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

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

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

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

#rockbox log for 2005-02-22

00:00:18amiconn_Ah. With calendar.c modified to use fdprintf(), it now compiles to the end...
00:00:41 Quit amiconn (Nick collision from services.)
00:00:42 Nick amiconn_ is now known as amiconn (
00:01:26elinenbeforget the calendar problem... there is a problem with recursive insert!
00:02:35Bagderthen fix it! ;-)
00:04:52 Part hubble
00:07:22Bagderamiconn: I'll make a little fix I think may fix the icon/image too
00:07:45amiconnShould I wait? I was about to apply simfix-10
00:07:56Bagdertry the -11 instead
00:10:30BagderLinux host detected
00:10:33BagderEnabling win32 crosscompiling
00:10:37BagderUsing i386-mingw32msvc-gcc 2.95.3-6 (295)
00:10:42LinusNelinenbe: can you explain what is wrong with the insert?
00:11:19Bagderalmost worked!
00:11:43LinusNelinenbe: ok, i see it
00:15:14amiconnBagder: Still those 'conflicting types for built-in function' warnings, and a couple other. I'll try again with reconfigure...
00:15:56 Quit Chamois ("CGI:IRC")
00:16:42Bagderhehe, ftruncate() is defined to NULL in the old sim
00:17:02Bagderbut no plugin uses it
00:18:50HClwow, it actually compiled
00:18:55HClwithout any compiler errors.
00:19:04HClno internal compiler errors
00:19:05HCli mean
00:19:42amiconnBagder: Some warnings remain (in plugin.c and uisimulator/win32/io.c), and the image & icon are still missing
00:20:10BagderI ignore the warnings for now, focusing on getting the build on all combos
00:21:21HClwhat does this mean xD
00:21:35amiconnYou can leave the Win32 specific warnings for me to fix. However, I don't understand one of the warnings in abovementioned io.c
00:21:36HClat 30066BB8
00:21:53LinusNHCl: illegal opcode
00:21:54amiconnHCl: That means an illegal 68k instruction
00:21:59HClokay XD
00:24:12 Join webguest83 [0] (
00:24:38amiconnBagder: Rather, I understand the warning, but don't know the best way to fix it. Somehow this file needs the struct plugin_api declaration...
00:26:53BagderTremor doesn't build cross-compiled for win32 ;-)
00:26:55 Quit Sucka`away ("( :: NoNameScript 3.81 :: )")
00:27:13Bagderlooks like a C99 requirement
00:27:15amiconnBagder: Similar warnings when building the X11 sim with cygwin
00:27:32amiconnBagder: Did you try the iriver sim?
00:27:57Bagderworks fine, except the win32 crosscompile
00:28:01amiconnI could try the same. Do you have an idea what causes the still missing bg image + icon?
00:29:02Bagdernot really, my win32 build at least builds a uisw32-res file...
00:29:12Bagderdoes it need to be name like the output perhaps?
00:29:29BagderI don't know much about resource files
00:29:47amiconnYou renamed the .exe a while ago, it still had image + icon
00:29:59Bagderah, I fixed a little thing
00:30:05Bagder$(WINDRES) −−include-dir $(OBJDIR) -i $< -o $@
00:30:14Bagderis the new line, in win32/Makefile
00:30:36amiconnCurrently building iriver-win32 sim, to check whether tremor builds...
00:30:54Bagderyou have a newer gcc than my win32
00:31:07Bagderso it might just work fine
00:31:43*HCl goes to figure out whats wrong with his opcode
00:34:11amiconnBagder: Building tremor only shows a line 'AR /home/Administrator/rb-patched/simulator-build/iriver/libTremor.a' !??
00:34:57amiconnmpa2wav.c errors out
00:35:11amiconnmpa2wav.c:94: undefined reference to `_mad_stream_init'
00:35:16amiconnmpa2wav.c:63: undefined reference to `_mad_stream_buffer'
00:35:36Bagderwell, those are issues beyond this current work
00:35:50amiconnYes of course. Only wanted to mention that
00:36:13amiconnThe important thing is that the simulator itself and most plugins build, and work
00:36:31amiconnThe rest is then solved case by case
00:37:53amiconnBagder: Haha, funny: 'make install' now copies the executable into archos/ on the sims...
00:38:16Bagdervery useful!
00:38:53amiconnHmm, plugins don't work on both sims... :(
00:39:05LinusNelinenbe: i have fixed the recursive insert, please try the next bleeding edge, or tomorrow's daily build
00:39:07amiconnThe X11 sim says 'Win32 error 5'
00:39:46 Join AC [0] (
00:40:42Bagderso, I'll hold off the commit
00:40:47linuxstbAC: Hi. How are things with wv?
00:41:55AClinuxstb: i should have it in about 30 mins
00:42:04HCli don't get this...
00:42:12HClnevermind >.>
00:42:15linuxstbAC: Any indication of decoding speed yet?
00:42:27*HCl kind... of... forgot the return instruction at the end of a dynarec block >.
00:42:43amiconnBagder: The plugins miss the "execute" permission. I had that before with the Win32 sim, but not the X11. When I manually change that, they work...
00:43:16AClinuxstb: not yet... in a few minutes
00:44:12Bagderamiconn: ok, I duplicate the logic for it
00:45:58amiconnAnd I found the problem with the missing image / icon: You do not link with uisw32-res.o ...
00:46:03rashurlinuxstb: the line that says "Speed" in the *2wav viewers, is that % of realtime?
00:46:33 Join ashridah [0] (
00:46:43Bagderamiconn: I need to add that special somewhere?
00:47:19amiconnThis needs to be linked to the sim executable.
00:49:34Bagderbut I don't see why it isn't
00:49:58amiconnIt does not appear on the linker command line...
00:50:23Bagderno, since it is in the libsim.a
00:50:45Bagdertry 'ar tv libsim.a' to view the contents of it
00:51:24amiconnYes, it's there. However, this may not be enough
00:51:58amiconnThe old build included it directly in the linker command. Maybe there is a special way to tell the linker to use this version from libsim.a
00:52:36amiconnBtw, I can't find the place where the executable is linked. Is that now part of apps/Makefile?
00:53:15ACis there a way to debug a plugon on the iriver?
00:53:20Bagderamiconn: yes it is
00:53:38BagderI like the new look on the cvs table while a build is in progress!
00:54:48Bagdercrappy estimate though
00:54:59Patr3ckif I want to add a few functions that should be used by all codecs libraries, do I need to create a separate library?
00:55:11linuxstbAC: Not really. You could open a file and write debugging information to it - or display information on the LCD.
00:55:43AClinuxstb: ok
00:55:46BagderAC: simple plugins can be run in the simulator, but I guess yours isn't ;-)
00:56:05amiconnBagder: That's it. I added the .o directly on the linker command line as a test, this works
00:56:24*rashur wonders how much it'd cost to transfer money directly to a Swedish bank account
00:56:25Bagderok, then we just need to make that happen in a nice way
00:56:46amiconnOr find out how to link kin a resource from a library
00:56:55amiconn*link in
00:57:05ACsemms that there is an error in WavpackUnpackSamples function... will look closer at it
00:57:27Bagderrashur: probably not that much from .dk
00:57:44BagderI need to sleep now
00:57:49rashurThat's what I figured. And I'm not too keen on paypal really.
00:57:58amiconnBagder: No commmit?
00:58:10linuxstbAC: Does your plugin crash at the moment?
00:58:14BagderI don't feel like sitting up fixing the quirks now
00:58:19BagderI'll try to do it tomorrow
00:58:32AClinuxstb: yep... i am looking at the moment why
00:58:48amiconnBagder: Hmm, okay :-/ (no offense)
00:58:48AClinuxstb: i know where it crashes
00:58:57***Saving seen data "./dancer.seen"
00:59:05Bagderamiconn: worst case, you can do the commit, there's a fresh patch uploaded
00:59:19linuxstbAC: Does it make use of the stack? If I remember correctly, the stack is currently about 32K.
