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

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

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

#rockbox log for 2009-07-15

00:01:01*gevaerts thinks that the build page looks weird
00:02:58CtcpPing from gevaerts!n=fg@rockbox/developer/gevaerts
00:02:59cg_hmm, seems to be now building two targets at once?
00:03:23gevaertsthat too
00:04:45Zagorgevaerts: yeah I tried some "hot updating" but that didn't work too well yet
00:04:50 Quit jgarvey ("Leaving")
00:05:09 Join Thundercloud [0] (
00:07:27gevaertsZagor: I'd really like a -uploadspeed parameter to help my poor adsl line survive a bit better (or maybe -curlopts?)
00:08:00*gevaerts decides to see if he can manage that
00:08:01Unhelpfulamiconn: double-checked. the -meabi=4 flag alone does not break vorbis, only removing -mlong-calls also so that stubs will be generated causes breakage. this feature might need to wait for a newer binutils
00:08:20Zagorgevaerts: yeah I was just thinking about ways to deal with that
00:09:16gevaertsZagor: I'd just use .curlrc, but I don't want to limit my download speed, and someone forgot to make the rate limiting in curl direction specific
00:10:21gevaertsmaybe use a curlrc file in the checkout if it's there?
00:10:36 Quit Rob2223 ()
00:10:37amiconnUnhelpful: I can only think of two possible causes, either a binutils bug or a hidden bug in our vorbis decoder
00:11:09kugelgevaerts: I think Bagder accepts bug reports
00:11:50gevaertskugel: I think he prefers patches, but I don't feel like working on that right now :)
00:13:54 Quit robin0800 ("Nice Scotty, now beam my clothes up too!")
00:17:32 Join Rob2222 [0] (
00:17:32 Join robin0800 [0] (
00:23:45scorche|shlinuxstb: i have the intention, yes...but have no clue when i will ever have the time to work on it...
00:24:44amiconnUnhelpful: Did you check whether the resulting call paths change?
00:28:42 Quit Zagor ("Leaving")
00:29:54 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
00:31:36 Quit bertrik (Read error: 113 (No route to host))
00:39:21 Quit DarkDefender ("Leaving")
00:43:01 Quit evilnick ("Page closed")
00:44:59 Quit r4v5 (Remote closed the connection)
00:45:28*funman just sent an email to an atmel contact who might have informations on which SD controller is used in the Fuzev2 and Clipv2
00:48:09 Join stripwax [0] (
00:48:40 Quit shotofadds ("Leaving")
00:49:25 Quit CaptainKwel ("Page closed")
00:50:58 Join Sajber^ [0] (
00:52:19 Quit stripwax (Read error: 104 (Connection reset by peer))
00:52:22 Quit mirak ("Ex-Chat")
00:54:05 Quit Sajber^ (Client Quit)
00:54:33 Join stripwax [0] (
00:54:40 Join Sajber^ [0] (
00:57:48***Saving seen data "./dancer.seen"
00:59:06kugelgevaerts: see what I mean :D
00:59:59 Nick punyhumanright is now known as karma (
01:11:27 Quit flydutch ("/* empty */")
01:17:19 Part toffe82
01:22:41 Quit Thundercloud (Remote closed the connection)
01:23:55 Quit kugel (Remote closed the connection)
01:31:59 Join evilnick_home [0] (
01:36:17 Quit graey (Read error: 110 (Connection timed out))
01:41:33 Quit CIA-71 ()
01:50:10 Part wincent ("Kopete 0.12.7 :")
02:00:47 Join CIA-69 [0] (n=CIA@
02:05:28 Join fdinel [0] (
02:16:48funmanwe might have to check the SD status register sent by cmd13
02:16:50Unhelpfulamiconn: what exactly should i be checking? the basic process here is for gcc to always emit bl <symbol> for calls, which is what it does without -mlong-calls, anyway. the assmebler then labels the callsites as R_ARM_CALL relocations, and then the linker either copies in the correct offset if they're in-range, or generates a stub to use as the bl target, the stub being __[label]_veneer: ldr pc, [pc, #-4] ; .word <address>
02:18:34 Join z35_ [0] (
02:18:52 Quit z35_ (Read error: 104 (Connection reset by peer))
02:21:17Unhelpfulthe #-4 seems a little weird to me, as the address is stored after the ldr...
02:21:39 Quit saratoga ("Page closed")
02:22:07funmanUnhelpful: pc points 8 bytes after the current instruction when read (4 bytes in thumb, so always the size of 2 instructions)
02:22:19 Quit robin0800 ("Don't follow me")
02:22:38 Quit efyx (Remote closed the connection)
02:23:46 Quit CIA-69 ()
02:23:54funmanmy attempts to use cmd55 (app cmd) after card init fails with CMD TIMEOUT ...
02:24:21Unhelpfulah, so #0 would be the place after the address, and -4 is the address, and that's entirely right. :)
02:25:30Unhelpfulso, what i should be looking for is 1) any code aside from the call instructions themselves that differs 2) any calls to the wrong veneer function 3) any veneer that do not point where they claim to ?
02:26:52Unhelpfuli have difficulty imagining how any of those things would happen... especially how #2 and #3 would produce the result i see (vorbis files skip on attempting playback without error report)
02:35:08amiconnIf a codec skips a file, it is due to an error. That error might or might not be displayed
02:35:37amiconnDid you replace both core and codecs in your test, or did you mix them?
02:35:38 Join CIA-69 [0] (n=CIA@
02:37:51 Quit JdGordon| ("Miranda IM! Smaller, Faster, Easier.")
02:42:47amiconnHmm, vorbis is the only codec using setjmp...
02:43:16Unhelpfulreplaced both. since calls between the two are always via pointers, i *could* mix them... it might make it clear whether it's a core or a codec issue.
02:50:35 Quit bzed (Remote closed the connection)
02:50:38 Join bzed [0] (
02:51:46amiconnYou can only mix them if you're not using different abis
02:53:19Unhelpfulamiconn: i'm not - i've built both with stock gcc except for the spec change to pass the -meabi=4 assembler option, but one of them is built with -mlong-calls, the other is not
02:53:48Unhelpfulwould being called via a veneer perhaps mess up setjmp or longjmp?
02:53:55*amiconn wonders how a real eabi build performs in this case
02:55:07Unhelpfulthat's a few days ago for me, but i believe what happened was that vorbis stalled at the start of the file, and that after stopping playback other things went wrong (that had worked before attempting playback)
02:55:30Unhelpfulthere are no stubs for setjmp or longjmp...
02:55:58Unhelpfulnm on the .elf, since it's been through linking, tells you exactly which functions have stubs :)
02:55:59amiconnI suspected setjmp for a while, but the generated asm is okay, and being called via a veneer shouldn't matter as that doesn't touch lr
02:57:34Unhelpfulprecisely... the veneer doesn't do anything but load the new address into pc
02:57:51***Saving seen data "./dancer.seen"
02:59:55Unhelpfulthere are only four stubs, i could either 1) go around enabling or adding debug logfs or 2) try marking individual stubbed functions as long-call to see if that fixes it
03:00:03amiconnYeah, but if you drop -mlong-calls, the rest of the code changes. All long calls are turned into 'bl'
03:00:28amiconnThis may in turn change regiser allocation
03:00:32 Quit Stephen_ ("Leaving")
03:01:05Unhelpfulit may, because with interworking it needs to load the target address into some other register first, doesn't it?
03:01:05*amiconn wonders whether there are asm blocks involved - another "popular" method for producing hard-to-find bugs
03:01:41amiconnAsm blocks with incorrect clobber lists or similar
03:02:35Unhelpfulamiconn: and perhaps a missed clobber is harmless in one case due to different register allocation?
03:03:26*amiconn alraedy had to deal with such asm blocks
03:03:41Unhelpfulan asm function that fails to save a clobbered register may have a similar effect, right?
03:03:45amiconnThis kind of bug also tends to show up when switching to a newer gcc
03:06:34Unhelpfulif the clobbered register is some calculated flag, or an test value for a loop condition, or such, that would also explain the behavior i'm seeing...
03:07:48Unhelpfulwhereas the idea of it having to do with the relocation itself would make it terribly hard to believe it would do anything but crash
03:12:05 Nick agaffney_ is now known as agaffney (n=agaffney@gentoo/developer/agaffney)
03:22:01 Join JdGordon| [0] (i=ae919271@gateway/web/freenode/x-8eefa34bf5a88223)
03:22:05Unhelpfuladd doesn't need a "cc" clobber, right? only adc would?
03:25:42UnhelpfulCLIP_TO_15 could be optimized on armv6 with ssat, i believe...
03:26:57 Join saratoga [0] (i=9803c6dd@gateway/web/freenode/x-c2db52f4af3f4c9f)
03:27:06saratogaUnhelpful: have you tested vorbis with ASM disabled?
03:27:16saratogashould give you a pretty big clue if its the problem or not
03:28:07Unhelpfulit looks pretty easy to do, at least ;)
03:30:31 Quit funman ("free(random());")
03:30:34Unhelpfulsaratoga: it looked to me as if using 16-bit constants in mdct might possibly be helpful... especially if the large const array in the main loop is packed as well. it basically cuts loads of constants in half if they're packed two to a register.
03:30:56saratogaUnhelpful: i think thats worth trying
03:31:09saratogafor reference, cook uses the same imdct entirely in 16 bit precision with good results
03:31:24saratogaand I think theres an 8 bit version of it floating around thats supposed to be listenable
03:32:28Unhelpfulthe gains from the other things i mentioned amounted to about 2%
03:32:32saratogathough i'm curious if this would apply to ARM4 or just the ones with 16 bit multiply?
03:32:52saratogai saw that, though I didn't see a commit for mp3?
03:33:31Unhelpfuli didn't look at the mp3's mdct... for all i know it has both optimizations already :)
03:33:41saratogaoh I thought you mentioned MP3 before
03:33:50Unhelpfuland no, it would not apply for arm4, so only some unsupported targets that are arm5e and arm6
03:34:02Unhelpfuli tested mp3. i stopped when you said it had its own mdct
03:34:07saratogaah ok
03:34:47saratogai bet simply rearranging ops in the mp3 filterbank and imdct would improve performance on later arm targets, it was written targeting ARMv4 (and for the ifp port IIRC)
03:35:07saratogaand arm7tdmi has fewer restrictions due to pipelining
03:36:09Unhelpfulyou *can* load values stored as halfwords together with arm4, but you'll need 3 ops of fixup to use those constants if they're to be multiplied... an asr #16 will get you the top value, and an lsl #16 ; asr #16 will get you the bottom one
03:36:53saratogayeah i remember looking into doing that and deciding it wasn't worth it
03:36:55Unhelpfulif you're going to be *adding* the values, and the other side of the add need not be rotated, you may need only one fixup op, in which case it's more likely to pay off to save the load
03:38:04Unhelpfulthe lack of input shifts or immediates for the multiply and parallel-math ops really stings sometimes ;/
03:38:58 Quit JdGordon| (Ping timeout: 180 seconds)
03:48:07Unhelpfuldisabling arm assembly in libvorbis = playback proceeds :)
03:58:09 Quit JdGordon ("Leaving.")
03:59:55 Join wannaplaylist [0] (n=813b5dd5@gateway/web/cgi-irc/
04:00:46 Join faemir [0] (n=faemir@
04:02:49 Join AndyI [0] (i=AndyI@
04:02:59 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
04:03:05wannaplaylistI have a question for anyone about the 'on the go playlist creation'. I have never heard of rockbox til now so feel free to assume that i know precisely jack about it. I want to get an mp3 player that is cheaper and less 'locked in' i guess, than the ipod. Lots of them seem to not allow customized playlists though. Is the on the go feature in rockbox something that will allow me to make a folder of son
04:04:02scorche|shwell, 'the ipod' is just a is the firmware that makes it 'locked in'...
04:04:14scorche|shalso, your message got cut off at allow me to make a folder of son
04:04:39 Join Strath [0] (
04:04:49wannaplaylistand is rockbox like firmware that you install on an mp3 player?... ok let me post the rest...
04:05:01wannaplaylist"Is the on the go feature in rockbox something that will allow me to make a folder of songs and put them on a device as a playlist? Thanks for any info."
04:05:32scorche|shand yes...rockbox is a replacement firmware as it says on the front page...
04:06:43wannaplaylistok here is when the techidiot comes out... i really hesitate to alter the firmware on my ipod bc i don't want it be irreversible. just letting itunes update the ipod firmware killed all of my music once today...
04:07:55 Quit TheSeven (Nick collision from services.)
04:08:11 Join The_Seven [0] (
04:08:15 Nick The_Seven is now known as TheSeven (
04:08:38wannaplaylistbut i want a second mp3 player anyway and might get a sandisk type.
04:08:39scorche|shit is extremely low risk to install and can easily be reversed....that said, some parts of rockbox can be a bit complex, so might a recommend a read through the manual to get yourself aquainted with how things work/what you should expect?
04:08:41wannaplaylistdoes that on the go playlist that rockbox enables mean that i can drag and drop 1 or more folders to a device and have it keep that folder intact as a playlist?
04:09:27scorche|shit means that you can load whatever music you want and while out and about add/move/delete whatever folders or files you wish from a playlist
04:10:29wannaplaylistoh, i didnt know it was reversible... so you can have multiple customized playlists? thanks for your help. once i get a new player i will read the manual before i install stuff since i'm not familiar with it.
04:12:03scorche|shyou can do whatever you want with a playlist...i usually dont use pre-created playlists though and just make one of what i want to hear at any one time...the manual will probably give you a better image of how these things work though...
04:13:34wannaplaylistok. it sounds like the feature is what i was looking for at least. i didn't see any point in having 4G of music if i couldn't organize it into lists that i wanted... it seems most force you to use playlists they make that are an artist/album/genre, etc.
