloaders are seriously overvalued. I'm now generating a set of new and improved scores
21:25:08Zagorgevaerts: we should probably use normal kill instead of -9, to allow cleanup code to run
21:25:24r4v5kill and then kill -2 if it's still running
21:25:30mtsaratoga: Let's work on that first then.
21:26:17J-23Unable to get setup-2.ini from <>
21:26:37*linuxstb would prefer to see more codecs - optimisation can be done by anyone at any time, but few people will add a completely new codec...
21:27:43 Quit timc (Remote closed the connection)
21:28:19gevaertsZagor: I'm measuring total used CPU time (using /usr/bin/time) for non-ccache builds. If those aren't good enough, nothing is
21:29:08Zagorgevaerts: sounds good. feel free to add the new values to the builds file when you are done
21:29:10mtlinuxstb: Isn't aac already supported? (i.e, it's not completely new ;) )
21:29:21gevaertsOK. Give me a few hours :)
21:29:26 Join advcomp2019_ [0] (n=advcomp2@unaffiliated/advcomp2019)
21:29:32 Join tomcat_ha [0] (
21:30:45linuxstbmt: True... ATRAC then! ;)
21:31:10amiconnmt, linuxstb: Using the suggested get_uint*be() optimisation works. Coldfire cook.codec becomes 60 bytes smaller. No measurable speedup (probably because these functions aren't used often)
21:31:30saratogaAAC in rm would be nice, our current AAC in MP4 support is buggy due to bad parser
21:31:34 Quit advcomp2019 (Nick collision from services.)
21:31:36saratogaso the rm version wouldn't be redudant
21:31:41 Nick advcomp2019_ is now known as advcomp2019 (n=advcomp2@unaffiliated/advcomp2019)
21:31:42saratogaand of course aac in rm isn't entirely uncommon
21:32:41Unhelpfulsaratoga: that sounds like a call for separate "codecs" for the containers. :)
21:33:42amiconnlinuxstb: Do you think such a change is worth being committed? It's an additional #ifdef with little effect...
21:34:39*mt still wishes we could have multi-line CIA commit messages.
21:34:58saratogaUnhelpful: i think the plan is to just link the codec more then once: aac-in-mp4.codec, aac-in-rm.codec, etc
21:35:25saratogamt: they're still multiline on the website and in svn, so its not a big loss i think
21:35:38 Quit FOAD ("I'll be back")
21:35:43Unhelpfulsaratoga: that at *least* suggests we want to have a somewhat standard interface between codec and container handler
21:35:51linuxstbamiconn: I've no real view either way. As you say, it's an additional #ifdef (bad), for a tiny improvement...
21:35:54mtsaratoga: Still, it would be nice. :)
21:36:23linuxstbmt: Not when someone commits a 20-line message...
21:36:37mtAh, good point.
21:36:44linuxstbA url would be nice though...
21:37:11*gevaerts does get a URL, thanks to some magic :)
21:37:13saratogaa 20-line message in the logs would be fine i think, they're so rare
21:37:21tomcat_haquestion: how can i do whats described in the last post of this thread: ?
21:37:35tomcat_hado i need to edit the bootloader?
21:38:18linuxstbtomcat_ha: You need to edit the Rockbox source code (text files), and then compile a bootloader (and Rockbox itself) from that modified source code.
21:38:38 Join FOAD [0] (
21:38:54kugellinuxstb: my ruby script appends the url :)
21:39:27kugelalthough it would be nice if it came from the bot in the first place indeed
21:40:13tomcat_hawhere can i find the source code?
21:40:14mtsaratoga: What about working on both the optimisation and the new codec in parallel, with main focus being on the new codec ?
21:40:58linuxstbtomcat_ha: Start here - - specifically the "HowToCompile" or "SimpleGuideToCompiling" pages.
21:41:57JdGordon|we have a tag for the wps for when the volume is being changed yeah?
21:42:07saratogamt: that sounds fine
21:42:14LloreanJdGordon|: Weren't you the one who added it?
21:42:24JdGordon|cant remember
21:42:46JdGordon|I want to change the clips cabbie to show the volume as a number when its changing in the wps... any objections?