00:59:33Bagdergood night
00:59:35amiconnBagder: I think it's easier to fix the quirks within cvs
00:59:48Bagderamiconn: it is
00:59:55Bagderfeel free to commit if you want to
00:59:57amiconnDoes it at least build on Linux
01:00:04Bagderyes it does
01:00:10AClinuxstb: it looks now much better...
01:00:10Bagderboth plain and sim
01:00:53AClinuxstb: it is not working atm, but i got a step into the correct dir
01:01:06amiconnHmm, maybe I should stop here too for today. I'd say commit tomorrow, and then we fix the quirks
01:01:07 Quit webguest83 ("CGI:IRC (EOF)")
01:01:15*HCl sighs.
01:01:19HCli'm at the point
01:01:25HClwhere i just need an assembly-level debugger
01:01:25amiconnBagder: Good night.
01:01:26HCl :(
01:02:01Patr3ckI have created a few simple functions for profiling that use e.g. file.h, sprintf.h from rockbox
01:02:14Patr3ckthese should be used by the codec libraries
01:02:15amiconnBagder: I'll check for that odd windows linker problem meanwhile
01:02:50Patr3ckcan someone point me into a general direction how to get access to these from the codec libraries?
01:03:40Patr3ckat the moment I get undefined references
01:03:51Patr3ckwhen linking
01:04:07amiconnYou cannot use these rockbox includes directly from within plugins
01:04:20amiconnInstead, you have to use the plugin api
01:04:32 Quit pillo (Read error: 110 (Connection timed out))
01:04:40Patr3ckThese are used in the codec libraries...
01:04:49Patr3cknot in the plugin itself
01:05:03amiconnYes, and the codec libraries are use from within the plugins...
01:05:15amiconnThe codec libs are never linked against core rockbox
01:06:08ACsometimes it is nedded
01:06:17Patr3ckcan I access the plugin api from within e.g. libmad?
01:06:21ACe.g. strcpy
01:06:34ACit would be nice, if i could use it in the codec
01:06:48preglowwell you'll pretty much have to, heh
01:06:54amiconnHmm. I don't know how it is adapted currently. You would need a copy of the plugin api pointer
01:07:03preglowi don't think the ordinary functions are even available
01:07:16linuxstbThe current system is just temporary - there will be a "codec api" which will give codecs access to the standard functions like "strcpy" etc.
01:07:40Patr3ckso at the moment it is not possible?
01:07:42amiconnThe only other option is to compile and link a private version, but that's ugly imho
01:07:54preglowyes, it is
01:08:04linuxstbFor the moment, I have just implemented wrappers in plugins/lib/xxx2wav.c for functions like memcpy, and also added a simple malloc implementation (without free).
01:08:46amiconnlinuxstb: How do you wrap?
01:09:26Patr3cklinuxstb: i could add the profiling functions there too and access them from within the codec library?
01:09:48preglowperhaps i should try coding my profiling idea
01:10:08linuxstbamiconn: the codecs libraries themselves include firmware/include/*.h, and then I implement those functions in xxx2wav.c by calling rb->???
01:11:43amiconnIn the plugins, you don't need to include these. Plugins are allowed to only ever include plugin.h, or priovate include files. They must not include other rockbox includes
01:12:21amiconnThe question is, how do you wrap these function in the codecs?
01:14:47amiconnYou can use rb->* in the codecs too, but that means your test plugins always have to have a global plugin api struct pointer, and it always needs to be called rb
01:15:06amiconnI think this is okay for the test plugins
01:15:36ACso i can use strcpy via the plugin api?
01:16:06amiconnIf you want a somewhat cleaner implementation, look at the grayscale lib. This lib has its own private plugin api pointer, and an init function to provide its value
01:16:45linuxstbAC: Yes, your "wv2wav" program can use the plugin API, but the code in codecs/libwavpack/ can not. Have a look at the code in apps/plugins/lib/xxx2wav.[ch] to see what I did for similar functions.
01:17:23linuxstbI still think of this as a temporary solution until the codec API is written, which will give the codecs access to libc type functions.
01:18:12Patr3cklinuxstb: I am trying to add them to xxx2wav.h and use the plugin api
01:18:51Patr3cklinuxstb: they are only temporary anyway I think
01:19:03linuxstbPatr3ck: the problem with that is that the codecs shouldn't really be including xxx2wav.h
01:19:42Patr3cklinuxstb: ah, too bad, only the plugin itself should do that?
01:20:14linuxstbIt's a bit like a system library including part of your application.
01:21:08Patr3ckI could wait, but performance measurement would be helpful now
01:21:36amiconnlinuxstb: I just found that you define dprintf(). Imho this shouldn't be necessary, you can use debugf()
01:22:35linuxstbamiconn: Does debugf do the same as my dprintf?
01:23:12amiconnAlmost. It does printf() in the simulators (unless you hook a win32 sim into a windows debugger)
01:23:31linuxstbOK, I'll change to use that then. Thanks.
01:23:35amiconnOn the target, it does the same (i.e. nothing) unless you create a debug build
01:24:19amiconnA debug build can actually output this using a gdb stub. I know, no such stub does exist yet for iriver.
01:25:31HCltell me about it :/
01:25:46amiconnHCl: There is a debug stub for archos...
01:25:48linuxstbPatr3ck: I agree it would be very helpful to have your code and start profiling, I'm just not sure of the best way to incorporate it into Rockbox. Maybe put "profile.h" and "profile.c" in apps/codecs, and then apps/codecs/Makefile would compile it, and make it available to be linked against the ???2wav plugins.
01:26:02HClamiconn: tell me when you got some free time to help me with rockboy a bit...
01:26:11preglowis it hard to generate an address map of all rockbox functions?
01:26:17HCli think i need a bit of help with debugging it
01:26:22preglowi'd pretty much need that for the profiler
01:26:48amiconnpreglow: When you compile, a .map file should be created too...
01:27:22amiconnHowever, it does only list global functions. Statics are left out
01:28:05preglowamiconn: yes, so it seems...
01:28:16preglowthat might be a show stopper
01:28:47Patr3cklinuxstb: I will try that
01:29:48 Nick rashur is now known as rasher (
01:29:58preglowamiconn: no codec symbols are included, and that's pretty much the reason i want to make a profiler
01:30:59Patr3cklinuxstb: but I am wondering how this would give me access to the rockbox functions
01:31:21 Quit Renko ("Leaving")
01:35:24XShocKLinus: Maybe it is better to set ICR register each byte long, and have all 12 different ICR register than 4 compound ones as now?
01:35:39XShocK#define ICR0 (*(volatile unsigned long *)(MBAR + 0x04c))
01:35:40XShocK#define ICR4 (*(volatile unsigned long *)(MBAR + 0x050))
01:35:40XShocK#define ICR8 (*(volatile unsigned long *)(MBAR + 0x054))
01:35:40DBUGEnqueued KICK XShocK
01:35:40XShocK#define ICR6 (*(volatile unsigned long *)(MBAR + 0x052))
01:36:59linuxstbPatr3ck: You're right, it wouldn't.
01:37:00XShocKICR6 was added by me, so it is not in rockbox in cvs
01:37:16LinusNthe internal register map is not byte addressable
01:37:36XShocKaa ok. didn't think about that.
01:40:17Patr3cklinuxstb: seems there is no clean solution at the moment
01:40:54linuxstbPatr3ck: No. If I was you, I would just do an "unclean" solution for now, and we can try and think of something clean later.
01:42:18Patr3cklinuxstb: for now it seems wrapping the needed functions in xxx2wav.h and including this in the codec library is the only working way
01:46:40preglowwoot, iram for codecs
01:46:42 Quit AC ("CGI:IRC (EOF)")
01:47:21preglowspeed test it, quick!
01:48:19 Quit Aison ("( :: NoNameScript 3.72 :: )")
01:48:34 Join AC [0] (
01:50:53amiconnDoes the coldfire have a cache? I guess so...
01:51:01preglowamiconn: instruction cache
01:51:14amiconnAh, so this may be HCl's problem...
01:51:19elinenbeLinusN: thanks so much... works great for me!