04:13:45wannaplaylistthanks for the response
04:16:38 Quit AndyIL (Read error: 110 (Connection timed out))
04:18:50 Join dys`` [0] (
04:21:38 Join kamlurker [0] (
04:25:15 Quit wannaplaylist ("CGI:IRC")
04:25:41 Quit kamlurker (Client Quit)
04:30:37 Join GreatBeaver [0] (
04:30:50GreatBeaverdo people defrag their mp3 players? and what program?
04:31:11scorche|shthat doesnt really have anything to do with rockbox..
04:32:05GreatBeavercan rockbox have a built-in defrag?
04:32:35scorche|shit would be silly to have one in rockbox...your computer can do a much better job of it
04:33:32 Quit dys` (Connection timed out)
04:39:05 Quit saratoga (Ping timeout: 180 seconds)
04:41:10 Join thegeek_ [0] (
04:47:15 Quit faemir (Read error: 110 (Connection timed out))
04:52:24 Quit n00b81 (Read error: 110 (Connection timed out))
04:57:52***Saving seen data "./dancer.seen"
04:59:03 Quit thegeek (Read error: 110 (Connection timed out))
05:04:07Unhelpfulhah... replacing *either* of vect_mult_bw or vect_mult_fw with the C implementation gets things back to working
05:05:56 Join derekja [0] (
05:06:52derekjahow do you make rockbox ignore the "the" in front of a band namd
05:07:24derekjaalso, it keeps crashing on my ipod video
05:10:20Unhelpfuli believe the C and asm versions could both be made slightly more efficient by passing the end address rather than the count... especially seeing as the caller *has* the end address and *calculates* the count for the call
05:12:52JdGordonderekja: you dont.. and thats not really an argument which anyone can be bothered having...
05:13:14JdGordonthere is a correct way to do it... using the sort tag, noone has done it in a way we want it, so its not being done yet
05:16:00derekjaJdGordon: and idea why it keeps crashing on my ipod video?
05:16:20JdGordonit being what?
05:21:10 Part derekja
05:21:11 Join derekja [0] (
05:21:15 Part derekja
05:22:02 Quit GreatBeaver ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060215]")
05:25:12 Quit Horscht ("Verlassend")
05:37:32 Join dash32 [0] (
05:39:42 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
05:44:38 Quit gevaerts (Nick collision from services.)
05:44:51 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
05:47:02 Quit Llorean (Read error: 54 (Connection reset by peer))
05:47:40 Join Llorean [0] (
05:55:22 Quit Strath ("Leaving")
06:01:20JdGordonare image co-ordinates viewport relative? or screen relative?
06:02:00UnhelpfulJdGordon: isn't pretty much *everything* in core screen-relative?
06:02:06Unhelpfulerm, VP-relative?
06:03:01JdGordonhmm... customwps says it is...
06:03:10*JdGordon doesnt wan to do math
06:09:52JdGordonanyone know the volume keys in the clip sim?
06:20:20CIA-69New commit by jdgordon (r21877): display the volume as a number when its changing in the cabbie theme on the clip
06:25:47 Quit dash32 (Remote closed the connection)
06:28:17 Quit JdGordon ("Leaving.")
06:30:13 Join BHSPitMonkey [0] (n=stephen@unaffiliated/bhspitmonkey)
06:33:44*Llorean wonders why he only updated one WPS
06:38:25 Join webguest76 [0] (n=40ffb426@gateway/web/cgi-irc/
06:39:04 Join webguest51 [0] (n=40ffb426@gateway/web/cgi-irc/
06:39:08 Quit webguest76 (Client Quit)
06:39:30 Join webguest54 [0] (n=40ffb426@gateway/web/cgi-irc/
06:40:07 Join webguest13 [0] (n=40ffb426@gateway/web/cgi-irc/
06:40:15 Quit webguest54 (Client Quit)
06:40:15 Quit webguest51 (Client Quit)
06:41:52 Join webguest58 [0] (n=40ffb426@gateway/web/cgi-irc/
06:41:52 Quit webguest13 (Client Quit)
06:43:52 Join webguest15 [0] (n=40ffb426@gateway/web/cgi-irc/
06:43:52 Quit webguest58 (Client Quit)
06:44:02 Quit webguest15 (Client Quit)
06:44:09 Join webguest51 [0] (n=40ffb426@gateway/web/cgi-irc/
06:44:49 Join webguest41 [0] (n=40ffb426@gateway/web/cgi-irc/
06:44:49 Quit webguest51 (Client Quit)
06:45:33 Quit webguest41 (Client Quit)
06:52:57 Join webguest99 [0] (n=40ffb479@gateway/web/cgi-irc/
06:53:06 Quit webguest99 (Client Quit)
06:53:13 Part ybit
06:55:06 Join derekja [0] (
06:55:15 Quit derekja (Client Quit)
06:57:54***Saving seen data "./dancer.seen"
07:08:31 Quit dmb (Read error: 113 (No route to host))
07:09:08 Nick fxb__ is now known as fxb (
07:14:28 Join einhirn [0] (
07:14:34 Quit fyrestorm (Read error: 104 (Connection reset by peer))
07:19:35 Join bertrik [0] (
07:28:01 Nick fxb is now known as fxb__ (
07:31:18pixelmaJdGordon: will you update the other WPSs too? (I agree with Llorean and think the default WPS should be as similar as possible across targets and since all bitmap ones show a volume icon...)
07:32:33pixelmagenerally I like the info there though and even thought about changing them myself but there are version that don't use viewports at all so that means more work....
07:34:52LloreanI can understand some making it unique per target to fit the hardware, but things like this where there's no reason it shouldn't be the same, it should really be done for all or none.
07:36:57pixelmayes, that's what I meant to with "as similar"... on that note, I'm reminded that I wanted to make a suggestion for the c200 using a 10-pixel high font to get one more line in the WPS to use for the playback times info (and generally better fit)
07:37:58pixelmathey used the Sazanami-Mincho-Regular fonts and that one exists as 10 too (but of course is a bit smaller)
07:38:49pixelmaemm... the 11-Sazanami for the c200 currently
07:39:21 Join stoffel [0] (
07:42:03 Join AndyIL [0] (i=AndyI@
07:44:12 Quit r0b- ("( :: NoNameScript 4.02 :: )")
07:50:13 Quit stoffel (Remote closed the connection)
07:50:39 Join stoffel [0] (
07:54:19 Quit AndyI (Read error: 110 (Connection timed out))
08:10:21 Join r0b- [0] (
08:10:23 Quit bertrik (Read error: 113 (No route to host))
08:12:15 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
08:35:56 Join flydutch [0] (
08:38:16 Quit r0b- (Read error: 110 (Connection timed out))
08:38:35 Join Zarggg_ [0] (
08:39:18 Part safetydan ("Leaving.")
08:43:58 Quit Zarggg (Read error: 113 (No route to host))
08:44:42 Quit pixelma (Nick collision from services.)
08:44:44 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
08:45:00 Quit amiconn (Nick collision from services.)
08:45:02 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
08:45:03 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
08:45:22 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
08:47:32 Join Rob2223 [0] (
08:50:35 Join scorche [50] (n=scorche@rockbox/administrator/scorche)
08:51:28 Quit TheSeven ("ChatZilla 0.9.85 [Firefox 3.5/20090624025744]")
08:57:55***Saving seen data "./dancer.seen"
08:58:00 Join Zagor [242] (n=bjorn@rockbox/developer/Zagor)
09:05:27 Quit Rob2222 (Read error: 110 (Connection timed out))
09:14:58 Join petur [50] (n=petur@rockbox/developer/petur)
09:23:50 Quit scorche (Read error: 60 (Operation timed out))
09:24:46 Join scorche [50] (n=scorche@rockbox/administrator/scorche)
09:26:22 Quit BHSPitMonkey (Read error: 104 (Connection reset by peer))
09:31:44 Join Thundercloud [0] (
09:33:58 Quit stripwax ("")
09:48:31 Quit Thundercloud (Remote closed the connection)
09:56:54 Quit jfc (Read error: 104 (Connection reset by peer))
09:57:10 Join jfc [0] (
09:58:24 Join obo_ [0] (
09:58:32 Quit obo ("No Ping reply in 90 seconds.")
10:02:17 Quit JdGordon (Read error: 110 (Connection timed out))
10:03:21 Join Grahack [0] (
10:10:20 Quit stoffel (Remote closed the connection)
10:21:22 Join FlynDice [0] (
10:27:17 Join stoffel [0] (
10:33:14 Nick fxb__ is now known as fxb (
10:37:19FlynDicefunman: (for the logs) RE: SD card speeds P.6 sd spec 3.4 SpeedClass, RE : 12.5/25 MB/sec __interface__ speed p.39 sd spec 4.3.11 High-Speed Mode
10:39:05 Join Jaykay [0] (
10:39:18 Join efyx [0] (
10:41:39Jaykayi suggest closing of FS #5886, FS #6212, and FS #8895 with the reason "out of date", because FS #9873 is going to be committed
10:41:54Jaykaythey are all about adding some functionality to the rec button
10:48:40 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
10:57:58***Saving seen data "./dancer.seen"
11:19:59 Join graey [0] (
11:21:38gevaertsJaykay: how about waiting until it's actually committed?
11:22:11amiconnUnhelpful: Btw, if loading 16 bit values together allows you to use ldmia, it pays off to do so on arm7 even if you have to do some unstuffing. The libdemac filters used to do that before I switched them to using 32 bit ints on armv4.
11:22:29Jaykaygevaerts: they are out of date anyway because theres a newer patch with the same functionality
11:23:02gevaertsJaykay: a patch isn't better just because it's newer
11:23:09amiconnUnhelpful: Oh, and in my experience stating "cc" as clobbered is unnecessary. It won't hurt though
11:25:27Jaykaygevaerts: then its better because it supports all targets :)
11:27:44gevaertsJaykay: even that is not good enough. If all else is equal, sure, but is it?
11:31:56Jaykaygevaerts: i don't know, i don't know enough about the code. then i'll wait until FS# 9873 is committed :)
11:34:33 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
11:34:44 Quit bmbl (Remote closed the connection)
11:35:19 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
11:35:23 Quit bmbl (Remote closed the connection)
11:40:14 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
11:41:41 Join DarkDefender [0] (
11:47:39Unhelpfulamiconn: well, these two asm functions list everything they clobber properly, and they're only called in one place, and looking at the asm there i don't see anything suspicious - but replacing either one of them with the C version from misc.h stops vorbis from failing again... perhaps it's something that happens somewhere else because of a difference in size, as that function has no long calls in it, anyway...
11:49:02Unhelpfuland i was looking at the bits in libtremor... i'm quite sure the multiply and vector functions don't do anything too silly, although the multilpy ones are not needed as it's quite easy to get gcc to generate the same asm with very readable C math
11:52:00Unhelpfulvect_mult_fw might go a bit faster with all four smull in front and then the shifts, but i don't see how it can be trashing any registers without gcc knowing
11:53:37Unhelpfuli believe the pattern used is especially poor on the newer arms... i'm pretty sure the arm1136 manual states that a shifted input register must be ready a cycle early
11:56:38*amiconn wonders what the XPROD32 macro in asm_arm.h is actually doing
11:57:20amiconnIt calculates 'l', but doesn't return that as the result of the block...
11:58:50*amiconn also wonders why that file seems to be duplicated
11:59:17amiconnIt's in codecs/lib/ as well as in codecs/libtremor/
12:09:30amiconnUnhelpful: Hmm, if you're mixing arm asm and C there, the result will be incorrent. See the comments regarding delayed shifts
12:10:58 Quit Jaykay (Read error: 110 (Connection timed out))
12:14:10Unhelpfulamiconn: i noticed that... however, if *either* of those is swapped out for the C version, it stops skipping vorbis tracks. perhaps the problem has something to do with where in memory some other value is being stored, because if the problem were one of these functions, it would matter which one i replace... and if it were both, i would *have* to replace both.
12:14:43 Join robin0800 [0] (
12:16:15 Quit bubsy ()
12:17:58 Quit efyx (Read error: 113 (No route to host))
12:18:49 Join bubsy [0] (i=Bubsy@
12:21:51amiconnUnhelpful: Btw, if you're looking for possible optimisations - asm_arm.h has several functions which could be sped up a lot using 'clz' on armv5 and above
12:22:18 Join daurn| [0] (
12:22:19 Join daurnimator_ [0] (
12:23:04Unhelpfulthe arm asm division should be using clz as well... and i should really get around to writing that specialized shifted-output divide for pictureflow
12:24:37amiconnWhich division are you talking about?
12:25:07amiconnGcc uses 'clz' in its division routine on armv5 and above.
12:27:40 Quit daurn (Read error: 60 (Operation timed out))
12:28:24 Quit daurnimator (Read error: 60 (Operation timed out))
12:31:46Unhelpfulwe'd talked about this before, but basically fdiv calculates (num << 10) / den by first using clz to see how far it can safely shift the numerator, and shifting the denominator right if needed, and then doing the division.
12:33:37Unhelpfulsince arm doesn't have a divider, and we have to loop, this could be handled more efficiently, i think, by producing shifted output in the division routine directly. essentially, shift the bit-to-set value left by 10, and do a normal division, with up to 10 extra iterations to fill in the fractional bits
12:34:55*amiconn was talking about libtremor and hence didn't expect Unhelpful suddenly jumping to pf optimisations
12:35:31*Unhelpful has a long to-do list :)
12:37:04Unhelpfuli'd *really* like to know what's breaking libtremor, though... i think the asm blocks in question may be a red herring
12:38:16amiconnSince it doesn't crash, it should be possible to log the error. Adding log statements may change behaviour if it's really an alignment problem
12:38:52amiconnDid you try if it also breaks if you do an ordinary short-call build (i.e. not using iram at all - probably slow as hell)?
12:41:50Unhelpfulno, i didn't... i think a good way to prove an alignment bug might be to just check the sizes for the two versions of that function and add a pad
12:43:53Unhelpfulif i can "fix" the problem in short-call-with-stubs by changing code that shouldn't break or fix anything, i should be able to change the behavior just by adding a pad to whichever version of the function is smaller, if the size is causing the problem.