21:42:54markunmt: which new codec? Are you going to add AC3 support?
21:43:07markun(as a real audio codec I mean)
21:43:13saratogaregarding optimization, the two fixed point mulitply functions in cook_fixpoint.h look ugly to me
21:43:22LloreanJdGordon|: We don't do that on any other cabbie, do we?
21:43:28LloreanIt really shouldn't behave uniquely.
21:43:34saratogaand the call to av_clip in output_math isn't needed since rockbox can take 32 bit output from the codec
21:43:43JdGordon|no, but because on the clip its so small it would be a great addition
21:43:51mtmarkun: No, aac since it's the next most common codec after cook. (according to the ffmpeg devs)
21:43:53LloreanJdGordon|: I don't understand what its size has to do with it.
21:43:53 Join timc [0] (
21:44:05JdGordon|oh stop being difficult :)
21:44:08LloreanThere's no numeric volume indicator in the WPS on the Gigabeat and that's the largest screen we've got.
21:44:12saratogaLlorean: I don't think we should try and make Cabbie look exactly the same on every target if the targets are fairly different
21:44:14LloreanWell, largest supported.
21:44:20Lloreansaratoga: It's an issue of behaviour, not looks.
21:44:25markunmt: I only have cook and ac3 files
21:44:37saratogaIMO cabbie should look good and be functional on all targets, but maybe not the same on each
21:44:41markunbut that can come laer
21:44:47JdGordon|the difference is the size of the bar thing is so small its hard to tell if its moved much or at all
21:44:49Lloreansaratoga: If we want it to show numeric volume when volume is changing, there's no reason not to do it on all targets. And no reason *to* do it on the clip if we're not doing it elsewhere.
21:44:52saratogaactually I always wanted cabbie on the ipods to have a different color scheme then on black targets . . .
21:45:02AlexPJdGordon: I have no objection to it displaying numeric values when changing on all targets :)
21:45:21linuxstbsaratoga: What do you mean by "black targets"? ipods come in black...
21:45:23Lloreansaratoga: This isn't an optimizing it to a specific target issue though.
21:45:38mtmarkun: Are the ac3 files old ?
21:45:38saratogado we support any black ipods?
21:45:44LloreanAlexP: Agreed, I like seeing numeric values. I just think it'd be rather wrong to just change it on the clip
21:45:47linuxstbsaratoga: Nano, 5g, 5.5g
21:45:52*JdGordon| changes his suggestion to showing the number on any targets that people can be bothered fixing...
21:45:57saratogaah ok
21:45:58Lloreansaratoga: One of the first iPods to hear music in Rockbox was black.
21:46:05saratogathat'd be hard to detect at runtime
21:46:06markunmt: yes, quite old
21:46:12Lloreansaratoga: Actually, no.
21:46:21Lloreansaratoga: If the OF is there, at least, you can find out from its data what color the iPod is.
21:46:37Lloreanthe OF reports it in the inquiry to iTunes, so it must know somewhere
21:46:50linuxstbAh yes, so itunes can show the correct icon ;)
21:47:02*Llorean thinks it'd actually be rather interesting to have Cabbie default to matching the iPod's color, with two versions available to switch between manually later.
21:47:38mtmarkun: That's why then. Recently it's been rarely used as a codec in rm. (again, according to ffmpeg devs :) )
21:47:53saratoganeat idea
21:48:00 Join thegraey [0] (
21:48:07linuxstbmt: BTW, did you ever get a copy of "realproducer" so you can create your own test files?
21:48:22saratogachanging cabbie subtlely on different targets to make it blend in with the hardware more has been on my TODO list for ages but I never got around to it
21:48:36mtlinuxstb: Yes.
21:50:05 Join bluebrother [0] (n=dom@rockbox/developer/bluebrother)
21:50:36thegraeyjust got back at my dev machine. is there any documentation on what keymaps apply to what devices?
21:50:38bertrikOne thing I find slightly weird is that in the default theme cabbiev2 the icons are on the bottom, while they are on top in the main menu
21:50:47tomcat_hathe manual says i have to type in ../tools/configure in my new made folder to see all possible devices
21:50:59JdGordon|bertrik: you mean the statusbar?