01:51:25preglowamiconn: indeed
01:51:41preglowamiconn: plus there's 96kbyte of iram that's single cycle access
01:51:51amiconnHCl: The coldfire has an instruction cache. Doing dynarec is the same as having self-modifying code...
01:52:06HClwell, yes. but. um. huh?
01:52:19HClhow does that relate to what i'm doing?
01:52:26amiconnYou need to invalidate the cache in order to tell the cpu to re-read from memory
01:52:41preglowno, it's not really the same at all, i think
01:53:07HCli'm sort of doing the same thing when loading a module
01:53:15HClexcept i'm not loading code from disk but writing it on the fly
01:54:29amiconnYes, that's correct. However, if you load a new module, the area was not previously cached
01:55:08amiconnAnd if you load a new module in place of an old one, this may indeed cause the same problem
01:56:07amiconnThat just came to mind as I read a bit about a ZX Spectrum emulator for the Amiga.
01:57:10amiconnLinusN: Does rockbox already handle these caching issues?
01:57:36HClwhat do you mean the area was not previously cached...
01:57:46LinusNamiconn: the cache is invalidated when loading plugins
01:57:59HClLinusN: where's the code for invalidating the cache..?
01:58:22HClbut i don't think it has to do with it though
01:58:28HClcause earlier i got an invalid opcode error
01:58:31HCland now it just freezes
01:58:42HCli added a return from subroutine opcode to it..
01:58:49amiconnHCl: You may need to call that every time you compile a new sequence
01:58:55HClamiconn: yea, i get the idea
02:01:21HClkay, i added that.
02:01:38 Quit Patr3ck ()
02:03:10amiconnLinusN: I guess using invalidate_icache() is okay within plugins, as it is an inline function anyway?
02:03:27LinusNgotta go to sleep
02:03:32LinusNcu tomorrow
02:03:36ACgood night
02:03:39 Part LinusN
02:03:58HClnop, freezes
02:06:11HClamiconn: what're your experiences with casting an int to a function and calling it.. could you just return from it normally..?
02:06:26MO-Pantsuhow dare he sleep. he must work 24/7 on iRiver! :D
02:06:56amiconnHCl: Yes it worked. However, I changed that meanwhile if you have a look at apps/plugins/rockboy.c
02:07:13amiconnNow I use a struct instead of an array, within the correct pointer type
02:09:35preglowlinuxstb: what efficiency did flac2wav have again?
02:10:27CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
02:10:27*rasher feels sortof silly spending a few hours decoding an mp3
02:10:33 Join mecraw [0] (
02:10:58rasherMostly because I have no idea why I started it.
02:12:31HCloh.. well.. yea.
02:12:32HCli'm dumb.
02:12:41*HCl thinks he found the bug ^^
02:12:56HClhm, wait, nm :/
02:13:56*AC needs help
02:14:09ACwarning: this decimal constant is unsigned only in ISO C90
02:14:21ACwarning: integer overflow in expression
02:14:30ACmin_shifted = (min_value = -(long)2147483648 >> shift) << shift;
02:16:19HClugly code..
02:16:40HClfirst of all
02:16:44ACnot mine... libwavpack
02:16:45HClget the min_value bit out
02:16:53HCland make min_shifted = min_value << shift;
02:17:25linuxstbpreglow: It's somewhere between 7% and 8% of realtime for my tests. I think Linus said it was about 90% at 140MHz.
02:17:43preglowjust tested it at 6.50 here :/
02:17:57preglowdoes −−best encoding give more lpc frames?
02:18:01linuxstbI must have a faster iRiver then :-)
02:18:05XShocKand isn't it better to make that 2147483648 be 0x80000000
02:18:31preglowXShocK: LONG_MIN even better
02:18:41ACXShocK: i dont know...
02:18:52XShocKpreglow: yes.
02:19:01linuxstbpreglow: I forget the details, but has excellent documentation. "man flac" is also good.
02:20:10ACwhere is LONG_MIN declared?
02:20:22 Quit mecraw ()
02:20:38preglow−−best sets max order lpc = 12, hyes
02:20:47preglowAC: limits.h, might not be in rockbox for all i know
02:21:21linuxstbpreglow: According to the man page, "−−best" is equivalent to "-8", which means it searches for order-12 lpc predictors.
02:21:30ACpreglow: so we need now a limits.h
02:22:18preglowAC: just use 0x80000000
02:22:32ACpreglow: ok
02:28:33 Quit edx (Read error: 110 (Connection timed out))
02:35:07MO-PantsuMP3 b4 everything! ;)
02:41:19 Nick MO-Pantsu is now known as DeadMan (
02:46:40preglowbut bed
02:46:53 Quit preglow ("omglolmao")
02:48:36XShocKLinus: I think it does support byte addressing, and each different ICR can be done as in move.b d0,($40000053).l in original firmware
02:49:24XShocKoops, didn't realize that he is gone
02:58:26 Quit AC ("CGI:IRC (Ping timeout)")
02:59:00***Saving seen data "./dancer.seen"
03:03:32 Quit amiconn (" HydraIRC -> <- Leading Edge IRC")
03:05:46DeadMananyone fancy a giggle? beware contains adult content
03:26:59 Part Bluechip
03:34:46 Join StrathAFK [0] (
03:34:56 Quit Strath (Nick collision from services.)
03:34:58 Nick StrathAFK is now known as Strath (
03:47:15 Join bagawk [0] (Lee@bagawk.user)
03:48:34bagawk[IDC]Dragon, hey
03:56:55bagawkMe have forums :)
04:03:38 Quit bagawk ("Leaving")
04:09:57 Join edx [0] (
04:21:18 Quit skav ()
04:29:09 Quit YouCeyE ("Leaving")
04:34:18 Quit XShocK (" HydraIRC -> <- IRC with a difference")
04:59:01***Saving seen data "./dancer.seen"
05:00:05 Join midk [0] (
05:44:00 Quit cYmen ("leaving")
06:01:29 Join YouCeyE [0] (
06:35:02rasherSortof off-topic, but what could be wrong when the iRiver firmware tells me "playlist error" when trying to load a playlist?
06:36:35rasherokay, nevermind me
06:59:03***Saving seen data "./dancer.seen"
08:00:29rasheror not.. is it just that iRiver requires CR/LF instead of just LF?
08:04:00[IDC]Dragonoops, I had the IRC window on all night
08:12:15 Join LinusN [0] (
08:17:42dwihnohowdy ho
08:19:28 Quit pill (Read error: 110 (Connection timed out))
08:28:56 Quit midk ("Leaving")
08:30:28 Quit ashridah (Read error: 110 (Connection timed out))
08:36:43 Join pill [0] (
08:59:04***Saving seen data "./dancer.seen"
09:00:21 Join jyp [0] (
09:11:53 Join ashridah [0] (
09:20:40 Join bobTHC [0] (
09:20:59bobTHCgood morning !
09:33:01 Join midk [0] (
09:40:17*HCl falls asleep again
09:47:06 Join amiconn [0] (
09:50:51rasherIt sure is.
09:58:55amiconnBagder: R u there?
10:28:42 Quit Ka (Read error: 110 (Connection timed out))
10:34:30 Join Aison [0] (
10:59:08***Saving seen data "./dancer.seen"
11:00:22LinusNamiconn: how are the sim builds working for you?
11:03:12amiconnCurrently not at all (with Bagder's latest patch as well as without) - that's why I want to talk to Bagder...
11:03:38amiconnHis -11 fix works, sort of, while his -13 fix breaks
11:07:44DeadManlol can't be for real surely
11:08:54 Quit midk (Read error: 60 (Operation timed out))
11:10:34ashridahDeadMan: they've been trying to push the subscription services angle for ages. this is just another prong, and one that's going to be justifiable in the name of 'security' if someone bitches.
11:12:22DeadManwill be interesting to see what happens
11:14:44ashridahi imagine it'll create as many problems as it solves for everyday users, but any business with a reasonable amount of pcs that isn't using SUS or some other distribution system and just blocking outbound access to ms's site completely will do things the way they always have
11:15:33ashridahwhich, will typically just mean more money for smalltime computer shop techs, and slightly more headaches for large site sysadmins.