12:44:24 Quit courtc (Read error: 113 (No route to host))
12:47:45 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
12:47:56 Nick daurnimator_ is now known as daurnimator (
12:52:45amiconnThat almost sounds like an off-by-one access somewhere
12:53:09amiconnChanging the code size will shift variables in memory
12:53:25amiconnAre you testing on a PP target?
12:54:04 Join mt [0] (n=mt@
12:55:26 Join webguest20 [0] (n=ad581cda@gateway/web/cgi-irc/
12:56:10 Quit webguest20 (Client Quit)
12:57:59***Saving seen data "./dancer.seen"
12:59:43Unhelpfulthe asm versions both unroll the loop by four... they're actually *larger*
13:02:39amiconnHmm, the vorbis codec doesn't seem to use dualcore. Otherwise wrong cache line alignment would have been a potential cause
13:05:59Unhelpfulfrom other channel: "successs". if i swap in the C versions for the asm ones, vorbis no longer fails tracks (although the output is wrong because of the delayed shift). if i just add a global asm with ".size 20" (20 being the difference in size of the two objects), it's broken again.
13:09:44amiconnHmm, so we do have an alignment issue in the vorbis codec...
13:10:07*amiconn once tracked down such a beast by bisecting
13:11:21amiconnFirst, put the filler at the very beginning and find out which sizes make the code break and which don't. Then, bisect the filler position
13:14:29amiconnThis is how I found that PP5002 doesn't like its sleep instruction at certain addresses
13:15:39Unhelpfulick. i'll take a shot at it tonight... i need to be getting to bed now. i guess i could start by moving the filler around to see which function doesn't want to be offset?
13:16:38amiconnI'd first check whether it's code or data
13:19:28ZagorFYI: We plan to disable the old build system tonight
13:22:45gevaertsNice :)
13:31:12 Join mt_ [0] (n=mt@
13:39:47 Join TheSeven [0] (
13:40:11 Join mt__ [0] (n=mt@
13:40:46 Quit mt_ (Read error: 113 (No route to host))
13:45:11 Join bagderoid [0] (
13:48:12 Quit mt (Read error: 110 (Connection timed out))
13:48:23bagderoidzagor maybe se should announce the build switch on the dev list
13:48:48Zagorbagderoid: yup, I'll do that.
13:49:23 Join funman [0] (n=fun@rockbox/developer/funman)
13:50:13funmanFlynDice: i tried reading the status register with acmd13, but i got a data timeout when trying to read the register with DMA
13:50:23funmanboth on internal storage and µSD
13:51:10bagderoidwe should also make sure the instructions are accurate.
13:54:12 Nick mt__ is now known as mt (n=mt@
13:57:35 Quit aditya (Read error: 110 (Connection timed out))
13:57:44 Join aditya [0] (n=aditya@
14:00:45 Join Jaykay [0] (
14:02:47 Join wincent [0] (
14:03:44 Join courtc [0] (n=court@unaffiliated/courtc)
14:04:43soapcan a wiki admin explain the last two diffs here?
14:05:56soapis that just a file being attached, then a second version of said file being attached under the same name?
14:07:51 Join mcuelenaere [0] (
14:09:40 Part bagderoid
14:10:59 Quit Zarggg_ (
14:10:59 Quit Sajber^ (
14:10:59 Quit dionoea (
14:10:59 Quit archstech (
14:10:59 Quit Trista340 (
14:10:59 Quit crashd (
14:10:59 Quit avacore (
14:10:59 Quit Zambezi (
14:10:59 Quit Bagder (
14:11:08NJoincrashd [0] (
14:11:08NJoindionoea [0] (
14:11:23NJoinZarggg_ [0] (
14:11:26NJoinBagder [241] (n=daniel@rockbox/developer/bagder)
14:11:40 Join avacore^ [0] (
14:11:44 Join Trista689 [0] (
14:12:08NJoinSajber^ [0] (
14:12:21 Quit Llorean (Read error: 104 (Connection reset by peer))
14:13:08Zagorsoap: good question :)
14:13:49soapAnd, since the new theme site is long up and running - can we lock the Graveyard pages from upload? Perhaps plan a migration path once-and-for-all from the wiki theme pages to the web theme site?
14:14:42Zagorthat would be nice
14:22:09 Quit dfkt (Read error: 104 (Connection reset by peer))
14:26:07 Quit robin0800 ("Don't follow me")
14:40:03 Join muesli [0] (n=4d15fa43@
14:40:10mueslihi guys
14:40:58mueslihi guys
14:41:32 Quit muesli (Client Quit)
14:41:38 Join robin0800 [0] (n=robin080@
14:41:58 Join muesli [0] (n=4d15fa43@gateway/web/cgi-irc/
14:43:00funmanmuesli: hi
14:43:02mueslimmh..i have no rights to speak??
14:44:20Torneno, the channel is just not very busy right now
14:44:30Torneand people are probably waiting to see if you actually have a question :)
14:44:41muesliah, here we are...
14:44:46mueslig'day mates
14:44:54linuxstbmuesli: You can always log at the logs -
14:45:31muesliah using the web interface. takes some moments until my message shows up ;-)
14:45:31funmanTorne: can you give me again the list of maintainers for linux arm machines?
14:45:54Tornerussel king is the main linux-arm guy
14:45:58Tornethat's about all the list there is :)
14:46:33Jaykayare there any differences between the manual for recorderv2 and fmrecorder?
14:46:33funmanI remember you gave me a list of maintainers for specific machines (listed in arch/arm/tools/mach-types)
14:47:21mueslianyway... referring to
14:47:34muesliquestion 3: whats the usb diskmode?
14:47:51Tornefunman: oh, the machine type registry is on the linux-arm site
14:48:05Tornethere's no guarantee it's up to date
14:48:12Torneit's mostly "the person who asked for that machine type to be added"
14:48:24Tornerather than any guarantee that that person still cares about that port
14:48:40funmanthanks, i'll bookmark it anyway
14:48:49mtlinuxstb: AAC extradata in rm seems to always be 3 bytes. (first 2 are 2h and 12h for the stereo samples I have, I'm guessing the first one could be channel config ?). Still can't figure what the other two are for, assuming I got the first one right !
14:48:52funman for AS353X machine
14:49:47linuxstbmt: Have you looked at the existing aac codec? How big is the extradata there?
14:50:30funmansince extradata is used by the decoder i would expect it is the same ?
14:50:43linuxstbfunman: Yes, that's what I'm asking...
14:50:49linuxstbYou never know though.
14:50:56 Join muesli_ [0] ( try
14:52:06muesli_anyway... referring to
14:52:07muesli_<muesli> question 3: whats the usb diskmode?
14:52:07muesli_sorry 4 flooding the channel with my doublepost
14:52:11 Quit muesli ("CGI:IRC")
14:52:24linuxstbmuesli_: Then why ask again?
14:52:54muesli_i changed to mirc since the webinterface couldnt be used for a proper conversation
14:53:56mtlinuxstb: (Just making sure) In apps/codecs/libm4a/m4a.h, is "codecdata" equivalent to what we referr to as codec_extradata ?
14:54:14 Join LambdaCalculus37 [0] (i=44a0430d@rockbox/staff/LambdaCalculus37)
14:58:01***Saving seen data "./dancer.seen"
15:01:42Jaykaysecond try: are there any differences between the manual for recorderv2 and fmrecorder?
15:02:13amiconnNo, as these devices are technically identical
15:02:16 Quit antil33t (Read error: 104 (Connection reset by peer))
15:02:30 Join antil33t [0] (n=Mudkips@
15:02:30amiconn(except the v2 *may* have the radio, while the fmr *does* have the radio)
15:02:57Jaykaywhy do they have different manuals then?
15:03:37amiconnWell, they wouldn't, but they are two different targets
15:04:04amiconnThis is necessary because the scrambling is different - an fmr loader won't load a v2 firmware and vice versa
15:04:22amiconnBut once descarmbled the code is identical
15:04:48Jaykaybut the manual for all h1xx is also the same file
15:05:31funmanttp:// < an email from Ulrich.Herrmann@ams about SD card / uboot \o/
15:06:26amiconnJaykay: Perhaps someone may add a note about the radio for the v2 only, then they will be no longer identical
15:06:38 Join saratoga [0] (i=439f4426@rockbox/developer/saratoga)
15:08:29Jaykayamiconn: there is a note about the radio in the v2-manual... but this note could also be in the manual for both players
15:08:47amiconnWhy should it? The fmr always has the radio
15:08:51Jaykaya "Note: some v2's don't have a radio" is enough imo
15:10:28muesli_is there a way to contact David A Johnston?
15:10:57 Join n00b81 [0] (n=n00b81@unaffiliated/n00b81)
15:11:55saratogamuesli_: hes a regular on the mailing lists, have you tried email?
15:13:25Jaykayamiconn: imo there's no need for another manual for a single line, but it's not my choice :)
15:16:13muesli_saratoga uff..have no clue about the mailing list. just want to contact him since he seems one of the very ones who got a iriver h120 cf-modded
15:16:34*LambdaCalculus37 has to finish his work on FS #10431 already :)
15:16:57muesli_ thats what i found
15:18:22funmanmuesli_: you can see his email on
15:18:42muesli_ah sexy, cheers
15:20:20GodEateranyone else care to comment on this one : ?
15:20:37Zagorhmm. maybe the Big Green build table should only show builds with problems?
15:21:10GodEaterthat "Tango Digital Media Platform" thing in the dmesg output is making my LibGphoto2Bug bump itch. But clearly hal is implemented slightly differently on his distro.
15:21:21GodEaterZagor: that's a nice idea
15:24:46linuxstbZagor: Isn't it also used to get build details on targets that have worked? But I guess that could be moved to the binsize table (if it even needs to be kept...)
15:26:04 Join itcheg [0] (i=4117734b@gateway/web/freenode/x-a77289e156738188)
15:26:26Zagorlinuxstb: the only thing it does is give easy access to the logs. we could always keep the big on a separate page for when you want to look at the them.
15:28:17linuxstbWould it be possible to keep a larger archive of build info? Things like binsize for old commits would be interesting to be able to lookup.
15:33:13 Quit jfc (Read error: 104 (Connection reset by peer))
15:33:35 Join jfc [0] (
15:33:39*mt points linuxstb to his question about 50 minutes ago :)
15:34:06 Quit jfc (Read error: 104 (Connection reset by peer))
15:34:21*rasher points linuxstb to"> (although that doesn't have as much info as the build system can retain)
15:34:27 Join jfc [0] (
15:34:43muesli_funman could you please also provide a link for
15:34:47 Part n00b81 ("Leaving")
15:34:52 Join n00b81 [0] (n=n00b81@unaffiliated/n00b81)
15:35:56funmanno sorry
15:36:21 Quit antil33t (Read error: 113 (No route to host))
15:36:27 Join antil33t [0] (n=Mudkips@
15:40:12saratogaGodEater: I think the Tango Digital Media id is normal for one of the many USB modes in the PP ROM
15:40:33saratogaIIRC its the pre-manufactoring mode or something like that
15:43:40GodEaterah ha
15:44:19linuxstbmt: Yes, I think codecdata == extradata. It's used to init the aac decoder in the call to NeAACDecInit2() in codecs/aac.c
15:46:10mtI guess then I shouldn't bother what those three bytes represent .. I just read them all at once into the codecdata array.
15:46:12 Quit jfc (Read error: 104 (Connection reset by peer))
15:46:36 Quit Jaykay (Read error: 110 (Connection timed out))
15:46:46 Join jfc [0] (
15:47:10 Quit bmbl (Read error: 110 (Connection timed out))
15:48:30linuxstbmt: Yes, this "extradata" is something which is just passed to the decoder - you shouldn't need to worry about the contents.
15:49:31*mt just thought of doing the same for cook, which should decrease the bin-size a bit.
15:49:34linuxstbrasher: Is that done independently to the main build system?
15:50:03rasherlinuxstb: yeah
15:50:12gevaertsyes and no
15:50:22*linuxstb won't say the obvious...
15:50:45rasherWell, I *think* it gets some of the results from the build system, but I'm not entirely sure that bit of it works these days
15:51:27gevaertsit should still work I think
15:51:36gevaertsmaybe it will stop tonight though...
15:54:33 Quit mt (Remote closed the connection)
16:00:08 Quit antil33t (Read error: 110 (Connection timed out))
16:00:14 Join antil33t [0] (n=Mudkips@
16:04:15 Nick fxb is now known as fxb__ (
16:04:48*funman just mailed the person paid by austriamicrosystems to port uboot/linux to the as353x SoC used in Clipv2/Fuzev2
16:07:42 Join evilnick [0] (i=0c140464@gateway/web/freenode/x-348fb7d0e92a4bac)
16:08:37LambdaCalculus37funman: Now to cross your fingers and hope for a good response. :)
16:09:11 Join BCM43 [0] (
16:09:14funmanLambdaCalculus37: unlike the last persons i had contact with, this person seems to know what "patch" and "open-source" mean, so I have good hopes
16:09:43LambdaCalculus37funman: Maybe he'll want to join us in the porting efforts?
16:09:51BCM43Is rockbox stable on the ipod video 60gb? I installed it for a friend and it keeps crashing.
16:10:21saratogamt: do the floats in COOKContext do anything? if not would you remove them? it'll save some IRAM
16:10:28saratogai think they are left over
16:10:39funmanLambdaCalculus37: i don't know, i think he works on u-boot/linux on his paid time; i can't tell if he wants to work on open source for free, and i don't imagine austriamicrosystems paying him to work on rockbox
16:10:56funmanBCM43: crashing when doing what?
16:11:54linuxstbBCM43: Maybe you need to try the 32MB build
16:12:01linuxstb(but yes, more info is helpful...)
16:12:57saratogamt: also, the current decoder duplicates a very large trig table between the mdct library and libcook, but removing the duplication is giving me occasional audio glitches so removing it is apparently non-trivial or else I am missing something
16:15:32 Nick obo_ is now known as obo (n=obo@rockbox/developer/obo)
16:16:59CIA-69New commit by zagor (r21878): Master no longer needs time reporting, since it knows that itself.