21:51:01tomcat_hahowever my terminal says it cant find those folders
21:51:29JdGordon|it can be put on the bottom now
21:51:38JdGordon|takes a bit of getting used to though
21:53:16mtlinuxstb, saratoga: Since aac has already been ported, should it go through the same stages cook went through ? more specifically is a standalone test program for aac-in-rm needed at first ?
21:54:00linuxstbmt: That's up to you - the standalone test programs are just to help you develop it.
21:54:27kugelJdGordon|: I already use it at the bottom for cabbie since it's committed
21:54:38 Join mortti [0] (
21:55:26saratogamt: do you know how AAC is stored in RM? if its pretty simple to decoder I think a stand alone test program might be unneeded
21:55:40tomcat_hasimpleguidetocompiling is made with windows users in mind?
21:55:42saratoga"pretty simple to decode"
21:55:48tomcat_hai am running ubuntu here atm :B
21:56:03bertrikdo we have audible keyclicks for any rockbox targets yet?
21:56:18saratogatomcat_ha: NOTE: This guide is for Windows users only! Linux users can read LinuxSimpleGuideToCompiling.
21:56:47tomcat_haim smart am i not 8)
21:56:53saratogabut the steps should be the same aside from the parts about setting up cgywin/vmware so it probably doesn't matter
21:57:06AlexPbertrik: There is the button click on (some?) ipods, but I seem to remember it causing issues at some point
21:57:20AlexPDid that use the ipod piezo?
21:57:38JdGordon|there is a very old patch for the piezo which never got in
21:57:42mtsaratoga: Yes I know, that's why I asked. :)
21:59:07 Join petur [50] (n=petur@rockbox/developer/petur)
21:59:09 Quit LambdaCalculus37 ()
21:59:52thegraeyis there a function to check if an array contains a value or should I just loop through it?
22:00:16gevaertsdepends. If it's sorted, there's bsearch()
22:00:21gevaerts(or there should be)
22:00:40 Join Ubuntuxer [0] (
22:02:49linuxstbthegraey: Also, what kind of array?
22:03:59thegraeyarray of ints
22:04:01linuxstbmt: If you try and write a standalone program, you'll also need to make Rockbox's aac decoder compile standalone...
22:04:46bertrikAlexP, JdGordon I'll see if I can find that patch
22:06:28saratogalinuxstb: don't we have that patch that compiles any codec stand alone now?
22:06:50linuxstbsaratoga: I'm not sure if it's any codec, but yes, that should help.
22:07:17saratogaalso have you looked at Yoshihisa Uchida's recent storm of codec patches?
22:07:31obolinuxstb: would you be able to give FS #10436 a quick sanity check if you have a few moments?
22:07:51*linuxstb heads off to flyspray...
22:08:43linuxstbsaratoga: Not yet, I'll have a look now...
22:08:56saratogai think hes nearly doubled the number of formats we can play, provided you put them in .WAV or .AIFF
22:09:05linuxstbobo: Yet another ARM variant? :( will never finish...
22:09:21oboI'm afraid so...
22:10:43linuxstbIs the View's keypad also different to all the others?
22:11:05saratogalinuxstb: I think thats the same arm core as the beast, sans some of the addons . . .
22:11:41obolinuxstb: I think it's pretty much the same as an e200, but I don't actually have one of those... I'll check the e200 manual
22:12:05linuxstbobo: You would save a lot of work if you could pretend to have the e200 keypad...
22:12:27linuxstbThere are also images in uisimulator/sdl/
22:12:48saratogayeah wikipedia says its the same core as the imx31 except with added DRM hardware, so they should be able to use the same arm compiler i think
22:13:37obolinuxstb: it's different. the view has a combined power/hold button on the side, a home button, and 4-way clickwheel.
22:13:39linuxstbThat could save people needing to re-run
22:14:11linuxstbSo this nvidia is a massively souped-up PortalPlayer?
22:15:33oboI'm not sure how massive the increase is...
22:15:56linuxstbWell, it seems at least to be armv6 compared to armv4...
22:16:39oboyes. I may not have changed enough defines to handle that yet.
22:16:41linuxstbBut yes, it looks like you could simply use the gigabeat S's compiler.