11:27:13 Part LinusN
11:28:10 Join Renko [0] (
11:33:56 Join Patr3ck [0] (
11:44:59 Join Lurkski [0] (
11:48:02 Join midk [0] (
11:48:14HClmorning thing.
11:53:32Bagderamiconn: so what's the prob with -13?
11:55:22amiconnFor win32, it produces a linker error
11:55:36amiconnThe x11 build doesn't even compile to the end
11:57:16amiconnAnother hint: You don't need the fancy workaround for ftruncate() on Windows. There is _chsize() the same way as _commit corresponds to fsync()
11:59:20*HCl wonders whether amiconn has auto accept..
11:59:30amiconnI have...
11:59:30HClamiconn: here's my source for rockboy at the moment
11:59:53Bagderamiconn: but the win32 build failed to link with the sim_ftruncate for some reason
12:00:14amiconnYeah... it doesn't find the symbol when linking...
12:00:33amiconnI did not yet find the reason why
12:03:38amiconnI'll retry; perhaps it was my fault...
12:04:21 Join Ka_ [0] (
12:08:45amiconnBagder: Hmm, strange. Now it worked... (win32).
12:09:04Bagderprobably just the bad dependency stuff
12:09:07Bagderthat hit you
12:09:29amiconnI thought I did 'make clean', but now I'm not so sure...
12:09:30BagderI found the x11 problem too
12:09:52Bagderthere's a patch -14 now
12:12:24amiconnDid you add the direct resource linking?
12:17:18 Quit linuxstb (Read error: 60 (Operation timed out))
12:18:45Bagderah, no...
12:25:14 Join linuxstb [0] (
12:26:44 Quit linuxstb ("Leaving")
12:27:51 Join linuxstb [0] (
12:32:16 Join zezayer [0] (
12:33:34 Quit zezayer (Client Quit)
12:36:58Bagderthere's a -15 patch now
12:39:14 Quit midk ("Leaving")
12:39:26 Quit linuxstb ("Leaving")
12:40:01 Join linuxstb [0] (
12:42:20 Quit jyp (Read error: 110 (Connection timed out))
12:48:06HCldo instructions have to be memory aligned? and if so, how do i make sure they are?
12:59:11***Saving seen data "./dancer.seen"
13:00:38ashridahHCl: from my basic 68k knowledge, no. only the stack has to be two-byte aligned.
13:01:11amiconnHCl: I think 68k can work without, but proper alignment wil speed things up
13:04:10amiconnWhat's the problem with aligning them? If you start the block aligned, further instructions should be aligned as well
13:05:33HClyea, i know
13:05:42HCli'm just not sure on the definition of aligned
13:05:42HCl :p
13:06:13HCli'm still a bit annoyed at not being able to call a function that does a simple return from subroutine..
13:06:58ashridahanyone happen to have a link to a reference for gnu assembler's syntax
13:07:09HClamiconn has one..
13:07:49ashridahhmm. guess i can use the info page in a pinch
13:11:18ashridahamiconn: nice. that'll help too. definently unfamiliar with c/asm mixing.
13:16:53amiconnBagder: Both win32 and x11 build now. Image & icon are there for win32
13:17:21amiconnX11 is working nicely, Win32 still misses the execute permission on plugins
13:17:21Bagdershould I commit?
13:17:40BagderI'm gonna leave so I won't be around to fix any remaining nits
13:17:57 Join sox [0] (
13:17:58amiconnThere are some warnings left, but I think you could commit
13:18:05Bagderok, buckle up
13:18:09Bagderhuge commit coming
13:18:12 Join Patr3ck_ [0] (
13:18:13amiconnI'll try to fix the remaining warnings in the evening
13:18:46soxhoy all, I asked this before but havent found any way to solve it so I figure I might ask again
13:19:18soxI get this error trying to run make for the iriver build on my mac os x 10.3.8
13:19:29soxUsing m68k-elf-gcc 3.4.3 (304)
13:19:29soxCreated Makefile
13:19:29soxc-503fe255:~/rockbox/rockbox-devel/uibuild svante$ make
13:19:29DBUGEnqueued KICK sox
13:19:29soxmake[1]: *** No rule to make target `#pragma', needed by `/Users/svante/rockbox/rockbox-devel/uibuild/dep-firmware'. Stop.
13:19:29soxmake: *** [all] Error 2
13:19:57BagderI believe your gcc produces a #pragma file in the shell magic in the top of the firmware/Makefile
13:20:00soxthis time i tried building the uisim, but it's the same error building the normal version
13:20:07Bagdera #pragma line even
13:20:45soxthis error came from nothing three days ago, some change in the code must have caused it
13:20:55ashridahsox: can you show us a copy of the Makefile that's being generated by tools/configure ?
13:21:26soxthe whole file?
13:21:29amiconnBagder: Different thing: There is still no current daily for Ondio FM, and the daily windows installer builds are also missing...
13:21:44ashridahsox: toss it up on a website or something'd be easiest.
13:22:01ashridahalthough i suspect i'm asking for the wrong makefile.
13:22:11Bagderamiconn: I know, but I have no time to fix that now
13:22:23Bagderzagor or Linus will have to do it, or wait
13:22:50amiconnOkay, just wanted to mention...
13:23:12Bagderit is most likely because of the "flash full" error
13:24:26soxashridah: that's the root Makefile, did you want the firmware/Makefile?
13:24:34Bagdersox: can you show us how your dep-firmware looks like?
13:25:13sox is for firmware
13:25:59ashridahyeah, put up a copy of the dep-firmware file.
13:26:18soxBagder: where is it?
13:26:22ashridahjust run 'make dep' if it doesn't exist. (don't know if it gets destroyed)
13:26:26Bagdersox: in your build dir
13:26:29ashridahshould be in the directory you're compiling in
13:27:13soxsorry, it's not there
13:27:53ashridahwait. this is an osx box isn't it? are you using gnu make or bsd's make?
13:27:59ashridahtry gmake ?
13:28:07soxnot found
13:28:15Bagderyou must use GNU make
13:28:28ashridahokay. i'm not that familiar with osx, so i don't know which it includes.
13:28:56soxcan i run both simultaneously?
13:29:01 Join ripnetuk [0] (
13:29:25ashridahyeah, usually environments that add it give it a different name, hence 'gmake' for gnu make on these systems.
13:29:44ashridahi'm not sure why that wouldn't be there, unless you didn't install the development resources for osX
13:29:52soxi sure did
13:30:15ashridahokay, what's the first line of output from running 'make −−version'
13:30:48soxGNU Make version 3.79,...
13:30:56ashridahhm. that should be okay then
13:31:01ashridahrun 'make dep'
13:31:20 Quit elinenbe (" HydraIRC -> <- The dawn of a new IRC era")
13:31:22soxmake: *** No rule to make target `dep'. Stop.
13:31:34Bagderit is meant to handle them automaticly
13:31:39amiconnBagder: I now updated my two working copies. While on one I got the usual message about removed files ('xyz is no longer in the repository'), the other gave me a warning: 'xyz is not (any longer) pertinent'. What does that mean?
13:31:59Bagderamiconn: I don't know
13:32:11ashridahah, hang on. the dep target is only in the firmware makefile, not the simulator one
13:32:18Bagderamiconn: perhaps there was a sys/types.h in the past
13:32:34Bagderashridah: the sim uses the firmware makefile nowadays
13:32:55ashridahBagder: yeah, i'm curious as to why these two are markedly different.
13:32:59Bagderoh right, there's a 'dep' there
13:33:35Bagderbut there's no 'dep' in the root Makefile so it doesn't work
13:33:39amiconnBagder: It gave this warning for uisimulator/common/dir.h, uisimulator/common/file.h and uisimulator/x11/kernel.h
13:34:05soxare you talking about something that affects my error too?