16:17:42muesli_petur are you there?
16:18:37muesli_nice :) just found your postings for a cd-modded h120 in this thread,95239.0.html
16:20:10muesli_could you please have a further look on some details are still not to me
16:20:17 Join jgarvey [0] (
16:23:53peturI'm not really up to date on the h1x0 CF mod state wrt bootloaders. I wanted to work on it and bricked my h120 (my own stupid mistake). That was a year ago :/
16:24:01*petur looks at LinusN
16:24:13CIA-69New commit by zagor (r21879): Added -ulspeed parameter to limit upload speed.
16:24:27muesli_mmh..but can you recall what bootloader you have used?
16:24:38muesli_the standard v6 or the pre-one?
16:24:43peturI didn't...
16:24:45CIA-69New commit by zagor (r21880): Bumped client revision.
16:25:04GodEaterthat's how he bricked it
16:25:52peturI guess V7 is the way to go
16:26:52muesli_but could you use the player as an usb-usm device as when using the hdd?
16:27:01muesli_TODO list for the upcoming bootloader revision
16:27:01muesli_H100: Support USB Diskmode for CFModded players
16:27:02BCM43funman: I was trying to get to folders in the menu
16:27:05muesli_that scares me
16:27:44funmanBCM43: you need to provide a complete bug report: which steps exactly you did, what is the result, what is printed on screen
16:28:04peturmuesli_: I don't understand that remark in the wiki, USB is handled by a bridge chip in hardware, I don't think the bootloader can do much there
16:28:59linuxstbBCM43: Did you install the 60/80GB build or the 30GB build? If the former, then try the latter.
16:29:32 Join Ubuntuxer [0] (
16:31:09BCM43funman: ok, I was using the menu to get to diffrent items, and it would crash every 5 min or so, on artist most often, but on others too. It would simply reboot, showing the apple start up screen.
16:31:18BCM43linuxstb: but it is a 60gb.
16:32:03linuxstbBCM43: I know.
16:32:27soapBCM43, purchased new from Apple? There is a strong possibility if you bought anywhere else than you have the logic board of a 30GB model inside the case of a 60GB model. The 30GB model has less RAM, and would suffer similar problems if you used the 60/80GB build.
16:32:42BCM43soap: nowhere near new.
16:33:02BCM43soap: ok, thanks
16:33:38funman\o/ the ams guy replied
16:33:39GrahackHi everyone, I noticed that r21863 breaks the Lua plugin. Some statements don't work correctly anymore: string.format('simple %s', 'test') -> "simple"
16:33:57linuxstbfunman: Positively?
16:34:45funmannot really :
16:35:02daurnimatorGrahack, hows the lua coding
16:35:21funmanI understand that the SD controller is made by Synopsys ?
16:35:33Grahackdaurminator: broken at the moment
16:39:41funmanSynposys DesignWare Mobile Storage IP (I'm not sure what "IP" means there, it is used a lot on Synopsys website)
16:40:14 Join toffe82 [0] (n=chatzill@
16:42:20saratogafunman: intellectual property, in this case it usually refers to the combination of the HDL code for the controler, and the code for the drivers
16:42:41saratogadoes that person her tells you to contact work for AMS?
16:43:11funmanyes, i have already been in contact with him and he seemed interested in rockbox on as3525
16:43:40saratogafunman: is there any linux source at all for the 353x?
16:43:51funmanhe didn't answer my mail from april 30th
16:43:57funmanno there is no source at all
16:44:31saratogaah so thats probably the legal issue he refers to
16:45:11 Join brokenscreen [0] (n=5dd4a78e@gateway/web/cgi-irc/
16:46:01saratogaLambdaCalculus379: I don't understand your reply in the flashwriter thread
16:46:41funmanperhaps they want to release the code only to their customers, so there is less change it becomes available in the wild
16:47:31brokenscreenis there a help channel ?
16:47:44saratogathis is it, for rockbox anyway
16:48:38 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
16:51:09brokenscreenok well i dont get it baken to create my own menue / wps ~ is there ab GUI that can help me ?
16:51:28saratogaI don't think so
16:53:50brokenscreenok well how can i tell RB to use only rthe upper right screen part of the display ? | the rest of the display is broken!
16:55:55saratogain the wps you can simply not draw anything to other parts of the screen
16:55:56soapI'd use the existing voice system.
16:56:10linuxstbbrokenscreen: The only way would be to modify the LCD driver and compile your own version of rockbox - not trivial...
16:56:23linuxstb(if you wanted all of Rockbox to just use that part of the LCD)
16:56:29soapas there is no existing framework to rework any part of the user interface outside the WPS.
16:56:58mcuelenaereGrahack: that part should probably get reversed then
16:58:04***Saving seen data "./dancer.seen"
16:58:52mcuelenaeren1s: shouldn't non-Rockbox code not get changed? (RE your strncpy commit) It seems to've broken Lua
16:59:07funmanand here we go, i mailed this product marketing manager at AMS again
17:05:31linuxstbmcuelenaere: As strncpy is in the plugin lib, I think it makes sense to revert for lua...
17:05:46mcuelenaerelinuxstb: yep, I'm going to do that
17:07:12gevaertsbrokenscreen: you're probably better off using voice navigation
17:07:13 Quit evilnick ("Page closed")
17:07:28brokenscreennice idea :)
17:07:36brokenscreenthanks i will try that :)
17:07:44 Quit funman ("free(random());")
17:07:56muesli_petur thx 4 help, bye @ll
17:08:44brokenscreenthx 4 help making my ipod blind now :)
17:09:19 Quit brokenscreen ("CGI:IRC")
17:09:29 Quit muesli_ ()
17:11:29 Join BryanJacobs [0] (n=bryanjac@
17:11:57gevaertsZagor: maybe an idea for the future : if you send ulspeed to the server, it could conceivably be used for allocation purposes as well
17:12:41Zagoractually we already record actual upload speed
17:12:57oboZagor: are there still plans to add SSL and/or (verified) usernames/passwords to the new build system?
17:12:58gevaertsah, true. Probably more reliable
17:13:00Zagorand yes, it could be used. especially to avoid re
17:13:18Zagorreally slow uplinks getting big builds
17:14:04Zagorobo: yes, but not today. we'll start with a free-for-all.
17:14:16BryanJacobswhat are the current build clients like?
17:14:22CIA-69New commit by Ubuntuxer (r21881): Fix a bug in lib display_text.h, which inserts a unwanted blank line
17:14:29saratogaZagor: I'm curious if you think its worth using 7zip for uploads? the builds are MUCH smaller in that format
17:14:36ZagorBryanJacobs: in what respect?
17:14:41BryanJacobshardware specs
17:14:43Zagorsaratoga: it probably is, yes
17:14:47saratogathough it would require substanital time for the server to repack to zip
17:14:51ZagorBryanJacobs: _very_ varied
17:15:03Zagorsaratoga: aha, is it that much slower?
17:15:15gevaertsBryanJacobs: they range from 8-core xeon machines to arm :)
17:15:40saratogaZagor: 7zip (de)compression is quite fast, but you still have to rezip them on the other end since we distribute zips
17:15:44ZagorBryanJacobs: the new system is designed to accomodate any speed. from very slow to very fast.
17:15:51CIA-69New commit by Ubuntuxer (r21882): Tiny bug fix for help text in pegbox
17:15:54BryanJacobswould a dual Xeon 3060 with a 100Mbps uplink, 2TB of hard drive space, and 3GB of RAM be helpful?
17:16:03 Quit BCM43 (Remote closed the connection)
17:16:10gevaertsyes, very
17:16:14ZagorBryanJacobs: yes. every client is helpful. and such a beast is very much so.
17:16:47Zagorsaratoga: ah, true. I'll guess we'll wait a bit with 7zip
17:16:58CIA-69New commit by mcuelenaere (r21883): Revert r21863 partly: fixes Lua
17:17:13BryanJacobsI also have a dual Xeon E5310 kicking around
17:17:21saratogaits a shame we can't just use 7z for the downloads, zip seems to work particularly poorly for rockbox for whatever reason
17:17:28BryanJacobsalso with 100Mbps uplink
17:17:56BryanJacobsbuild crosscompilers now :-)
17:18:49BryanJacobswhy build after every commit - why not do bisection only in the event that something comes up red in a daily build?
17:19:25Zagorbecause we can :-) and we like to get fixes done as soon as possible.
17:19:34gevaertsby that time whoever made the mistake has gone to sleep
17:20:02BryanJacobsw/e, my power is included so no big deal
17:20:03Zagorshortly we'll be getting the CIA bots to yell here in the channel when a build fails
17:20:22gevaertsZagor: make sure to rate-limit that :)
17:20:50Zagorwell as long as the builds take a few hundred seconds each, I'm sure we're pretty safe from spamming :)
17:21:09BryanJacobswhere can I see a list of build clients?
17:21:33Zagorthere's no such list. clients come and go as they choose.
17:21:55Zagorhere's a list of those who participated in the build of r21877:
17:22:15BryanJacobshm. this is also rather insecure.
17:22:39BryanJacobsI mean, any rockbox dev can run arbitrary code on the build clients, and in addition the current implementation is vulnerable to a MitM
17:23:21soapWhat sort of SVN commit would run arbitrary code on a build client?
17:23:28Zagoryes it is. we'll be straigtening it up a bit later, but we're too impatient to wait for it :)
17:23:42BryanJacobsbut it'd be so easy to gpg-sign revisions...
17:23:43dionoeasoap: a makefile change for example
17:23:44saratogayou could edit the make files or configure script to run whatever you wanted
17:23:50gevaertsyou can run things in a VM if you like.
17:23:52Zagorbut in the end it will always be inherently unsecure as long as we build things from svn
17:24:06Zagoryes chroot or vm is a good idea
17:24:09BryanJacobsgevaerts: I think I will, I've got Xen already running VMs
17:24:12saratogabut honestly unless you're running this thing as root its probably not a huge deal
17:24:25BryanJacobsfilling up the hard drive would be bad enough
17:24:31BryanJacobseven /tmp can do that
17:24:48gevaertsBryanJacobs: in that case I think it's a no-brainer. That also seriously reduces the risk of accidentally breaking the build environment for the client
17:25:10BryanJacobsyep, a full VM sounds like the way to go
17:25:15BryanJacobschroots aren't good enough
17:25:26saratogabut yes some sort of secure communications would be nice so that I don't have to worry putting this on school machines and such
17:25:31BryanJacobsI'll even give it a dedicated IP address, Because I Can (TM)
17:25:43gevaertsBryanJacobs: now that *is* overkill
17:26:12BryanJacobsI have one anyway - so why not? No packet processing in the Dom0 this way
17:26:25BryanJacobsotherwise I'd have to NAT it
17:28:48Zagorsaratoga: secure comm doesn't help. any dev can still change the rbclient script.
17:29:31gevaertsZagor: not really. Update commands are for a specific revision
17:29:48Zagorgevaerts: ok, so change a Makefile then
17:30:17gevaertsthose are the real risk, yes, unless you personally turn evil :)
17:30:26saratogai'm not too concerned about developers
17:32:22 Quit toffe82 (Read error: 104 (Connection reset by peer))
17:32:36Zagorsaratoga: so basically you just want svn over ssl?
17:32:55 Join evilnick [0] (i=0c140464@gateway/web/freenode/x-0e567c69905fbac3)
17:33:02*BryanJacobs thinks that it should be svn over ssl so that nobody EXCEPT rockbox devs can inject code into the machine
17:33:10saratogasecure communication with the build master, yeah
17:33:10BryanJacobsotherwise, anybody can
17:33:28BryanJacobsalso it'd be nice to have verifiable revision signatures
17:33:44BryanJacobskinda like monotone enforces
17:33:51 Quit flydutch ("/* empty */")
17:33:54saratogai don't know enough about networking to comment on the best way to do that though
17:34:08BryanJacobsand while we're on the wishlist, how about a DVCS instead of SVN?
17:34:11ZagorBryanJacobs: does svn support that?
17:34:38ZagorBryanJacobs: we have git-svn support
17:34:42BryanJacobsZagor: svn has no built-in support for signed revisions. But you could still do a GPG signature on the manifest which would do the trick
17:35:04Zagorthe manifest? pardon my ignorance :)
17:35:07BryanJacobswhere is this mythical git?
17:35:17saratogacheck the wiki
17:35:25BryanJacobssorry, there's a unique identification for each revision
17:35:26saratogatheres a page explaining how to use it IIRC
17:35:46BryanJacobsif you sign that the whole revision is essentially verified provided that the hash is secure
17:37:42BryanJacobseww, this git-svn thing is icky
17:38:03saratogadon't use it problem solved
17:38:45*BryanJacobs wants to use HG
17:40:33 Quit Ubuntuxer ("Leaving.")
17:41:03 Join BCM43 [0] (
17:41:32BCM43why does rockbox not ignore the preceding "the"?
17:41:49gevaertsBCM43: because "the" is a drink in many languages
17:41:59BCM43gevaerts: really?
17:42:30ZagorBCM43: every "why does rockbox not" question generally has the same answer: because nobody has implemented it
17:43:03gevaertsBCM43: also, where do you sort things like "The The"?
17:43:22BCM43gevaerts: under the, you only ignore the first one.
17:43:46gevaertsBCM43: should it also ignore "le", "die", ...?
17:44:35BryanJacobsideally it would be based on the language of the file name
17:44:35BCM43gevaerts: you could set it to do or not do that, or use the system language
17:44:47saratogai always thought the simplist option was to pick a half dozen of the most common articles, always ignore them, and wait for this to annoy someone enough that they implement a language independent system with the lang files . . .
17:44:57BryanJacobsunfortunately that's not really feasible without some heuristics, so system language setting is a good second place
17:45:18BryanJacobssaratoga: I'm with you :-)
17:45:20*gevaerts thinks that there is no halfway good way to do this
17:45:22BCM43saratoga: lol, annoy people till they fix it.