22:16:54obobut the S has the floating point unit?
22:17:22saratogaso does the View apparently
22:17:43oboit's a jz not a jfz?
22:18:41saratogaboth have vfp
22:18:49saratogai don't know which it is though
22:18:56linuxstbAlthough what does the "KZ" mean in "ARMv6KZ" ?
22:19:39TorneK is "a few extra bits"
22:19:46saratogaoh perhaps not
22:19:52Torneeverything except ARM1136 is v6K
22:20:03Tornethings liek CLREX and other useful bits that they totally forgot in arm1136
22:20:09Tornethe Z is TrustZone
22:20:14saratoga"The only functional difference between the ARM1176JZ-S and ARM1176JZF-S processor is that the ARM1176JZF-S processor includes a Vector Floating-Point (VFP) coprocessor."
22:20:28linuxstbTorne: Ah, so that would mean that there are extra instructions on the 1176 vs the 1136?
22:20:35Torneonly a few though
22:20:42Tornethe 1136 is a dud
22:21:08saratogaso it doesn't have vfp afterall
22:21:11 Quit Ubuntuxer ("Leaving.")
22:21:50Torneif you can run code on it, just check
22:22:06Torneturn on cp10 and cp11 in CAR and read FPSID
22:22:12Torneif you die with an undef abort then it doesn't :)
22:23:23saratogawe don't actually emit vfp ops anyway on the beast, correct?
22:23:49Torneno idea
22:24:15CIA-71New commit by 03gevaerts (r21847): Replace scores by more carefully calculated ones. Now the user+system CPU time (as reported by /usr/bin/time) on an Intel L5410 are used, as opposed ...
22:25:16linuxstbobo: The other "unsupported" targets don't have that echo in configure do they?
22:25:23linuxstb(it's a nice idea though...)
22:25:31 Quit bertrik (Read error: 54 (Connection reset by peer))
22:25:36obolinuxstb: the Clipv2 does
22:25:44gevaertsZagor: the heaviest build now has score 26825. Maybe 30000 is a good value for the slow client cutoff now?
22:26:12linuxstbobo: You forgot to change "e200" in your config-view.h file
22:26:22gevaertsThat's about 47 times heavier than the lightest build
22:26:23obooops :)
22:27:17 Quit mt (Remote closed the connection)
22:27:39linuxstbobo: But it looks fine to me (including the change to the multilibs patch - it seems it needs a new gcc)
22:28:11JdGordon|gevaerts: thats too high... that means it could still try doing a single massive build
22:28:25JdGordon|got a listing of all the builds and their scores?
22:28:42obolinuxstb: I've updated the config file locally. Thanks for the review.
22:28:51amiconnNot yet. But I think the beast does have vfp, as it's an ARM1136JF-S
22:29:02gevaertsJdGordon|: no. Lower than the cutoff and it will get bootloaders preferably
22:29:36 Join bertrik [0] (
22:30:33gevaertsso the cutoff needs to be higher than the heaviest build
22:30:44linuxstbsaratoga: Hmm, looking at FS #10422, Uchida's actively removed comments saying what codecs are supported...
22:31:07oboZagor: could you copy the multilibs patch from FS #10436 to please?
22:32:03 Join matsl [0] (
22:32:52linuxstbsaratoga: Hmm, it looks nice on first glance...
22:33:25*JdGordon| is getting confused
22:33:30*linuxstb is now confused by the subsequent patches...
22:34:34kugelgevaerts: I diN't understand the logic
22:35:15 Quit bertrik (Read error: 104 (Connection reset by peer))
22:35:16gevaertsany specific bit?
22:35:22kugelthe heaviest build is 27K. And slow clients are those with <30K. Doesn't that mean slow clients are qualified for the heaviest build?
22:35:42Zagorobo: does it replace a file or is it a new one?