13:34:42Bagderouch, player build broke
13:34:46ashridahsox: okay, run 'make -n'
13:34:59ashridahand take a copy of the entire output, toss it into a file, and show us it
13:34:59soxmake -C /Users/svante/rockbox/rockbox/firmware
13:34:59soxmake[1]: *** No rule to make target `#pragma', needed by `/Users/svante/rockbox/rockbox/build/dep-firmware'. Stop.
13:34:59soxmake: *** [all] Error 2
13:35:06soxthat's it
13:35:28 Quit Patr3ck (Read error: 110 (Connection timed out))
13:35:45Bagdersox: cd to the firmware dir,then run this command:
13:36:33Bagdercat SOURCES | gcc -E -P
13:36:50Bagdereh, no...'ll fail on the include
13:37:41Bagdercat SOURCES | gcc -Iexport -I. -E -P
13:37:53BagderI suspect it adds #pragma lines to the output
13:38:04sox"no input files"
13:38:15Bagdercat SOURCES | gcc -Iexport -I. -E -P -
13:38:20Bagder(added a dash)
13:38:47soxthe first line says: #pragma GCC set_debug_pwd "/Users/svante/rockbox/rockbox/firmware"
13:38:53Bagderthere it is
13:38:58Bagderthe problem is identified
13:39:02 Join jyp [0] (
13:39:10Bagdernow we need a way to switch off that
13:43:14ashridahseems to be an apple-specific feature
13:43:47Bagdertry adding -xassembler-with-cpp to the gcc options
13:43:53 Join R3nTiL [0] (~zorroz@
13:44:01ashridahThe GCC 3.3 preprocessor inserts a new pragma, #pragma GCC set_debug_pwd, as part of the new Distributed Builds feature. (See below.) This may surprise tools and scripts that depended on the exact form of preprocessed output from GCC. These scripts should be rewritten to ignore unrecognized pragmas."
13:44:38Bagderthey should've used -E as a hint
13:44:49Bagdereh, no I mean -P
13:45:08Bagder(which is the disable #line lines option)
13:46:31soxis there any easy way to do this so that I dont have to do it manually every time
13:46:50Bagderwe need to fix the Makefiles to use this option
13:47:05Bagderthe same trick is used in most Rockbox makefiles
13:50:58amiconnBagder: Red builds for the player sims (didn't check these :( ) and an fprintf is left in radio.c
13:51:01soxI tried adding -xassembler-with-cpp to the GCCOPTS, didnt change anything, any other places I have to change
13:51:04ashridahit's sad to see all the stupid hacks people have to go through because apple screwed up the default behavior of the c preprocessor
13:51:16Bagder...I noticed, working on some of them
13:51:35ashridahsox: the preprocessor probably doesn't get passed GCCOPTS
13:51:42Bagdersox: the GCCOPTS is not used for that invoke
13:51:47ashridahsince it's not, strictly speaking, compiling anything.
13:51:47soxah ok
13:51:54ashridahit's just the preprocessor
13:52:08Bagdersox: try the extra defines
13:52:24Bagderbut it's a hack
13:52:41Bagdereven if it might work as a quick work-around
13:55:05soxBagder: you mean like this: "export EXTRA_DEFINES=-xassembler-with-cpp"
13:55:41soxit compiles, but with a shitload of errors
13:57:31ashridahBagder: for a stopgap, he could try using the preprocessor that he would have built targetting the m68k?
13:57:43Bagderah, yes
13:57:53soxstill, I wonder what made this problem come up, because three days ago I was compiling without problems, and the pragma line was added then too?
13:57:53ashridahit probably won't do the trick for the simulators, but it'd get him a built firmware
13:58:07Bagderin fact, we were using that before...and I changed to the native one, which is why this problem appeared
13:58:11ashridahsox: apple modified the c preprocessor to add extra junk for their IDE xcode
13:58:41ashridahwithout doign the sane thing of turning it off by default and having xcode set a flag to specifically add it (which is the intelligent approach)
13:58:42soxBagder: ok
13:59:01Bagderthis is truly the fault of Apple's gcc doing stupid things
13:59:17Bagderstupid stupid stupid
13:59:29Bagderbut it doesn't help you
13:59:32soxOK Ill call them asap :-/
14:00:19ashridahsox: i wouldn't bother. knowing them, they won't fix it until the next major release, if then.
14:00:29soxnah I was kidding
14:00:40ashridahyour best bet is to join the chorus bitching on their development forums and placing bug reports.
14:01:04soxso there's no simple workaround then...
14:01:18ashridahsox: yes, there is.
14:01:23ashridahuse m68k-elf-cpp
14:01:26ashridahinstead of 'gcc'
14:01:47Bagder$(GCC) is the target one
14:02:03ashridahin the SRC variable inside the firmware/Makefile file.
14:02:09Bagder$(CC) even
14:03:23Bagderthe problem was that my win32 cross-compile gcc doesn't support the -include option
14:03:29ashridahie, make it SRC := $(shell cat SOURCES | m68k-elf-cpp -DMEMORYSIZE=$(MEMORYSIZE) $(INCLUDES) $(TARGET) $(DEFINES) $(EXTRA_DEFINES) -E -P -include "config.h" - )
14:03:52ashridahwhich should do in a pinch
14:04:14Bagdersox: you need this fix in all Makefiles that do this trick
14:04:15ashridahbagder's right. m68k-elf-gcc
14:04:33ashridahhrm. and more than one place? ick.
14:04:58ashridahBagder: ouch. could at least have used CC instead of 'gcc' in the command :)
14:05:00Bagderbut we need to fix this
14:05:46soxthis works, but you're right, I have to modify all Makefiles
14:05:56amiconnBagder: Is it possible to replace your oldish windows cross-gcc with something newer? What would be necessary to do this?
14:06:00ashridahwell, googling for the parameter shows a lot of people bitching, the occasional reference to xcode, and the odd dodgy workaround
14:06:00Bagderashridah: it wasn't a nice fix anyway, it should use $(CC) and we should make it not use -include but instead add an #include "config.h" first in all SOURCES
14:06:37Bagderamiconn: I was about to the other day, but it is very sparse with docs and help for anything but the version we already use
14:07:47amiconnIiuc, this basically means building a gcc on Linux as a cross compiler to build windows binaries the same way as the cross-gcc for the targets build sh1 and m68k binaries, right?
14:08:24Bagderthe general idea would be that, but I've gotten the impression that it isn't supported in the main gcc branch
14:08:51amiconnSo how did you create the old cross gcc?
14:08:52Bagderthat's why this is a mingw cross-compile thing
14:09:13BagderI'm afraid to tell ;-)
14:09:22Bagder... I downloaded a binary
14:09:34soxashridah: I had to modify the makefiles in apps/, plugins/, plugins/lib/ and firmware/ but after doing it it compiled fine
14:09:42soxashridah: so thanks for the help
14:10:21soxashridah: i hope this is something that you can solve (since I can't, lack programming abilities ;-()
14:10:54amiconnBagder: Maybe this url is helpful:
14:12:17Bagderfor "egcs on Linux"
14:12:42Bagderbut I'll keep the URL and try next week
14:13:27ashridahsox: it's really something apple can 'solve'
14:13:30ashridahwe can only really work around it
14:13:51soxashridah: I understand that
14:14:20soxashridah: and this workaround is OK, not very time consuming
14:14:20amiconnBagder: "EGCS was an experimental step in the development of GCC, the GNU C compiler."
14:14:27BagderI know
14:14:33Bagdereven older than the version we use atm
14:14:45Bagderbut it might still work with newer versions
14:16:27ashridahsox: yeah, well, someone will probably roll something into the makefiles shortly. it's just not really a perfect solution
14:17:40CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
14:17:40*sox ponder about what "rolling something" into a makefile means
14:17:51Bagder"fix it"
14:17:55ashridahapply patch
14:19:25*sox feels enlighted!
14:21:44Bagderadded a note about this on ThingsTodo
14:21:55 Join cYmen [0] (
14:25:57BagderI'm off for a week now
14:25:59Bagdersee ya
14:26:02 Quit Bagder ("Off to search for that connect-resetting peer guy!")