17:45:33gevaertsBCM43: you *can't* fix it
17:45:55oboThere is a patch to use artistsort tags in the DB: FS #7287
17:45:58BryanJacobsgevaerts: well, first build a database of every song ever written (and those that have yet to be written)...
17:46:43BryanJacobsfrom a theory perspective, if we had a black-box that accepts a song name and outputs the language in which it's written, this is a solvable problem
17:46:55BryanJacobsprovided that there are no words that are articles or not depending on context
17:47:34*gevaerts thinks that sort tags are the right way to solve this
17:47:48saratogajust because you can't make a solution that works for 100% of all possible inputs, doesn't mean you acn't make one that works for all common inputs and fails harmlessly on the others . . .
17:48:09BCM43BryanJacobs: I geuss you could just chage the band name to Doors, the
17:48:17gevaertssaratoga: I'm just not thinking about this from an anglocentric point of view
17:48:37saratogagevaerts: I don't think i'm being anglocentric here
17:48:48BryanJacobswe can just throw up our hands if the tag is encoded in UTF-foo
17:48:50saratogawe could easily enough make the article list be present in the lang file for instance
17:49:00saratogaits only centric to languages that have articles
17:49:15saratogawhich at least would include french, spanish and german
17:49:27gevaertssaratoga: that would work if articles in one language wouldn't be nouns or verbs in another
17:49:32BryanJacobssaratoga: what if I have my spanish songs pop (Paulina Rubio) on my english-language player?
17:49:57BryanJacobssong tile "La Costa Verde" wouldn't be sorted as "Costa Verde, La"
17:49:58saratogaagain, you can't be right everytime but you can be right most of the time for most people
17:50:10saratogathats why you make it a setting
17:50:35saratogaand let people use funny file name, sort tags, or whatever if they can't manage
17:50:47amiconn[17:17:21] <saratoga> its a shame we can't just use 7z for the downloads, ... <== Why is that?
17:50:57 Quit Windlord ("")
17:51:03saratogaamiconn: well I guess we could :)
17:51:12saratogathough i bet we'd have people complaining about it
17:51:12 Join bertrik [0] (n=bertrik@
17:51:33BryanJacobsI dunno... 7z is about the best free compression program for Windows
17:51:41BryanJacobsand it's readily available on all major Linux distros
17:51:57saratogai'm sure we'd annoy Mac users though who tend to use oddball programs
17:52:04amiconnsaratoga: It would probably be a very good idea if rbutil would support 7zip
17:52:12saratogaof course they're still welcome to use rbutil . . .
17:52:21BryanJacobsno kidding about Mac users. Cyberduck? Who came up with that name?
17:52:25saratogahmm maybe not such a bad idea afterall
17:52:54saratogamake sure rbutil works with 7z, tell users to use rbutil, or if they want to do manual installs get 7z support
17:53:08BryanJacobsthat doesn't sound half bad
17:53:24BryanJacobsthe nontechnical people use rbutil anyway, and the technical ones have no problem decompressing a 7z archive
17:53:28saratogawould save us a lot of bandwidth
17:53:52 Quit Zagor ("Don't panic")
17:53:53*BryanJacobs has a few TB of hosting bandwidth to spare
17:53:54BryanJacobsyou need some?
17:54:14saratogayou could probably setup a download mirror if you wanted to
17:54:33BryanJacobshow much traffic do you guys get in a month?
17:56:51saratogahundreds of thousands of .rockbox downloads
17:57:19saratogawhich incidently means 7zip woudl save us hundreds of GB of bandwidth a month
17:57:30 Join webguest49 [0] (n=400789a2@gateway/web/cgi-irc/
17:57:40 Quit webguest49 (Client Quit)
17:57:47BryanJacobshundreds of thousands, eh?
17:57:53*BryanJacobs goes to do some math
17:58:24gevaertsBryanJacobs: have a look at
17:58:28at0mrasher: i just launched
17:59:01gevaertsthis doesn't include mirros
17:59:36BryanJacobsthis isn't bad, that's less than a terabyte of traffic total
17:59:57BryanJacobsI have 3TB to spare per month
18:00:41BryanJacobsgive me a week or two to set up a dedicated Rockbox VM
18:00:52gevaertsBryanJacobs: have a look at Those people might want some help with hosting
18:01:27BryanJacobshosting a converted version of wikipedia?
18:01:36BryanJacobsI already have an XML mirror
18:02:15gevaertsthe plugin isn't ready for commit yet, but if it gets in, lots of people will want a pre-converted version
18:02:20BryanJacobstorrents sound like a good idea for something like that
18:02:32BryanJacobslarge files are best served P2P
18:02:44BryanJacobsthe 600k rockbox build zips are not
18:02:47rasherat0m: it should let you in now. But it's not terribly important - gevaerts' host manages to keep up with the builds
18:03:24 Join JdGordon| [0] (i=ad81d397@gateway/web/freenode/x-4201b2af032c720e)
18:03:37gevaertsBryanJacobs: just suggesting things :)
18:04:12 Quit saratoga ("Page closed")
18:04:46at0mrasher: ah ok.. will let it run a couple days for a try. didn't know how much of a bottleneck building was
18:06:49rasherat0m: it might not let you in for another 30 minutes iirc the system caches the list or hosts to allow
18:07:05 Quit AndyIL ()
18:09:33 Join Strife89 [0] (n=michael@
18:09:36 Join funman [0] (n=fun@rockbox/developer/funman)
18:10:26funmanBryanJacobs: i was wondering if you have fixed yourself milestones for your gsoc project
18:10:56funmanto know if you have fullfilled all the requirements, and are now doing "extra work" outside the gsoc program
18:12:18BryanJacobsfunman: yeah, I did come up with milestones
18:12:30BryanJacobsI'm a little more than halfway done with them
18:12:57BryanJacobsleft would be seeking support, sound for the more esoteric (read: non-stereo) hybrid files, and getting a patch integrated
18:13:20funmannon-stereo as in mono or more channels?
18:13:27BryanJacobstake your pick?
18:13:27funmanor both
18:13:38BryanJacobsmono, 6-channel, and 32-bit?
18:13:50BryanJacobsI'll definitely do mono and maybe the other two
18:13:54BryanJacobsalso floating point
18:14:12funmanI don't know if there is a channel mixer already in rockbox
18:14:21BryanJacobsthere must be for the ac3
18:14:22pixelmagevaerts: HID less build seems to do the trick with my c250 on this Mac... I automatically get the two volumes when I connect to USB. Although compared to last week I also updated the bootloader, maybe that could be a reason why it works now too. Any ideas what to check?
18:14:25BryanJacobsit supports 5.1
18:14:48funmanBryanJacobs: what do you mean by "floating point" ? I thought no rockbox target had a fpu
18:15:14BryanJacobsWavpack has an integer floating point system
18:15:17 Quit Strife89 ("Vamoose!")
18:16:29funmanaren't "integer" and "(floating) point" mutually exclusive?
18:16:51funmanor do you mean a floating point representation in integer registers
18:16:52 Join _zic [0] (
18:16:53BryanJacobsno, it does no floating point operations (thus, "is an integer system") but represents floating point data
18:17:05BryanJacobsyou got it there on the second line :-)
18:17:12funmanthanks for the explanation
18:17:22funmanand good luck for your work!
18:18:10funmanwhen you are done with wavpack, i might have work for you related to buffering .. ;)
18:18:48BryanJacobsbut I'm already working on buffering...
18:19:00BryanJacobsone of the things I have to do is modify buffering to support two files at once
18:19:09funmani have started reading buffering.c/playback.c to fix problems on the Clip/m200v4/c200v2 (the ones with a tiny tiny audio buffer) but I didn't continue
18:19:17 Join AndyI [0] (i=AndyI@
18:19:33BryanJacobsI think I understand how they work, so feel free to ask any questions
18:19:41funmanI saw that, and thought now you would be more familiar with that code than most of us
18:19:55funmanthanks, i'll ping you when i open up those files again ;)
18:20:25bertrikfunman, I've been thinking a bit whether we could implement a kind of a continuous built-in-self-check
18:20:34 Quit petur ("work->....")
18:20:45funmanbertrik: what do you want to check exactly?
18:20:59bertriklike adding magic values between data that we check at each thread switch, so we can check if something is corrupting something else
18:21:26funmani don't think the problem has to do with corruption, more like an infinite loop when the buffer is nearly full, which can be aborted under certain conditions
18:21:28bertrikadd checks on missing initialisation / double initialisation on kernel/thread functions
18:21:31BryanJacobsthose would be called cookies btw
18:21:57BryanJacobsthey're used as part of many modern stack- and heap-protection schemes
18:22:09bertrikadd checks to some functions to see if they are called from interrupt context while not allowed
18:22:15funmanat least it is exactly what i saw on c200v2 when loading AA, not sure if the exact same problem affects the clip/m200v4 which don't use AA
18:22:51BryanJacobshow much memory do these targets have?
18:22:52funmanthere is already a cookie mechanism for threads stacks (the last entry of the stack = 0xdeadbeef)
18:23:07bertrikIIUC, the 'ticks tasks' run in interrupt context, but this may not always be obvious when browsing through the code
18:23:13funman2MB of SDRAM + 384kB of another kind of SDRAM (IRAM)
18:23:50BryanJacobsew. So maybe you should just eliminate the disk buffer and use only a PCM buffer? (as in, make the buffering calls just read from the disk directly and pass through into the codec)
18:23:57funmanbertrik: right, but afaik only the button reading is made in tick tasks?
18:24:22bertrikif you have to ask means you don't know ! :P
18:24:37 Join toffe82 [0] (n=chatzill@
18:25:04bertrikI wouldn't be surprised if a couple of these checks would already trigger if we added them to rockbox
18:25:14funmanBryanJacobs: right, nico_p even made a patch for this some time ago
18:25:36bertrikfor example, quite a few mutex locks were used uninitialised until recently
18:26:28LambdaCalculus37Heh... according to talk in #linux4nano-dev it turns out that the nano2g may have a similar, if not same, LCD as the Meizu M3. :)
18:26:50LambdaCalculus37linuxstb found a few addresses and other bits of info that indicate that.
18:26:53funmanwell there is much more tick tasks
18:27:07bertrikfunman, indeed the playback stops on the clip et al. seem to be too "clean" to be actual crashes
18:27:18funmanLambdaCalculus37: wasn't it already known that the lcd _screen_ is identical?
18:27:36amiconnBryanJacobs: Tests have shown that direct-read playback on flash targets still costs battery runtime compared to proper buffering
18:27:49amiconnNot as much as on hdd targets, but still significant
18:27:59funmanamiconn: i see no battery bench on fs#9332
18:28:04pixelmagevaerts: got to go now though - so any testing and checks need to be delayed
18:28:05LambdaCalculus37funman: Not to my knowledge, no.
18:28:18bertrikthe LCD controller addresses will be the same as they are part of the s5l8700 SoC
18:28:24LambdaCalculus37funman: But now I do know. :)
18:28:25bertrikLambdaCalculus37, would be nice if it could be re-used
18:29:14BryanJacobsamiconn: I can guarantee that would only be the case in the event of data reuse from the buffer (codec calls bufseek backwards)
18:29:34BryanJacobsfor linear-access files that CAN'T be true
18:30:00LambdaCalculus37bertrik: I think it can be done.
18:30:28amiconnBryanJacobs: It *is* true, and there's even an explanation
18:31:10BryanJacobsbut but but... if you have buffering, the cost to get a byte to the codec is cost(disk read)+cost(memory copy)+cost(memory read) - with direct-from-disk it's cost(disk read)+cost(memory read)
18:31:32TorneBryanJacobs: but codecs don't read in conveniently flash-block-aligned increments :)
18:31:46BryanJacobsI've got it. Thanks, Torne
18:31:50amiconnMost (all?) flash storage devices auto-sleep after a few milliseconds without access. So if you access it constantly, reading only small blocks at a time, it will never sleep
18:32:01Tornealso that
18:32:04BryanJacobsand write batching, that too
18:32:07Torneconstantly opening pages, even different pages, is expensive
18:32:12amiconnBut if you do proper buffering, you read a large chunk, then let it sleep
18:32:20BryanJacobsI was just thinking of the flash chips as true random-access devices
18:32:43funmanamiconn: then perhaps we can tweak the buffer size if we know this auto-sleep time
18:33:05BryanJacobsfunman: "as large as possible" is the global optimum
18:33:18amiconnThe tweak is to make it as large as possible, like on hdd targets
18:33:34amiconnIt's the same effect, only to a lower degree
18:33:52bertrikI can imagine there's some kind of knee, above which any improvement is marginal
18:34:03amiconnYes, sure
18:34:25BryanJacobsnot if the flash chip staying asleep matters...
18:34:25bertrikand my guess is that it's a lot smaller than the full RAM
18:34:25Tornebertrik: yes, but that's also true of hdd targets, no?
18:34:26amiconnTypical auto-sleep times seem to be around 20..50ms
18:34:33BryanJacobsthe larger the buffer the lower the active/inactive ratio
18:34:43Tornebertrik: which is why the 64mb ipod video doesn't get significantly better battery performance than the 32mb version if you normalise for the battery capacity
18:35:18BryanJacobsTorne: that might also be because we don't buffer up a user's whole playlist?
18:35:24Torneyes we do
18:35:40BryanJacobsmy "Woman in White Suite" is larger than the whole buffer
18:35:41amiconnTorne: Someone should test that using lossless audio files on a 64MB ipod. I'm sure the difference is measurable
18:35:45Torneon the ipods it's because if the buffer is twice as big it takes twice as long to fill
18:35:59Torneeventually the power drawn by the actual transfer dominates instead of the spinup power
18:36:14Torneit's still an improvement, but much less than it is at lower memory sizes
18:36:15amiconnThe power drain by the actual transfers is the same
18:36:26Torne..uh, yes
18:36:27Tornehang on
18:36:29BryanJacobsgotta love asymptotic behavior
18:36:31*Torne pokes self
18:36:53Torneamiconn: why would lossless make any difference?