22:35:56oboZagor: it's new
22:36:18kugelI thought maybe 2x the heaviest bootloader would be a good cutoff
22:36:19 Quit matsl (Remote closed the connection)
22:36:24gevaertskugel: all clients are qualified for all builds. What "slow" means is you start of with light builds and get heavier builds as the round progresses instead of starting of with heavy builds
22:36:43gevaerts2x the heaviest bootloader would mean we get no slow clients
22:37:09 Join bertrik [0] (
22:37:24kugelif a client can only manage bootloaders, it will never have an avarage score of 2x the heaviest bootloader
22:37:25 Join B4gder [241] (n=daniel@rockbox/developer/bagder)
22:37:53Zagorkugel: yes it will. average score is _per round_, not per target
22:38:13Unhelpfulgcc says we need to build for the "softfp" float ABI for VFP support - the "hard" float ABI is not supported on it
22:38:27kugelI thought the score is "avarage of the sum of the build-scores"
22:38:43Zagorobo: can you suggest a name for it? specifically how does it differ from "rockbox-multilibs-arm-elf-gcc-4.0.3.diff"?
22:38:53amiconnUnhelpful: On eabi? That's at least how I understand it
22:38:56gevaertskugel: it is. The sum of the build-scores is the round-score
22:39:03amiconnOld abi doesn't seem to support vfp at all
22:39:16oboZagor: the current file has a _2 suffix, this one has a _3?
22:39:35Unhelpfulamiconn: the gcc manual says that for the -mfloat-abi option without any mention of EABI.
22:39:44Zagorobo: ok
22:39:50linuxstbWasn't there talk about putting that patch in svn, and changing to get it via viewvc?
22:39:58Zagorobo: done
22:40:03oboZagor: thanks
22:40:07linuxstbAlthough I guess it updates rarely, so....
22:40:17kugelgevaerts: I mean the client score
22:40:34B4gderlinuxstb: or simply just have the patches in a subdir for
22:40:34gevaertskugel: yes
22:40:41amiconnUnhelpful: Well, as I understand it, old abi would support soft float (subroutines) and hard float using hardcoded fpa instructions, but not vfp
22:40:56linuxstbB4gder: D'oh... Of course, lives in a SVN checkout...
22:41:05kugelgevaerts: and if we avarage this over the number of builds done?
22:41:07amiconnAnd eabi *should* use vfp if it's compiled as soft-float
22:41:17Zagorkugel: the server keeps the score each client manages for each round. then the average score per round is used to see if it is a fast or a slow client.
22:41:23Unhelpfulbasically there's a different calling convention for floats with the soft float-abi. softfp == "use soft float calling convention, but allow hardware fp instructions"
22:41:31gevaertskugel: then you get the average build weight, unrelated to client performance
22:42:01gevaertsZagor: is it the total average, or the average of say the last 15 rounds that client participated in?
22:42:21Zagorgevaerts: currently it is the total average
22:42:21 Quit pixelma (Nick collision from services.)
22:42:22 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
22:42:39 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
22:42:46linuxstbB4gder: Are you #rockbox-gsoc?
22:42:49gevaertsZagor: maybe clear the numbers when you get the new score values in service?
22:42:51AlexPZagor: And if a client did no builds for that round, it gets 0?
22:42:57Zagorgevaerts: yes
22:43:01ZagorAlexP: yes
22:43:06kugelgevaerts: ah yes, it won't work then
22:43:18ZagorAlexP: which means it is more likely to get easy builds the next round
22:43:34 Quit amiconn (Nick collision from services.)
22:43:37 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
22:43:56 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
22:44:09Unhelpfulamiconn: i also did get a working toolchain for long-call stubs without EABI. oddly that still breaks one or two things, somehow, like vorbis.
22:44:32gevaertsZagor: I guess a total average is OK, as long as people pick new names if they switch to different hardware, or if they ping you about it
22:45:20kugelUnhelpful: I thought ogg has many assembly, it probably breaks there
22:45:51Unhelpfulkugel: this was without the struct changes, though - basically gcc 4.0.3 and the relocation changes.
22:47:23kugelUnhelpful: can't it still break there? codecs have a mix-up of IRAM and DRAM stuff, I'm thinking only the long-call stuff might affect it already
22:48:22 Join BdN3504 [0] (n=5ce22528@gateway/web/cgi-irc/
22:48:22Unhelpfulkugel: i don't really see how... the long-call stuff amounts to it emitting short-call instructions everywhere, and if it finds one of them out of range at link time, it generates a one-instruction stub that jumps to the proper address.