14:26:17 Quit lostlogic ("Going to the moon")
14:31:44ashridahuh. what's bagder's actual name? :)
14:34:48ashridahheh. i should have known
14:35:02ashridahtoo late now. i'll probably get told this on the mailing list too
14:35:15ashridahnevertheless, i've summarised the situation there.
14:38:23 Join elinenbe [0] (~elinenbe_@
14:41:16ashridahsox: you still around?
14:41:36soxashridah: japp
14:42:36ashridahcan you try using 'gcc -traditional -D__GNUC__' instead of m68k-elf-gcc in those makefiles and see if it makes a difference? i'm not expecting it to, but it's worth a short
14:42:38ashridahshot even
14:43:40ashridahhm. might be able to drop -D__GNUC__
14:44:26soxashridah: tried both, the first one gave me this error: <command line>:6: warning: "__GNUC__" redefined
14:44:26sox<command line>:1: warning: this is the location of the previous definition
14:44:34ashridahyeah, that's okay.
14:44:49ashridahyou can drop the -D__GNUC__ bit.
14:44:51soxthe other one the usual : make[1]: *** No rule to make target `#pragma'
14:44:58ashridahno fix there then
14:45:28ashridahi suppose an easier check would be to get you to have run echo "hi" | gcc −−traditional -P -E -
14:45:29ashridahanyway :)
14:45:41*ashridah keeps searching for a fix
14:45:56soxahridah: thanks for your work....
14:45:56 Join AC [0] (
14:46:02ashridahthe fixes i'm seeing seem to suggest that an older cpp still exists on apple boxes.
14:46:35soxahridah: i got that impression too
14:46:41ashridahsox: okay, try echo "hi" | gcc -undef -P -E -
14:46:56ashridah(just run it)
14:47:06ashridahsee if it still outputs the pragma
14:47:10sox#pragma GCC set_debug_pwd "/Users/svante/rockbox/rockbox/build"
14:47:34*AC checks wav generated from wv2wav
14:48:05*ashridah resumes kicking apple
14:48:41*sox joins in kicking
14:49:28ACwav file is 152 kb big, but it seems to be damed :(
14:53:33ACwavpack codec seems hell fast
14:54:53amiconnashridah, sox:
14:56:53soxashridah: do you think that solution might apply to rockbox as well, adding the -x
14:56:53soxassembler-with-cpp to all makefiles and hoping it works on other platforms?
14:56:59ashridahyeah, that's what bagder suggested before, but sox seemed to think it didn't work.
14:57:14soxashridah: I didnt?
14:57:15ashridahsox: it'd be easy enough to test the platform in tools/configure
14:57:17linuxstbAC: Am I correct in saying that you have a running wv2wav, but you have a problem with the wav output?
14:57:28ashridahsox: i thought you tried it
14:57:35ashridahof course, you might have tried it in the wrong place ;)
14:58:07AClinuxstb: yep... it seems to decode correct, but must look if i get a wav-checking tool, to see whats wrong
14:58:13amiconnanother solution:
14:58:15soxashridah: tell me again and i will try it
14:58:21ashridahamiconn: the problem is, that's a command for the preprocessor specifically
14:58:45ashridahso 'gcc -P -xassembler-with-cpp' isn't right. need to modify it to send it to the preprocessor (which i don't know, but is probably in the manpages)
14:59:12***Saving seen data "./dancer.seen"
14:59:16AClinuxstb: will try a bigger file
14:59:25ashridahsox: hang on.
14:59:34soxashridah: im here waiting
15:00:19linuxstbAC: If you have a copy, the command-line "sox" program is very useful. It includes a program called "play", and you can use that to play the wav file as if it was raw data - so you can try different combinations of parameters (such as endianness, signed/unsigned etc). That will tell you if the WAV header is wrong, or if your data is not signed little-endian.
15:01:17AClinuxstb: will look if portage has sox in it :)
15:02:40*sox wonders why everybodys talking bout me ;-)
15:03:14ashridahsox: because the problem is something every apple user will face
15:03:18ashridahas they update
15:04:38ashridahhrm. the manpage suggests one can use echo "hi" | gcc -Wp,xassemble-with-cpp -P -E -
15:04:50AClinuxstb: about 120000% with wv2wav.. its very hard to ready, as it is changeing so fast
15:05:00ACwill take a foto
15:05:01ashridahbut that bitches about xassemble-with-cpp not being found as a file, which is clearly wrong
15:05:14dwihnoAC: cool
15:06:13ACfoto is quite ugly..
15:06:14amiconnashridah: What do you think about the solution of passing the generated file through sed to get rid of the #pragma line?
15:06:26*AC hast only a handy with a shit camera in it
15:07:09ashridahheh. sounds good enough to me.
15:07:37*AC cant ready how many seconds the decoding is running
15:07:38ashridahgot to make sure it deals with directories with spaces in the name tho
15:07:48ashridahwhich could get a little frustrating.
15:08:19ashridahmight just be able to get away with matching the pragma and then trimming the entire line
15:08:38ashridahit seems like a common (and fugly, imho) workaround
15:09:17amiconnYeah, it doesn't look nice, but the the apple gcc behaviour isn't nice either
15:09:50soxtreat them like they treat us ;-)
15:09:55amiconnThe second link I posted does drop the whole line iiuc
15:10:37 Join lolo-laptop [0] (~lostlogic@
15:10:56ashridahit's just silly. they HAD to have realised it'd break other things. i can't understand why they just didn't add an -xcode option that does funky-magic[tm] for their specific build environment
15:11:38ashridahof course, apple do all sorts of strange things
15:11:48ashridahlike not rolling out bugfixes for opengl until the next MAJOR release
15:12:02linuxstbAC: Something is obviously wrong with the statistics your wv2wav is keeping. The "file_info.current_sample" variable should contain the number of samples decoded so far (one sample is normally 4 bytes (2 16-bit words)). You also need to set the file_info.sample_rate variable to the correct value (or just hard-code it to 44100 for now).
15:12:04ashridahso people who want to play recent games are forced to buy an upgrade for osX to play it
15:12:49 Quit R3nTiL ()
15:13:27AClinuxstb: i am filled in all infos
15:16:55linuxstbAC: What about file_info.samplerate ?
15:17:25ashridahamiconn: anyway, if you can figure out how to get gcc to pass the xassemble-with-cpp option to the preprocessor, we're on a winner.
15:18:07AClinuxstb: i have seen it yet.. mom will fix it
15:18:31ashridah... 'mom will fix it'?!
15:18:33amiconnHmm. I cannot try that. No Mac anywhere near me. I don't even know someone who uses a Mac
15:18:49ashridahamiconn: it should work as a command on all platforms
15:19:01ashridahie, it should be a generic gcc option
15:19:15ashridahas -xassemble-with-cpp is a generic cpp option.
15:19:15amiconnYes, but I cannot test if it does the desired thing
15:19:21ashridahthat's okay
15:19:29ashridahas long as it actually runs at all
15:19:44ashridahie, it doesn't toss out a weird error about the command line options
15:19:48ashridahthen we can get sox to test it
15:19:53ashridahwho, funnily enough, has a mac :)
15:21:57*sox feels almost like someone is making fun of me ... ;-/
15:22:38AClinuxstb: runns now wirh more then 100%.. speed is growing
15:22:41soxhoy then i demand that we make fun of windows users and their cygwin mess too ;-)
15:23:08ashridahsox: that's okay, windows makes fun of itself
15:23:16AClinuxstb: 115%
15:23:26ashridahsox: i'm just picturing the happy-nowhere mac advert parody, that's all
15:23:59soxashridah: im with you
15:24:05AClinuxstb: 118%
15:24:15*amiconn just doubts whether he should dig into the #pragma fun any further...