18:37:11amiconnBecause it consumes the data much more quickly
18:37:45amiconnTorne: There is *one* usage case where a very large buffer is actually worse than a smaller one - if the user is undecided what to listen
18:37:53Torneamiconn: yah, that's true
18:38:08*Torne is only considering battery benches, really, rather than reality :)
18:38:16Tornei can't see how the codec matters, though
18:38:32Tornei would expect the same relative difference between 32/64mb regardless of codec as long as the playlist is bigger than the buffer
18:38:52Torneas far as i've seen from mine and other people's benches, the difference is 50%
18:39:03Tornewhich is the difference in stock battery capacity between the two models
18:39:12amiconnIt doesn't matter whether the playlist is bigger or smaller than the buffer, due to the way rockbox buffering works
18:39:18BryanJacobsI wouldn't expect the same relative difference, because the ambient consumption matters more if that data are consumed more quickly
18:39:39TorneBryanJacobs: well yes, sorry. you would have to also normalise for playback current usage
18:39:42soapcodec matters in that the effect of the large buffer is likely hidden behind the rather disproportionately large "static" consumption of all the common components.
18:39:58amiconnRockbox buffering doesn't go back. If the playlist is smaller than the buffer and repeat is enabled, it will buffer another copy of the playlist after the first, and then another one etc
18:40:16soapa low CPU / high disk consumption pattern would be needed to see a significant variation in power consumption.
18:40:17Torneamiconn: ah, didn't know that, and all the batter ybench instructions explicitly say to use one larger than ram, so i assumed it was
18:40:24BryanJacobswhich is one of the things I'd like to fix, but OK...
18:41:06Tornesoap: yes, sorry
18:41:08amiconnYeah, that's just to play safe, in case someone forgets to change the instructions when changing buffering behaviour
18:41:08Torneyou are right
18:41:09soapsince EverythingButTheCPU appears to be eating most the battery anymore.
18:42:01amiconnHigher bitrate will cause more freqquent spinups, so the disk access becomes a larger fraction of overall power consumption.
18:42:26*Torne blames his inability to do arithmetic :)
18:42:31*Torne can only do math
18:43:36amiconnJust imagine you have an album that is exactly 4 times the size in lossless (flac) as it is in lossy (mp3), and that album fits exactly into the buffer on a 64MB ipod
18:43:37BryanJacobsor you could look at it as "the access rate from the disk averages to the bitrate of playback, so higher bitrate means more power drawn by the disk"
18:43:41BryanJacobseven with no cost for spinning up
18:43:52Torneamiconn: yesyes, i see it now
18:44:03 Quit mcuelenaere (Remote closed the connection)
18:44:13Tornethough for the ipodvideo specifically the effect might still be small compared to the fact that the 64mb version has a 50% larger battery :)
18:44:17amiconnSo on a 64MB ipod you would have 1 spinup perhour (assuming the album plays 60min), on a 32MB ipod you would have 2 spinups per hour
18:44:24Tornewhich is a pretty hard improvement to beat
18:44:36amiconnFor lossless this would mean 4 spinups per hour on 64MB and 8 spinups per hour on 32MB
18:44:38 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
18:44:57BryanJacobsa compelling case for using a lossless codec over raw PCM! :-P
18:45:09TorneBryanJacobs: as if there weren't cases fo rhtat before? :)
18:45:18BryanJacobsI saw a retarded thread the other day about it
18:45:49BryanJacobs there we go, for a laugh
18:46:30Tornei guess if you had a flash based player with a large enough ram the cpu cost of decoding the codec might actually make it more power efficient to just buffer pcm, since the difference between idle and active flash consumption could be less than the difference between boosted and unboosted playback
18:46:40amiconnBryanJacobs: Btw, if I did understand the explanation of hybrid wavpack properly, it should be possible to reconstruct a lossless stream from lossy+correction without full decoding
18:47:01amiconnI'm not 100% sure - this is a kind of math I don't understand well
18:47:05Tornebut if yo uhave a flash based player you probably don't have room for pcm files :)
18:47:07BryanJacobsamiconn: it almost is full decoding, you're only skipping the entropy step
18:47:20BryanJacobsit's certainly NOT something you want to do inside the buffering thread
18:47:34BryanJacobsunless that thread runs on a COP or something
18:48:23soapTorne, in that case you should likely decrease the CPU frequency at boost, or add a third frequency as you're apparently wasting cycles when at full boost - else you're consuming the same number of cycles to get compressed->uncompressed regardless of when you do it - be it upon buffering or later.
18:48:52ej0rgeBryanJacobs: what's depressing is not a month goes by on head-fi when you don't see someone questioning whether lossless is really lossless
18:48:54amiconnIiuc lossy+correction is produced by quantizing rice codes. So that step should be reversable without decoding
18:49:18Tornesoap: that would still have such a point, jus tlater. If your boost freq was exactly the rate required to, say, decode flac in realtime then it's sitll possible for boosting the cpu to be more expensive than reading from flash, in theory
18:49:26rasherat0m: errr... your binsize client is broken
18:49:35Tornewhether this is likely on any particular hardware, i have no idea. probably not.
18:49:38BryanJacobsej0rge: I wrote a program which sample-for-sample compares WAV files, I don't suppose that would be enough to convince people?
18:49:39Tornebut it's not impossible
18:49:41ej0rgeBryanJacobs: too much magical thinking at that site, and i get so tired of telling people that flac sounds exactly like pcm for the same reason that the math in a spreadsheet isn't corrupted by putting it in a RAR file
18:50:26soapthe modern audiophile claims that the CPU load being higher during lossless is what accounts for the audible difference.
18:50:32bertrikit could be that the processor leaks noise into the audio
18:50:38soap*compressed lossless
18:51:09ej0rgesoap: or they raise the bogeyman of 'jitter'
18:51:16 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
18:51:35*BryanJacobs facepalms
18:55:15 Quit JdGordon| (Ping timeout: 180 seconds)
18:58:08***Saving seen data "./dancer.seen"
19:00:01dzsoap: we clearly need to do some proper double-blind testing to make said audiophiles look like clowns
19:00:46 Join Llorean [0] (
19:00:53linuxstbbertrik: I'm assuming from reading the drivers that we don't know the different lcd controllers in the Meizu M3?
19:01:48bertriklinuxstb, I know next to nothing about the meizu lcd drivers
19:02:00bertrikmaybe markun does
19:02:01linuxstbDenes isn't around any more?
19:02:30bertrikdenes signed off 4 months and 13 days ago
19:03:10n00b81He left the project?
19:05:25 Join liqix [0] (n=user@
19:07:41 Part liqix ("ERC Version 5.2 (IRC client for Emacs)")
19:13:47funmanstill can't get SD status register transferred even after correcting my code .. :(
19:15:18 Join Lynx_ [0] (
19:22:27 Quit BryanJacobs ("Java user signed off")
19:23:23funmanif you want to help :
19:25:39 Join JdGordon| [0] (n=Miranda@
19:26:20 Quit itcheg (Ping timeout: 180 seconds)
19:26:43 Join M [0] (n=7bf3fc5d@gateway/web/cgi-irc/
19:29:48 Quit M (Client Quit)
19:29:52 Join M [0] (n=7bf3fc5d@gateway/web/cgi-irc/
19:31:23linuxstbbertrik: BTW, the Nano2G seems to have the LCD controller at a different address to the 8700 - it's at 0x386000xx instead of 0x3c1000xx
19:32:00 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
19:32:41 Join itcheg [0] (i=4117734b@gateway/web/freenode/x-3fcdec5575013c45)
19:32:42bertriklinuxstb, ok, interesting, didn't expect that
19:33:02linuxstbNo, but it looks like TheSeven has confirmed it - he's making the LCD do things now...
19:33:26TheSevenred/green/blue screen works
19:33:37TheSevenblack/white and all that, too, of course :-P
19:34:00bertrikvery cool
19:34:16gevaertsTheSeven: can you look at the meizu M6 LCDs one of these days? ;)
19:34:43TheSevenwell, depends on what you mean by "look"
19:34:58gevaertsI mean "make them work" of course!
19:35:20 Quit M ("CGI:IRC (Ping timeout)")
19:36:02itchegQI have a Ipov video, i'm using a build from abt a month ago and when I use it with a FM tranmitter via the dock (using IAP) the info for next song shows in giberish
19:36:04TheSevenwell, i don't have such a thing...
19:36:35itchegwhen I use it with out the FM transmitter all shows fine
19:37:00itchegThis is what I'm using for next %s%alNext: %Ia;%t4%s%alNext: %?It<%It|%Fn>
19:37:55itchegany ideas on why using the IAP would cause this?
19:38:02LambdaCalculus37linuxstb: What builds so far with the nano2g code in the trunk?
19:38:28linuxstbLambdaCalculus37: The bootloader builds, but I haven't tested it, or checked that any of the code shared with the Meizu will work...
19:38:40 Quit Grahack ("Leaving.")
19:38:56linuxstbI should be able to spend some time on that over the next few days, and get an lcd driver working...
19:39:10 Join saratoga [0] (i=9803c6dd@rockbox/developer/saratoga)
19:39:37LambdaCalculus37linuxstb: I can try sending a bootloader to the nano2g using iBugger and see what it does. Hopefully we get something.
19:39:39bertrikI tried to take care to append -s5l8700 to any source file strictly related to the s5l8700, and -meizu to source files that do something meizu-specific
19:39:55linuxstbLambdaCalculus37: You won't see anything...
19:40:02LambdaCalculus37linuxstb: Ahh, right, right.
19:40:17linuxstbbertrik: First thing will be to add a s5l8701 define - as registers look to be in different places...
19:40:22LambdaCalculus37linuxstb: You also added a tool to the trunk for working with the nano2g, didn't you?
19:40:44linuxstbYes, it's "bin2note" - the same as my "bin2htm" included with iBuggerLoader.
19:41:01 Join mc2739 [0] (
19:41:05linuxstbYou can use that (and the Makefile) to turn "loader.asm" into "loader.htm" if you so desire.
19:41:25 Join QncfOpO [0] (
19:41:26bertriklinuxstb, yes, it would be cool if we could share drivers even if they are at different addresses
19:41:59linuxstbbertrik: Maybe it will be as simple as defining things like "LCD_BASE", and specifying the registers relative to that.
19:42:14linuxstbBut I guess we'll find out...
19:42:28linuxstbI expect we can even share the M3's lcd driver...
19:42:52mc2739funman: should line 133 be mutex_unlock(&sd_mtx); instad of mutex_lock?
19:43:13linuxstbAlthough there are two LCD variants on the Meizus?
19:43:30 Join evilnick_7 [0] (i=0c140464@gateway/web/freenode/x-cf574c5f9e7b90cc)
19:43:31gevaertsthere are indeed
19:43:46linuxstbLovely #ifdef hell...
19:44:09funmanmc2739: right..
19:44:22soapDo you wish to discuss the user-list here, Llorean?
19:44:49linuxstbfunman: You said earlier that it was known that the Nano2g had the same LCD as the Meizu?
19:44:51evilnick_7What is the objection to having a moderated (user) mailling list?
19:45:12saratogamoderated as in how?
19:45:31funmanlinuxstb: i remember reading that someone had replaced a nano2g screen with a meizu screen (or the reverse)
19:45:32Lloreanevilnick_7: If a new list needs started, its an unofficial one to be unmoderated.
19:45:34linuxstbevilnick_7: I would guess the objection would be that someone would have to modify it...
19:45:48bertriklinuxstb, the m3 comes with either a wolfson or a philips codec, the OF supports both by what seems to be an if-statement for each time the codec has to do something
19:45:59itchegsaratoga any ideas? (I think you commited the IAP patch)
19:46:02linuxstbfunman: Ah, interesting.
19:46:07evilnick_7Good question; I guess moderated as in a certain group of people can approve messages as they are sent in??
19:46:11Lloreansoap: So the suggestion is "if you're having difficulty solving the problem, give up"
19:46:12soapFor I go back to my primary point: Repeating a behavior expecting different results is insane. So long as you narrowly define the possibilities so as to justify the behavior you repeatedly engage in you are guaranteeing the same (I would argue harmful) results.
19:46:14funmani can't find it again on google though
19:46:24Lloreansoap: OFFER POSSIBILITIES.
19:46:31soapI have offered three
19:46:35soapstop shouting, please.
19:46:39LloreanList them again, I've missed them clearly.
19:46:46soap1 - Moderate the list.
19:46:49LloreanI acn't do that.
19:46:53soapwhy not?
19:47:09LloreanBecause that's not something I have access to do?
19:47:19saratogadelete the list, use the forums instead
19:47:35soapI never said "Llorean, you make the list moderated"
19:47:43saratogaitcheg: I committed it but I didn't write it and I don't have an ipod
19:47:57soapThe only Rockbox user mailing list run by CAN be moderated.
19:48:00Lloreansoap: Well, we were discussing what *I* should or should not be doing, or at least you kept directing a lot of "you" at "me"
19:48:38soapIt appears to me you would be unwilling to accept your powerlessness over the problem.
19:49:12LloreanAccording to what?
19:49:17soapClearly the "need" for these long emotional threads has not diminished over time. I therefore would say the success rate of current policy to be a big fat zero.
19:49:25 Quit funman ("free(random());")
19:49:25saratogaitcheg: I think the usual response though to devices that don't work with the IAP is to either investigate whats missing and implement support for whatever your device needs from it, or wait for someone else to take an interest in the problem and do it for you
19:49:25Lloreansoap: that's dumb, honestly
19:49:36LloreanHave you seen me respond to any of the half dozen top posters since I left?
19:49:48LloreanNo, I'm only responding to the people who've responded to this one thread I created upon leaving.
19:49:53gevaertsitcheg: do you mean the text is wrong on the player?
19:49:54soapGo ahead - continue to piss off the masses in your self-righteous crusade.
19:50:00LloreanAnd that's because they are already engaged in a conversation with me.
19:50:14soapyou didn't have to respond. You could leave their sick behavior to be their own problem.