22:48:37Unhelpfulthere's no calling convention or structure packing change involved.
22:48:54saratogawhich codecs fail? just vorbis or others too?
22:49:27Unhelpfulsaratoga: vorbis was all that i found. it silently skipped tracks.
22:49:52Unhelpfulfull eabi breaks more things in a rather more dramatic fashion, but that's not surprising.
22:49:59saratogaUnhelpful: undef arm for that codec and see if it starts working
22:50:11saratogaso that it reverts to plain c code
22:50:12kugelWe definitely need a ring-like album list in PF
22:50:19BdN3504hey, i just got myself a 16 gb microsd card for my sansa and must say this is really awessome. Is the sansa going to work with any card you throw at it, or has anyone encountered some incompatibilities over time?
22:50:21amiconnUnhelpful: Why are there two different relocation types, btw?
22:50:24kugeli.e. going from the beginning to the end
22:51:28BdN3504i thought to maybe set up a wiki page like the one for the support cf cards for modded players. but if you guys tell me that the sansa will accept any card, then that page would be obsolete.
22:51:50kugelUnhelpful: would it be hard to do?
22:51:56AlexPBdN3504: It should be fine
22:52:12BdN3504to set up a wiki page?
22:52:31Unhelpfulamiconn: i'm not really completely sure, i believe the eabi objects have relocation types that specify exactly what is done with the address reference, while the pre-eabi R_ARM_PC24 reloc can represent several things, so that the linker needs to decode the instruction to figure out what happens
22:52:33saratogai don't think anyone has ever found a card that it didn't work with
22:52:40bluebrotherOMG. We should delete the users ML.
22:52:57 Quit BdN3504 (Client Quit)
22:53:14 Join BdN3504 [0] (n=5ce22528@gateway/web/cgi-irc/
22:53:26kugelUnhelpful: did you wipe the entire build dir? I was having problems with not doing that
22:53:32BdN3504hooray for rockbox!
22:53:36AlexPBdN3504: No, I meant all cards should be fine
22:53:37 Quit BdN3504 (Client Quit)
22:53:55 Join ej0rge [0] (
22:53:57kugelI.e. my code was perfectly fine, but codecs didn't work because I didn't think of rm -rf'ing the build dir
22:54:06Unhelpfuli'll try again, but i can't this very moment :)
22:54:07kugel(some sort of linker problem, I don't know)
22:54:28JdGordon|bluebrother: haha yes we should.. what is it this time though?
22:54:41 Quit bubsy (Read error: 60 (Operation timed out))
22:54:46kugelDon't, they'll come on the -dev list then
22:54:51linuxstbJdGordon|: The same as always...
22:54:52gevaertsJdGordon|: same as last time. Someone revived it
22:55:06 Quit B4gder ("It is time to say moo")
22:55:22bluebrotherJdGordon|: yet another top-posting thread. And even in a completely unrelated thread.
22:55:23JdGordon|what the FUCK!?
22:56:02kugelclose it temporarily is enough
22:56:04bluebrotherI get the impression that users simply aren't worth something like Rockbox and our support channels :/
22:56:42bluebrotherand those users whining about an etiquette won't come back?
22:57:04JdGordon|well... we could just abandon the ml and leave it to rot... closing it would mean more of them in here and the forums
22:57:12***Saving seen data "./dancer.seen"
22:57:16AlexPJdGordon|: That pretty much happens anyway
22:57:25bluebrotherwell, I've left the list quite a while ago for exactly that reason
22:57:40AlexPHands up how many people ever read it except for when someone points out a stupid thread?
22:58:02obolinuxstb: on further digging the keypad seems to be the same as the sansa fuze, so I'll change the CONFIG_KEYPAD to match
22:58:06kugelI need to take a weekend and read the whole thread, lol'ing all the time
22:58:07AlexPI looked at it once, and decided I'd rather do many many many unpleasant things than join it
22:58:21*gevaerts thinks that we should send an announcement on the user ML stating that the list will be closed on the next top posting or netiquette rules complaint, and then actually follow up on that
22:58:22linuxstbobo: Congratulations! ;)
22:58:31kugelAlexP: I only read it when someone points to a stupid thread :p
22:58:57saratogai've said before that i think the mailing lists are an absolute mess
22:59:44bluebrotherthey are. Unfortunately. But the users ML is the worst, -dev is bearable
22:59:51JdGordon|I reckon its all Llorean's fault for letting the rabble get away with blue bloody murder! thats what happens when you dont lay down the law occasionally!