15:25:35 Join cYmen_ [0] (
15:25:39AClinuxstb: 166 sec for a song that is 5 min and 50 sec long
15:25:57 Quit cYmen (Read error: 104 (Connection reset by peer))
15:26:03AClinuxstb: and about 8,6 MB big
15:27:41ashridah"i don't feel i'm operating the mac so much as i'm just there, sharing the mac experience. And if i can do something useful, while the mac is willing, so much the better"
15:30:15 Quit DeadMan ()
15:31:40 Quit cYmen_ (Connection reset by peer)
15:32:05 Join cYmen [0] (
15:36:45soxashridah: still looking for solutions
15:38:09ashridahbetter solutions, yeah.
15:38:34ashridahfor the time being, replacing the 'gcc' with 'cpp -xassemble-with-cpp' should do the trick
15:39:21ashridah-xassembler-with-cpp even
15:40:02soxthat works
15:40:23soxbut we still have to modify all makefiles, no?
15:40:33ashridahwell, ultimately, yes.
15:40:33soxor can this work for other platforms as well?
15:40:37ashridahbut we'll do that in cvs
15:40:46ashridahwell, since all platforms use gcc, it should be fine
15:40:59soxwell, then its solved?
15:41:11*sox goes to bake a cake
15:41:13ashridahthere's workarounds.
15:42:53linuxstbAC: That's excellent news. Can you make a tarball, and I'll add it to CVS.
15:43:24ashridahi can't commit anything to cvs myself tho.
15:44:00soxashridah: neither can I (who couldve guessed that...)
15:45:42soxlinuxstb: how far away is getting sound out of rockbox on iriver?
15:46:09AClinuxstb: later... i want to get a correct wav file out of it... i am looking at the moment whats wrong. Could i have later a own cvs account.. current libwavpack doesn't support wv files version 3 and encoding isn't there also
15:46:25linuxstbThe actual sound playback is working, but no work has started on the actual infrastructure to link rockbox, the audio codecs and the hardware yet.
15:46:56linuxstbAC: I'm not able to give others CVS accounts - one of the core developers will have to do that (Bagder, LinusN or Zagor).
15:47:16AClinuxstb: ok.. i will ask one of them
15:47:30soxlinuxstb: aha, and that work is next on the list, or are other stuff more prioritized?
15:48:49linuxstbsox: Everyone has different areas they are working on. LinusN and others are working on the low-level hardware stuff, others are working on optimising the codecs, But I think a few of us (including me) are currently working on the audio playback part of things.
15:48:51ACiriver is little endian?
15:49:09linuxstbAC: No, big-endian!
15:49:26AClinuxstb: oh.. that is the probelm with wv2wav
15:49:38soxashridah: got to go, thanks for your work today, means a lot to me and three or four other mac users....
15:49:55linuxstbCheck my flac2wav code - I think that has an LE_S16 macro defined which you can use to swap bytes.
15:50:09AClinuxstb: thanks
15:50:22*AC makes a short break
15:54:04 Join R3nTiL [0] (z0rr0@
15:54:38 Quit sox ("time to go to kindergarten and get the kid...")
16:12:04amiconnlinuxstb: I wonder why you reinved macros... Rockbox has SWAB16, SWAB32 and SWAW32 to handle endianess
16:12:13amiconn*reinvent even
16:25:31 Quit ripnetuk ("Arrrrghhhh")
16:26:18 Quit R3nTiL ()
16:27:17 Join methangas [0] (
16:30:13linuxstbamiconn: Simply because I didn't spot the existing ones - or even expected them to exist. I'll start using them, thanks.
16:36:30amiconnWell, rockbox had to handle endianess for a long time. The SH1 is big endian too.
16:38:32amiconnHmm. The abovementioned macros are not asm optimised for coldfire, unlike the SH1 ones. Are yours asm optimised?
16:39:29 Quit [IDC]Dragon ("CGI:IRC (EOF)")
16:39:33linuxstbNo, they are only used is a couple of the WAV output routines at the moment - not in the core codec code.
16:40:17amiconnWell, the SH1 has byte and word-swap asm instructions. So SWAB16 and SWAW32 are one instruction, SWAB32 uses 3
16:40:39amiconnI guess similar instructions exist for the coldfire
16:40:53 Join preglow [0] (
16:41:30preglowcoldfire has swap instruction
16:42:08preglowonly swap.w, unfortunately
16:42:20 Quit ashridah ("Leaving")
16:42:40amiconnHmm. So at least SWAW32 can be coded as one instruction
16:43:20preglowasm("swap.w %0" : "=d" (input) : "d" (swapedinput)); should do the job
16:43:36preglow%1, i mean
16:44:05amiconnNot quite. First, %0 is correct for the first parameter
16:44:35amiconnSecond, if you only have the option that input and output are identical, you have to use the combined input/output sysntax.
16:44:47preglowyes, i don't know that syntax, so that was i guess ;)
16:45:04amiconnasm("swap.w %0" : : "+d"(data));
16:45:24amiconnOops, one colon too much
16:45:42preglowyep, then we agree
16:46:31preglowbut how to swap bytes quickly
16:46:45 Quit bobTHC ("( :: NoNameScript 3.81 :: )")
16:49:46amiconnpreglow: For SWAB16, there might be no faster solution than shift high byte right, shift low byte left, then or
16:50:18amiconnHowever, for SWAB32 there is probably a better method than shifting around all 4 bytes
16:50:55preglowswaw is probably the least used of them as well, so bah
16:56:14preglowlinuxstb: shouldn't linus' lates change have sped our libflac up somewhat? if i'm not mistaken, it has moved lpc_restore_signal_order to iram
16:57:58HCldammit. i need someone to help me with writing an return from subroutine instruction in memory, and calling it successfully without dying.. on iriver
16:59:14***Saving seen data "./dancer.seen"
17:00:13 Join XShocK [0] (
17:03:55preglowwhat makes that so hard?
17:05:57 Quit XShocK ("CGI:IRC")
17:09:54 Join XShocK [0] (
17:10:13HClpreglow: you tell me
17:10:15HCl :(
17:10:47HClsomeone should port a gdb stub :/
17:13:24XShocKhi. someone actually can change the status of sound in wiki, since it is interrupt based, and quality is perfect, the only DMA is not working yet, but digging the original firmaware helps.
17:13:45HClis it in cvs yet?
17:13:53XShocKno, not in cvs
17:14:12XShocKin original firmware DMA is used
17:17:03XShocKit can be put in cvs, but should it? because audio framework is not there anyway.
17:17:36HClnah, but update the link maybe? and put it online?
17:19:32 Join R3nTiL [0] (~zorroz@
17:19:39HClmaybe change status to 90% (dma not working yet)
17:19:40HClor something
17:20:08CtcpPing from R3nTiL!~zorroz@
17:20:10CtcpVersion from R3nTiL!~zorroz@
17:20:12CtcpTime from R3nTiL!~zorroz@
17:20:24XShocKsound input does not work yet.
17:20:32HCl50% then
17:21:16HClplus digital out.. digital in.. line out.. line in..
17:21:56XShocKyeah. :)
17:24:45 Quit R3nTiL ()
17:25:21 Join mecraw [0] (~mecraw@
17:33:26 Quit XShocK ("CGI:IRC (EOF)")
17:39:47preglowHCl: have you tried dumping the code you generate and running a disassembler on it?
17:42:55HClit seems fine.
17:43:15preglowand you're aware of the alignment issues on the coldfire?
17:54:10 Join R3nTiL [0] (~zorroz@
17:57:02 Quit AC ("CGI:IRC (Ping timeout)")
18:02:02 Join hubble [0] (
18:03:07 Quit elinenbe (" HydraIRC -> <- Chicks dig it")
18:07:37 Part Lurkski
18:07:49 Quit Patr3ck_ ("User pushed the X - because it's Xtra, baby")
18:14:40 Quit jyp ("poof!")
18:17:37 Quit R3nTiL ()
18:44:15 Join ]LoKi[ [0] (
18:47:49]LoKi[I've got a couple of questions, can anyone help?