19:50:34Lloreansoap: You honestly think I feel *superior* because of this?
19:50:42soapdid I suggest you did?
19:50:49LloreanYou called me "self righteous"
19:50:56itcheggevaerts: the issue is that for next song it works under normal circumstance but if I use an fm transmitter via the dock is shows gargles text
19:51:04LloreanSo, yes, you said so at least. I guess you may have mis-used it.
19:51:07soapwhat is your motivation?
19:51:10gevaertsitcheg: on the player?
19:51:26itchegipod video 64 (5.5 gen)
19:51:31Lloreansoap: Well, prior to leaving, it was "nobody else enforces them, we need to either remove them or do so"
19:51:49gevaertshm, that really shouldn't happen, regardless of whether the transmitter works or not...
19:52:00 Join M [0] (n=7bf3fc5d@gateway/web/cgi-irc/
19:52:17itchegthe interesting thing is that I had used the patch with my own build before it was committed and it worked fine
19:52:35Lloreansoap: The easiest solution is to simply remove the rule about top-posting.
19:52:40soapwhat you just said in quotes is the sort of black-and-white thinking which leads, IMHO, to these threads.
19:52:46 Join n1s [0] (n=n1s@rockbox/developer/n1s)
19:53:16Lloreansoap: So what, we should pick and choose who's allowed to top post to maintain a nice healthy shade of gray?
19:53:33LloreanYou're using hyperbole again, rather than suggesting changes.
19:53:39gevaertsitcheg: I can't really test this (I don't have the needed toys). Can you submit a bug report?
19:53:39soapIf by correcting every single misstatement in the thread under question you believe you accomplish anything except perpetuating the sick discourse, you are mistaken.
19:54:13soapI call it self-righteous as there appears to be no error, no matter how small, you are willing to let go uncorrected.
19:54:13itchegsure, but need to make an account...
19:54:29itcheglet me try with latest build first
19:54:55soapThere seems to be an amazing inability to simply walk away from a conversation until the other side has either left or submissively appeared to agree to your points.
19:54:57Lloreansoap: That would be something else, since self-righteous contains a rather different meaning that that.
19:55:33LloreanI'll also walk away when the other side says "I disagree, but have no new points to offer on this" or when I personally feel I've run out of new points to offer.
19:55:36itcheggevaerts: is there anything wrong with this: %s%alNext: %Ia;%t4%s%alNext: %?It<%It|%Fn> that could be causing something like that
19:56:05JdGordon|itcheg: IIRC you cant put scrolling in sublines?
19:56:06gevaertsitcheg: I'm not too familiar with wps syntax, but I wouldn't think so
19:56:19soapreally? What new facts did you bring to the table in your most recent post to the user mailing list?
19:56:31soap(the reply to DavidD)
19:56:40gevaertsAt least, whatever you do wrong there (if any) it shouldn't cause the behaviour you're seeing
19:57:17Lloreansoap: Well, the fact that I've never banned anyone from the list for one.
19:57:29soapValuable information there.
19:57:29 Quit M ("CGI:IRC (Ping timeout)")
19:57:40soapWorth continuing a public argument I'd say~
19:57:52LloreanOh, yes, I should instead let someone spread a lie about what's actually happened?
19:57:53soapCure is worse than the disease.
19:58:00soap"spread a lie" -
19:58:08LloreanYou think it's truth that I've banned people?
19:58:09soapThat, in a nutshell, is the problem again.
19:58:47soapYou honestly believe that DavidD was spreading lies that you kick people?
19:59:01LloreanI honestly believe that DavidD posted to our public list with information that was wrong.
19:59:10LloreanAnd that if other people didn't know better, they could believe it was true.
19:59:15soapThere isn't a _chance_ that the line of his you quote could not be taken something less than 100% literally?
19:59:28saratogaI don't think arguing about people's arguments is productive
19:59:37Lloreansoap: What's it supposed to be taken as, exactly?
19:59:40saratogaeither propose a solution to whatever problem you percieve or let it go
19:59:49Llorean"He said I've banned people, but he really meant I've asked them politely not to top post"?
19:59:50 Quit n1s (Read error: 104 (Connection reset by peer))
19:59:51soapSelf-righteous: Your responsibility to correct every little factual error.
20:00:01LloreanThat's not what self-righteous means.
20:00:08soapsaratoga, I have proposed three solutions.
20:00:14soap1 - make the list moderated.
20:00:21LloreanAnd frankly, the idea that I've been banning people from the list is not a *little* error
20:00:27soap2 - try someone other than Llorean dealing with problem users.
20:00:28evilnick_7"Being a dick to people then kicking out anybody that complains is not an acceptable way to interact with a community." Is not explicitly saying that anyone is doing/has done this.
20:00:38Lloreansoap: I've stopped dealing with problem users anyway.
20:00:45soap3 - take the discussions of, and handling of, problem users off-list.
20:01:31saratogaall of those presume that the list is kind of useless as a way to discuss things
20:01:47soapand I think a strong case can be made that it is.
20:01:48saratogain that they discourage people from using the list to discuss things
20:01:59saratogatake this logic to its conclusion
20:02:00evilnick_7saratoga: It pretty much is, the way things are right now.
20:02:21soapAs these threads continue, and continue to be a sizable percentage of the total list. If they were "useful" or "functional" the need for them would cease.
20:02:22 Join M [0] (n=7bf3fc5d@gateway/web/cgi-irc/
20:02:41Lloreansoap: As to #3, have you actually tried that?
20:02:53soapyes, I email about two people a week.
20:03:16soapI have no powers to remove anyone from the list - so I simply point out the rules, do not threaten.
20:03:34LloreanI've never threatened.
20:03:36saratogai don't think any of those are likely to work, but I like #1 best since in practice it would mean that most people wouldn't use the list
20:03:41soapdidn't suggest you had.
20:04:32soapsaratoga, perhaps they won't work. I think it is reasonable, though, to say what is currently being done isn't working.
20:04:33Llorean#3 has never worked for me. Even a simple "We've asked that you please not top post to the list. If you're having difficulty, let me know what client you're using and I'll see if I can help" has gotten me extremely rude responses.
20:04:36MHey, could anyone tell me how I might add a virtual led toggle to the ui simulator?
20:04:46soapLlorean, I have NEVER gotten a rude response.
20:04:46LloreanAnd by never I mean, literally, never.
20:04:58soapno response? Often.
20:05:21LloreanI've gotten no response from people who've also not changed their behaviour.
20:05:36evilnick_7soap: Maybe you could let us know what you put in an example off-list message?
20:05:38 Nick M is now known as MHael (n=7bf3fc5d@gateway/web/cgi-irc/
20:05:42soapand justifying current behavior, if we take it as broken, as the only option when "no behavior" is always an option is also false.
20:06:01Lloreansoap: The option "no behaviour" should be rewritten as "drop the rule" then
20:06:11LloreanIf you aren't enforcing the rule, it doesn't really exist.
20:06:17soaplol, here we go again with taking it to extremes.
20:06:27LloreanHow is it not the case?
20:06:41LloreanIf you say "we've written this rule down, but will never enforce it" you're saying "this rule isn't really there"
20:06:42soapA holiday from an "enforcement" policy which doesn't work is hardly a suggestion to "drop the rule".
20:06:53LloreanHow is it a holiday?
20:07:14LloreanWhy *can't* we just drop the rule, exactly?
20:07:34LloreanIf nobody's willing to try to enforce it, why do we have it?
20:07:35soapYou are the one defining your current enforcement policy as the only possibility. A comfortable definition for it pretty much ensures the answer is "continue as before".
20:07:56Lloreansoap: Elaborate, please.
20:07:56soapRegardless of the utter failure of current enforcement tactics.
20:09:10LloreanYour suggestions for better policies are 1) Make the list moderated (but nobody has stepped up to do this in the last several years), 2- someone else do it (nobody has stepped up to do this, and I've left the job anyway so 2 is currently inaction) and 3- take it off list
20:09:12soapElaborate? Half of what you have said to me in this channel the last 20 (?) minutes and all that you have said to me in the mailing list is a narrow definition of the possibilities. All with the self-evident result that current policy = only possible policy = no need to change visibly broken ways.
20:09:31LloreanI've asked you time and again to offer broader definitions.
20:09:38soapof what?
20:09:46Lloreanyou're the one saying I've narrowed them
20:09:47LloreanShouldn't you know?
20:09:52soapYou dismiss my broader definitions before I even get started.
20:09:59LloreanGet started on what?
20:10:30soapA holiday from an "enforcement" policy which doesn't work is hardly a suggestion to "drop the rule".
20:10:44LloreanAgain - holiday how
20:10:53soapyou let the current thread die.
20:11:03LloreanYou're either not enforcing the rule at all, in which the rule is dropped, or you're replacing it with another enforcement policy.
20:11:05soapregardless of how many "wrong" things are said in it.
20:11:24LloreanWhat does the current thread have to do with the enforcement policy, at all?
20:11:58 Join _lifeless [0] (n=lifeless@
20:12:04soapno - you act much more pragmatically and realize that sometimes you cause your ultimate goal more harm by continuing to press your point (no matter how valid) in the face of resistance than to let all parties calm down and address it again at a later date.
20:12:15soapWhat is your goal?
20:12:28LloreanNow? Fuck all. I'm not talking to top posters.
20:12:56 Quit _lifeless (Client Quit)
20:13:52 Quit MHael ("CGI:IRC (Ping timeout)")
20:13:53LloreanSeriously, is this a discussion about what *I'm* doing, or about what the list needs to be doing.
20:14:16 Quit mc2739 ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060215]")
20:14:44LloreanIf it's about what I'm doing, you can stop now. I've stopped attempting to ask people to follow the rules in anywhere but this one thread, and you can either get it closed when it actually qualifies as no longer about Rockbox or the Rockbox list, or you can leave me alone because it's technically on-topic if those are on topic.
20:14:56LloreanIf it's about the list, do what you want with it. I left that to you guys quite some time ago.
20:18:25 Join MHael [0] (n=7bf3fc5d@gateway/web/cgi-irc/
20:20:05soapas of 40 hours ago you were still carrying on a conversation in defense of your handling of top-posters.
20:21:41LloreanAnd I didn't deny I was doing that.
20:23:41 Quit MHael ("CGI:IRC (Ping timeout)")
20:23:44 Join Thundercloud [0] (
20:30:53 Join MHael [0] (
20:31:23 Join funman [0] (n=fun@rockbox/developer/funman)
20:32:35MHaelHey, could anyone tell me how I might add a HDD activity virtual LED toggle switch to the UI Simulator?
20:32:40 Join efyx [0] (
20:33:24funmanamiconn: what's the point of declaring a const char testbasedir[] in test_disk.c when TESTBASEDIR could be used?
20:33:56funmani want to extend test_disk to test multiple drives
20:34:54 Quit BCM43 (Remote closed the connection)
20:35:18JdGordon|MHael: uisimulator/*
20:35:30JdGordon|cant really be more specific atm though :p
20:35:56MHaelI figured that much. :P
20:36:57JdGordon|umm... its not like the hold or usb faking which is pretty easy
20:37:15*JdGordon| isnt sure the best way to do it
20:40:54 Quit stoffel (Remote closed the connection)
20:44:59 Join stoffel [0] (
20:45:01MHaelI figure it's controlled by ata_idle_notify so if I could modify that to toggle on and off?
20:47:10JdGordon|its not, but you could toggle it there..
20:47:22JdGordon|but I assume you would actually want to control it somehow?
20:47:57 Quit stoffel (Remote closed the connection)
20:48:56JdGordon|you want this to make sure your WPs is good yeah?
20:50:22JdGordon|so my suggestino would be to use the %mh tag (hold button) instead of the diosk activity tag while you are testing
20:51:20MHaelI never would've thought of that
20:51:41JdGordon|or... work on the wps editor app :)
20:52:55MHaelI haven't even made a wps yet, this is my first. I haven't even put rockbox on my iPod.
20:58:12***Saving seen data "./dancer.seen"
20:58:43 Nick n00b81 is now known as rabbit (n=n00b81@unaffiliated/n00b81)
20:59:05 Nick funman is now known as something (n=fun@rockbox/developer/funman)
20:59:28 Nick rabbit is now known as n00b81 (n=n00b81@unaffiliated/n00b81)
21:00:25 Nick something is now known as funman (n=fun@rockbox/developer/funman)
21:01:10 Nick n00b81 is now known as n00b81_ (n=n00b81@unaffiliated/n00b81)
21:03:26 Nick n00b81_ is now known as n00b81 (n=n00b81@unaffiliated/n00b81)
21:04:01 Nick aditya is now known as adi|away (n=aditya@
21:06:28 Quit funman ("kick kick kick")
21:07:52 Quit antil33t ()
21:08:24LambdaCalculus37linuxstb: I build a nano2g bootloader. Going to try it out.
21:10:56LambdaCalculus37bin2note is complaining "Payload too big!" whenever I point it to the bootloader.bin file I just built for the nano2g bootloader.
21:11:31JdGordon|its not trying to add a bootloader splash image is it?
21:11:49linuxstbLambdaCalculus37: Yes, you don't upload it directly with bin2note - you use iBuggerLoader to load it.
21:12:17linuxstbLambdaCalculus37: But I'm not sure yet how to do that...