22:59:58kugelcommitters is lovely :D
23:00:36saratogai do like committers
23:00:54saratogawhat are the arguments in favor the users list exactly? seems like those people should really just be on the forums
23:01:07AlexPsaratoga: You git, don't send them there!
23:01:28*amiconn still prefers the ml over the forums by far, regardless of top posting
23:01:30gevaertsAlexP: on the forums, deleting posts is easy :)
23:01:37AlexPgevaerts: true :)
23:01:39saratogaI mean theres the equivilent of maybe 5-10 real threads a month in the mailing list, and half are just arguing about nothing
23:01:47*JdGordon| wonders why AlexP is calling saratoga a version control system?
23:02:14*bluebrother still likes MLs too but only if they are useable
23:02:15saratogaI mean on the users list that is, the dev list is nice for annoucements
23:03:14JdGordon|we have our fair share of arguments ont eh dev ml also
23:03:19Unhelpfulkugel: i'm rebuilding a fresh long-call-stubs-only toolchain, and will try a completely fresh build, but if that works where make clean doesn't i'm inclined to think we do something wrong
23:03:44bluebrothertrue, but the users list is worse, especially when it comes to ignorants
23:04:02kugelUnhelpful: I don't know what wrong, but sometimes make clean just isn't enough
23:04:36saratogaI like to think many of the arguments on the dev list are ultimately productive in that they lead to some sort of agreement on how things ought to be
23:04:44saratogathough to be honest I rarely read them
23:05:52 Join shotofadds [0] (n=rob@rockbox/developer/shotofadds)
23:05:53kugeloften threads also are discussed so endlessly so that minor things lead to people getting tired and no agreement being reached
23:06:25kugelRegularly that are the ones I start ;)
23:06:29amiconnkugel: That may e.g. happen if source files get renamed and their related objects are located in the build root
23:06:45n1si usually read them if i care about the subject or if they are < 20 mails
23:07:13amiconnSinec 'make' clean' only cleans objects an earlier 'make' created, they will stay there. And depending on the order the linker picks up objects, and outdated object may get linked
23:07:57amiconnSimilar things happen if an arbitrary source file gets renamed (or a different source file gets used for a target, e.g. an asm version) and you don't 'make clean'
23:08:05Unhelpfulamiconn: but that's the sort of thing that could happen over time *without* toolchain changes, and *can't* happen because of toolchain changes alon.
23:08:41amiconnThis often leads to both objects residing in an .a, and the linker usually picks the first one it finds, which is the old one
23:09:43CIA-71New commit by 03robert (r21848): FS #10436 - add the Sansa View to the build system. The bootloader builds but doesn't do anything useful yet.
23:10:18amiconnOn cygwin I often tried to delete only the affected subdir in the build dir, because a full rebuild is so slow. On linux with ccache enabled this is much less of an issue
23:11:43amiconnUnhelpful: Right, but 'make clean' is mandatory when switching toolchains
23:12:19Unhelpfulalso in this case it would lead to link failure - gnu ld will not link objects of different eabi versions. this is why you can't just try adding -Wa,meabi=4, unless you manage to completely avoid any use of toolchain libraries
23:13:01Unhelpfullibgcc must be built with -meabi=4 passed to gas, and then your other binaries likewise, for them to correctly link with stub generation
23:13:03 Quit dash32 (Read error: 110 (Connection timed out))
23:13:14saratogamt: what is the meaning of samples per channel? is it derivived from the imdct block size?
23:13:27Unhelpfuland obviously letting such objects meet "true" eabi objects is a very very bad idea
23:13:34 Join dash32 [0] (
23:15:41kugelhmm, with the View and PP6100 in the target tree, could we eventually get around fixing the mess there?
23:17:23Unhelpfuli've not found any struct-passing-to-asm in libmad yet
23:17:38 Quit _zic (Read error: 101 (Network is unreachable))
23:17:53 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
23:18:11linuxstbkugel: Which mess in particular? And who is "we" ? ;)
23:18:15Unhelpfulit may be that the *only* breakage in mp3 was that of using gcc-4.3.3
23:18:38kugellinuxstb: We is MrSomeone :P And I mean the manufacturer vs soc mess
23:18:43 Quit martian67 (SendQ exceeded)
23:19:29linuxstbkugel: Yes, he should.
23:19:37 Join stripwax [0] (
23:21:52*JdGordon| volanteers kugel for the job
23:22:05JdGordon|Mr Someone is pretty useless for getting things actually done
23:23:52kugelwell, I thought about it
23:24:14kugelbut the problem is to decide which subdir structure PP needs
23:24:17itchegquestion I have an Ipod 5.5 with a build from about a month ago that the settings get reset on relatively ofter, what would be the best way to troubleshoot this
23:24:53kugela generic PP will need a dir extra which our target tree doesn't handle, PPXXXX in arm/ will lead to many PP-files in arm/ still
23:25:32itchegI should mention I use it with an FM transmitter via the recently implemented iap
23:26:31 Join __lifeless [0] (n=lifeless@
23:27:44 Join advcomp2019_ [0] (n=advcomp2@unaffiliated/advcomp2019)
23:28:37 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
23:29:17 Quit advcomp2019 (Nick collision from services.)
23:29:18 Nick advcomp2019_ is now known as advcomp2019 (n=advcomp2@unaffiliated/advcomp2019)
23:29:41 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
23:30:34 Quit martian67 (SendQ exceeded)
23:30:40 Join r0b- [0] (
23:31:17bertrikhm, simply changing the DC level on the meizu m3 piezo when keys are pressed or when sliding over the touch strip is hardly audible
23:31:19 Join dmb [0] (n=Dmb@unaffiliated/dmb)
23:32:25 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
23:34:17rasherZagor: Setting LC_ALL=C when building seems like it'd be rather useful
23:34:27Zagorrasher: yes, I'll add that
23:34:49 Quit n1s ("Lämnar")
23:35:44 Join bubsy [0] (i=Bubsy@
23:37:25saratogabertrik: you need to apply a tone to a PZ
23:40:02 Quit evilnick ("Page closed")
23:42:08 Quit _lifeless (Read error: 110 (Connection timed out))
23:42:14 Quit itcheg (Ping timeout: 180 seconds)
23:44:15 Quit stripwax ("")
23:44:35 Join safetydan [0] (n=deverton@rockbox/developer/safetydan)
23:45:40 Join stripwax [0] (
23:46:23 Quit thegraey ()
23:46:23 Quit TheSeven (Read error: 54 (Connection reset by peer))
23:46:50 Join TheSeven|Zzz [0] (
23:46:53 Nick TheSeven|Zzz is now known as TheSeven (
23:48:39 Quit Horscht (Read error: 110 (Connection timed out))
23:50:28CIA-71New commit by 03kugel (r21849): Try to make configure respect the TMPDIR environment variable.
23:50:50kugelI don't think it will help :(
23:53:53kugelwhy is the new build system not running?
23:54:02Zagorit is
23:54:11ZagorBuild: build mrobe100sim rev 21849 client kugel-x-kugel
23:54:31kugelah, I was tricked
23:54:45kugelthere's still no progressbar/indicator on new.cgi
23:55:10kugellooking for my client takes a few keystrokes so I didn't bother :(
23:55:11bertriksaratoga, I think the OF doesn't use a tone, I hear just a click (maybe a very short tone)
23:55:48saratogabertrik: if you just change a voltage, you're sending it one cycle of a square wave, which means you get just a very little bit of sound out
23:55:54bertrikI'l try driving it with opposite polarity for the two piezo outputs
23:56:01saratogamaybe they turn it on and off a couple times in a row?
23:56:39kugelindeed, it didn't help
23:57:38JdGordon|the clips batt hasnt been calibrated yet has it?
23:57:48JdGordon|29h remainig with 42%?
23:58:12bertrikhours remaining is not related to calibration
23:58:54bertrikthe calibration just maps voltage to percent charge, not to absolute hours remaining

Previous day | Next day