21:12:41linuxstbThere's at least one change in crt0.S that should be removed - switching the CPU into big-endian mode. It's four lines ending with the comment "// set big-endian"
21:13:16 Quit timc (Remote closed the connection)
21:17:11 Quit jfc (Read error: 54 (Connection reset by peer))
21:17:17 Join timc [0] (
21:17:33 Join jfc [0] (
21:18:03 Quit jfc (Read error: 104 (Connection reset by peer))
21:18:24 Join jfc [0] (
21:18:55 Quit jfc (Read error: 104 (Connection reset by peer))
21:19:16 Join jfc [0] (
21:19:47 Quit jfc (Read error: 104 (Connection reset by peer))
21:20:07 Join jfc [0] (
21:21:14 Join petur [0] (
21:25:09MHaelThanks for your help JdGordon|
21:25:12 Quit MHael ("ChatZilla 0.9.85 [Firefox 3.0.10/2009042316]")
21:28:19 Join BdN3504 [0] (n=5ce2257e@gateway/web/cgi-irc/
21:31:41BdN3504OK guys i have to bother you one last time about the gigabeat flashwriter plugin. i was now able to compile a build conatining that plugin and now the only question i have is to how it works: do simply execute it on target? do i have to put any of the bins provided on the patch tracker to the device as well, or will this plugin write to the flash autonomously without me providing any further files?also, i didn't
21:36:04 Quit robin0800 ("Don't follow me")
21:36:41gevaertsBdN3504: I think you got cut off. That line ends with "also, i didn't"
21:39:35 Join webguest93 [0] (n=5628a84d@gateway/web/cgi-irc/
21:41:42 Quit saratoga ("Page closed")
21:41:45 Quit itcheg (Ping timeout: 180 seconds)
21:42:43 Quit webguest93 (Client Quit)
21:42:53 Join stephen_ [0] (
21:45:14BdN3504enable rtc in the config-gigabeat.h, will the plugin not work properly now?
21:49:41 Join freetripleg [0] (n=44122fbd@gateway/web/cgi-irc/
21:50:08freetripleghey anyone know how to tell if a fuze is v1 or v2? forum search function is fail.
21:50:43BdN3504afaik, there are fuze v1 and v2s
21:50:58advcomp2019freetripleg, only with the firmware version
21:52:33 Quit QncfOpO (Remote closed the connection)
21:52:46freetriplegahhh, updater says v2 nuts, that means I must wait much longer for rockbox
21:53:43freetriplegthanks for the quick answer
21:53:46 Quit freetripleg (Client Quit)
21:53:49BdN3504erm, how do i use this plugin again? i try install/update bl but i get ABORT: this is an untested etc. but people have reported they flashed anyway, how did they do that?
21:56:24 Join B4gder [241] (n=daniel@rockbox/developer/bagder)
21:59:31 Quit LambdaCalculus37 ()
22:04:53BdN3504ok, i have to copy the bootloader from the patchtracker to the root directory and then try again right? that's why i asked...
22:05:49 Join kugel [0] (n=kugel@rockbox/developer/kugel)
22:05:53bertrikhas anyone ever seen a reference to (or a datasheet of) a WM1800 codec?
22:06:05bertrikI assume it's a wolfson
22:06:41bertrikOr a touch key controller from a manufacturer called Melfas?
22:10:56 Join Grahack [0] (
22:11:16pixelmahmm... if e.g. the Pegbox menu isn't using custom bitmaps anymore, those could be removed or am I missing something?
22:12:18 Join dash32 [0] (
22:12:27linuxstbpixelma: That would seem reasonable...
22:13:01linuxstbbertrik: Never heard of it... Are you sure that's the right model?
22:13:21*pixelma looks around for Ubuntuxer... in vain once more
22:13:43bertrikthe WM1800 is mentioned in a disassembly video and I also find it as a string in the firmware upgrade image
22:14:23linuxstbWell, there seem to be 100s of wolfson codecs, and only some are mentioned officially...
22:14:55pixelmaalthough... I put a bit of work in the adaptation of these bitmaps to different screens... but I understand that custom bitmap menus can cause trouble
22:15:03 Quit HellDragon (Read error: 104 (Connection reset by peer))
22:15:07 Join HellDragon [0] (
22:15:27linuxstbbertrik: Google tells me the PSP uses it though...
22:16:18 Quit BdN3504 ("CGI:IRC")
22:16:19pixelmaI just thought that maybe the background could be kept or so...
22:17:42JdGordon|pixelma: nag various people to allow viewport background images so they can be kept :)
22:18:05linuxstbWhy would you need a viewport background image?
22:18:13 Quit HellDragon (Read error: 104 (Connection reset by peer))
22:18:25pixelmajust set a backdrop I thought
22:19:47JdGordon|sure if you want to do it the boring way....
22:19:48bertriklinuxstb, thanks, didn't know that yet
22:20:17 Join HellDragon [0] (
22:20:27linuxstbbertrik: Odd that there's no mention at all on wolfson's website though...
22:20:54pixelmawhat does that have to do with boring?
22:25:01 Join flydutch [0] (
22:31:01 Join robin0800 [0] (
22:33:19 Quit stephen_ ("Leaving")
22:36:37CIA-69New commit by amiconn (r21884): Further ARMv6 imdct optimisation, ~5.5% speedup.
22:39:49amiconnThat is overall decoding speedup
22:42:19amiconnHmm, new build table still has no progress indicator...
22:43:01gevaertsamiconn: that's to motivate you to run a build client, so you get the countdown messages
22:44:14 Join stephen_ [0] (
22:45:02amiconnI don't get to see those when running it in teh background
22:46:05gevaertswe clearly need to be able to run a user-contributed script on certain events
22:49:35 Quit wincent (Read error: 110 (Connection timed out))
22:51:28 Join stripwax [0] (
22:51:58 Join saratoga [0] (i=9803c6dd@gateway/web/freenode/x-9b73980ae5d89280)
22:52:09saratogafaster iMdct for mpegplayer ;)
22:53:07 Quit robin0800 ("Don't follow me")
22:54:58amiconnsaratoga: By-product of learning how to use armv6 extensions...
22:58:17***Saving seen data "./dancer.seen"
22:59:13linuxstbbertrik: Are interrupts enabled in the bootloader on the Meizu/
23:01:29 Quit _zic (Remote closed the connection)
23:07:17 Quit Grahack ("Leaving.")
23:09:24 Join Zagor [242] (n=bjst@rockbox/developer/Zagor)
23:14:13 Quit dash32 (Read error: 110 (Connection timed out))
23:18:55stripwaxanyone seeing data aborts on pp (ipod 5g) usb happening when windows mounts the device? seems to have started with a recent (local) build, where it's 100% reproducible and I don't *think* anything usb-related is patched, but i'll test an official build too
23:19:09 Quit jfc (Read error: 104 (Connection reset by peer))
23:19:09 Join pingosimon [0] (
23:19:31 Join jfc [0] (
23:19:50 Join robin0800 [0] (
23:20:07pingosimonHey guys, I'm on Mac OS X and my newly Rockbox'd ipod won't show up on the desktop or disk utility when I plug it in
23:20:34pingosimonthe iPod screen shows the USB icon
23:21:04gevaertsWhich version are you using? Current build or release?
23:21:09 Quit jfc (Read error: 104 (Connection reset by peer))
23:21:30 Join jfc [0] (
23:21:30pingosimonthe current build, I did a manual install
23:22:00gevaertsok. Which version of OS X is this?
23:22:01 Quit jfc (Read error: 104 (Connection reset by peer))
23:22:22 Join jfc [0] (
23:22:51 Quit jfc (Read error: 104 (Connection reset by peer))
23:23:13 Join jfc [0] (
23:23:31CIA-69New commit by zagor (r21885): Removed use of bogomips. Changed uname -o to uname -s.
23:23:50gevaertshm, it looks like os x 10.4 has a bug with USB composite devices...
23:24:17 Quit jfc (Read error: 104 (Connection reset by peer))
23:24:38 Join jfc [0] (
23:25:09 Quit jfc (Read error: 104 (Connection reset by peer))
23:25:30 Join jfc [0] (
23:25:34 Quit J-23 (Read error: 110 (Connection timed out))
23:25:37pingosimondo you know of a way to fix that?
23:25:41gevaertsThis is the second report I've seen about this sort of thing. You basically have three options : boot to the original firmware manually when you need USB, install the latest release (3.3) (that one boots to the emergency disk mode on connect), or compile your own build
23:25:59 Quit jfc (Read error: 104 (Connection reset by peer))
23:26:20 Join jfc [0] (
23:26:45 Quit jfc (Read error: 104 (Connection reset by peer))
23:26:59pingosimonyeah I tried to install the latest release using Rockbox Utility, but I got the "could not open iPod" error that nobody has seemed to be able to fix
23:27:06 Join jfc [0] (
23:27:14gevaertsyou can install that one manually as well
23:30:41pingosimonwhere can I find the version number within rockbox?
23:30:53gevaertsJdGordon|: I think we need something like FS #10198 to handle this case as well (i.e. disable HID while using a storage connection)
23:31:00gevaertspingosimon: System->rockbox info
23:31:06ZagorJdGordon: can you try the latest rbclient on mac osx?
23:31:10 Quit DarkDefender ("Leaving")
23:31:14JdGordon|gevaerts: ?
23:31:33 Quit stephen_ ("Leaving")
23:31:40JdGordon|Zagor: yeah sure, but unfortunalty that configure issue means i cant participate uin the builds for a while
23:31:40pingosimonhmm I think this is the latest, r21869-090714
23:32:26ZagorJdGordon|: oh right, the path thing
23:32:51gevaertsJdGordon|: apparently older versions of OS X can't handle composite devices with both usb storage and HID. Either we disable HID totally, or we make it optional. I don't think os x 10.4 users will like us if we don't make things work for them
23:33:16*JdGordon| has no idea what you're talking about... but ok
23:33:27gevaertspingosimon: that's a "current build" from yesterday. The release is something different (currently 3.3)
23:33:44pingosimonah ok, thanks a lot! I'm gonna try 3.3 right now
23:33:52JdGordon|Zagor: should be version 27?
23:33:54gevaertsJdGordon|: FS #10198 is the "USB charge only setting" patch
23:34:04JdGordon|OH... ok
23:34:33*gevaerts thought that JdGordon| knew all FS numbers by heart!
23:34:34CIA-69New commit by dave (r21886): Introduce S5L8701 CONFIG_CPU definition for Nano2G and a new CPU_S5L870X "family" define - the 8700 and 8701 are proving to be different. Also move ...
23:34:52JdGordon|gevaerts: it did sound familiar :)
23:35:07ZagorJdGordon|: oh, I didn't bump the client version yet. I'll do that after this build round.
23:35:44bertriklinuxstb, yes, interrupts are enabled in the meizu m3 bootloader
23:36:01bertrikI think sleep even works
23:36:50B4gderold build system: disabled
23:37:24JdGordon|RIP :p
23:37:49 Quit bmbl ("Bye!")
23:39:59 Quit jfc (Read error: 104 (Connection reset by peer))
23:40:04 Join safetydan [0] (n=deverton@rockbox/developer/safetydan)
23:40:20 Join jfc [0] (
23:40:28pingosimonRegarding installing 3.3 on an ipod that already had a daily build: can I just replace the old .rockbox folder with the new one?
23:40:35pingosimonor do I need to run ipodpatcher again?
23:40:54gevaertsjust replace .rockbox
23:41:12pingosimonexcept now even witht he ipod firmware loaded,, it won't show up on the computer
23:41:25linuxstbnot cool
23:41:35gevaertsdo you have itunes running? Maybe that hides it
23:41:41pingosimonnope, itunes is off
23:41:47linuxstbBut installed?
23:41:59pingosimonoh wait there it is
23:42:13linuxstbYou need to enable the "use ipod as a disk" option in itunes.
23:42:16pingosimonyes I do have itunes installed, but the auto sync feature is disabled
23:42:20pingosimonyeah that's enabled too
23:42:28gevaertsso it was just a bit slow?
23:42:33pingosimonanyway, it's here on the desktop now!
23:42:42pingosimonyeah it took about 2 minutes
23:42:47CIA-69New commit by bagder (r21887): Added Philips SA9200 sim and bootloader builds
23:43:24JdGordon|stop adding builds! arnt we trying to get under 5 min :p
23:43:37pingosimonDo any of you guys use rockbox to play NSF SPC or SID files?
23:44:02linuxstbB4gder: Did you disable the old build system half-way through it building my commit? ;)
23:44:52B4gdernot exactly intended, but hey
23:44:54*linuxstb tries not to be offended ;)
23:45:28Zagorlinuxstb: you got it built faster this way! :)
23:45:41linuxstbYes, if I wasn't staring at the old page... ;)
23:46:12*Zagor renames new.cgi to dev.cgi
23:48:43gevaertsso no chance of having to go back? We can clear the old setups?
23:50:33B4gderwhat setups?
23:51:13gevaertsthe rbclient login and corresponding checkout
23:51:24 Join DerPapst [0] (
23:51:53B4gderah yes those can be wiped, but you could possibly wait a day or two before you proceed with that
23:52:04pingosimonOk with Rockbox 3.3, the ipod does show up on the desktop, after leaving it plugged in for 2-3 minutes. Maybe I just wasn't patient enough when trying with the daily build
23:52:10gevaertsok. No problem with that
23:52:42gevaertspingosimon: you could try. If you do, please report back
23:53:02pingosimonwill do
23:53:29pingosimonthis seems to have the same USB speed problem as the nano did
23:53:34DerPapstpixelma: about the pegbox background, you could viewportify the menu and shift it down so that it is not drawn on top of the pegbox "logo"
23:53:38 Quit robin0800 ("Don't follow me")
23:54:14DerPapsti've done something similar for battleships (which is still not tracker ready. darn real life :P)
23:55:59gevaertspingosimon: yes. You're now in the apple emergency disk mode, which is slower than the normal USB code. The native rockbox code (which you get with a current build) is a bit faster, but not as fast as the original firmware (we know how to make it faster, but that's still not in shape for the official builds)
23:56:04Zagorwhich is the slowest build client yet? did someone run an arm machine?
23:56:26gevaertsZagor: monster is a 1.2GHz armv5 (sheevaplug)
23:56:42gevaertsIt seems to be faster than rasher's cygwin client though
23:56:57 Quit flydutch ("/* empty */")
23:57:41ZagorI'm about to add a 800 MHz Via C3
23:58:06gevaertswe'll see. hal is a 900 MHz Celeron M
23:58:17Zagorceleron is a monster compared to c3
23:58:36gevaertsceleron M is really not very impressive :)
23:58:51*Zagor dramatically throws a glove on the floor

Previous day | Next day