18:54:11preglowask away
18:59:16***Saving seen data "./dancer.seen"
19:00:26 Join [LoKi] [0] (
19:00:39preglow18:54 < preglow> ask away
19:00:51[LoKi]soz, I got disc'ed
19:01:38[LoKi]anyway :P
19:01:38[LoKi]How can I get Rockbox to pic up a microphone hooked up to the Line In jack
19:01:52[LoKi]I choose the input and it's just silent
19:02:12[LoKi](V2 Recorder, btw)
19:03:47 Quit ]LoKi[ (Read error: 60 (Operation timed out))
19:04:53[LoKi]and the other one is... is there anyway to go to the next song without going to the File List when in Repeat 1 mode?
19:09:17 Quit [LoKi] ()
19:10:34 Join Tang [0] (
19:10:56Tanghi World :)
19:11:52 Join DrRick [0] (
19:18:40 Join Stryke` [0] (~Chairman8@resnet-241-86.resnet.UMBC.EDU)
19:25:17 Quit preglow ("woop")
19:33:40 Quit Renko (Remote closed the connection)
19:50:51 Join blahdood [0] (
19:51:03 Nick blahdood is now known as drewslater (
19:51:44 Quit drewslater (Client Quit)
20:02:21 Join Sucka [0] (
20:04:21 Join condor9 [0] (
20:05:41condor9I'm having a config.cfg problem with 2.4 stable ... can anyone help?
20:07:16amiconnAny linux user around?
20:07:40 Join preglow [0] (
20:07:54condor9hey amiconn ... you were helping me the other day ... got a second?
20:08:00amiconnIf yes, please check what 'echo $OSTYPE' on the command line outputs
20:08:25amiconncondor9: just ask...
20:08:55condor9My fonts aren't loaded from the config.cfg If I browse and load them they're fine.
20:09:09amiconnWhere are you fonts located?
20:09:16condor9I tried a simple config.cfg with just the font loading and that didn't work either.
20:09:45amiconnDid you try saving a .cfg after loading the desired font?
20:10:09condor9Yes. 2.2 is ok, but 2.3 and 2.4 are not. I used the new convbdf to convert from bdf to fnt
20:11:20amiconnHmm, strange. Works perfectly here.
20:11:50condor9Yeah, seems to work for most people. Let me ask, could it be a problem with HD cache ... too large perhaps?
20:11:56amiconnPerhaps the font name is too long?
20:12:09condor9I renamed the font to "bold.fnt" :)
20:12:35amiconnHmm, that should definitely work.
20:13:06amiconnDidn't you have other problems with .cfg files as well?
20:13:10condor92.2 is works great but I would like to rombox my recorder, so I need 2.4.
20:13:20condor9Same problem. I isolated it to font loading.
20:13:48condor9Other settings are restored ... the font is the only thing that doesn't load.
20:14:18amiconnWhat line ends does your .cfg file use?
20:14:51amiconnOther method, simply drop me your .cfg and the font in question, I'll try it here
20:14:55condor9I tried the default ^M and then removed them ... doesn't seem to matter.
20:15:42condor9I'm in a coffee shop without my recorder, otherwise I'd send you my config.
20:16:01condor9I also tried moving the font line to different positions in the config ... that didn't work either.
20:17:29condor9It doesn't matter what font I use ... none of them autoload ... just the default.
20:19:52amiconnThe last selected font should even autoload on boot, without using a .cfg file, if it is (1) located in /.rockbox/fonts and (2) doesn't have a too long name
20:20:03amiconn...and (3) is a vaild font of course
20:20:33amiconnUnfortunately I have no idea what fails if it doesn't.
20:20:40condor9Yeah, I know ... that's why I'm asking here. :)
20:21:27amiconnHmm. Other settings are preserved across reboots?
20:21:48amiconnAH, you said 2.2 works... so I guess they are
20:22:50condor9I know the font format changed ... so I made sure to use the new convbdf. The font works if I just browse to it. Very strange.
20:23:34condor9Is there some way I can debug the loading of a config file?
20:23:42amiconnHmm. I'm afarid I can't help without being able to reproduce the problem
20:24:06condor9Ok. Thanks for your time amiconn.
20:27:00amiconnYou may drop me the font & config file when you have access to your box again.
20:30:35linuxstbamiconn: I've ssh'ed around a few machines I have access to: Debian - OSTYPE=linux-gnu, RedHat - OSTYPE=linux-gnu, SuSE - OSTYPE=linux, Solaris - OSTYPE=solaris, MacOSX 10.2 - OSTYPE=darwin
20:30:35 Quit condor9 ("Leaving")
20:31:10amiconnlinuxstb: Thanks. So this variable seems to exists almost everywhere... nice :)
20:31:21linuxstbDoes cygwin have it?
20:31:29amiconncygwin - OSTYPE=cygwin
20:31:54amiconnNow I only need to know how to check that env variable in a Makefile...
20:32:14linuxstbCan't you just use OSTYPE in the Makefile?
20:32:23amiconnNo, it's empty....
20:33:45amiconnIt seems that all variables which are accessible in all rockbox Makefiles are defined in the build-dir's Makefile, which is created by configure
20:35:50linuxstbThen just use configure to add it to the build Makefile.
20:38:47amiconnHmm, here's why I'm digging for that. There are 3 related problems with the cygwin build:
20:39:44amiconn- plugins are dlls on Windows. These need the execute flag set, otherwise they can't be executed. So there is a check in the makefile, comparing UNAME. However, this doesn't work
20:40:17amiconn- The uname variable is set in configure, but not exported to the Makefile. This alone would be easy to solve
20:41:11 Join Zezayer [0] (
20:41:15amiconn- But, uname isn't constant on cygwin, it has the form CGYWIN_NT-5.1
20:41:40amiconn- Makefiles can only check for equal/not equal, so this would need mangling
20:42:05amiconnSo OSTYPE seems to be much better...
20:42:34 Join XShocK [0] (
20:44:34 Quit Tang ("Chatzilla 0.9.66 [Mozilla rv:1.7.5/20041108]")
20:44:58linuxstbI've just tested, and the line "ifeq ($(OSTYPE),linux-gnu)" seems to work in a Makefile
20:45:34linuxstbGNU Make 3.80
20:47:49linuxstbamiconn: I think I'm mistaken, I think it's only used if you type "export OSTYPE" in the shell before typing "make".
20:51:40linuxstbUnless I'm misunderstanding, you need to do something extra only for the win32 simulator - so why don't you test "SIMVER" in the Makefiles?
20:52:07amiconnBecause the problem also occurs when building the x11 sim on cygwin...
20:52:34linuxstbBut what if someone is cross-compiling? either to or from windows?
20:54:12amiconnI don't know exactly what will happen in that case. The old check tried to do the same what I intend to do, but with UNAME, and it failed because of the non-exported variable
20:59:18***Saving seen data "./dancer.seen"
21:00:20linuxstbThere's already a lot of variables in the master make file, so I don't think adding OSTYPE would be a problem.
21:00:56amiconnI didn't figure how to access OSTYPe within the script, so I added UNAME instead, as was probably intended anyway
21:01:24amiconnThen I needed to change the check in apps/plugins/Makefile to use findstring
21:02:26linuxstbIn tools/configure, just add the line "export OSTYPE=$OSTYPE" to the end of all the export statements. You will then have the line "export OSTYPE=cygwin" in your Makefile.
21:04:07 Quit Aison (Read error: 54 (Connection reset by peer))
21:04:22amiconnNow there are only those 2 warnings left for the win32 sim, but I don't get these when building myself. It is most likely a problem of the oldish crosscompiler used on the official server. GCC 2.95 you know...
21:05:17amiconnThe dependencies also need fixing, but I don't know how to do that....
21:11:27 Join sox [0] (
21:12:37 Quit sox (Client Quit)
21:12:52 Join sox [0] (
21:15:18soxamiconn: couldnt that OS specific configuration be useful to get the workaround for the mac os x pragma problem as well?
21:16:23 Quit Zezayer ("ChatZilla 0.9.61 [Mozilla rv:1.7.2/20040804]")
21:16:51amiconnCertainly, the UNAME variable could be checked for this too. The question is how to hand the option to the preprocessor...
21:17:14soxamiconn: eg, cpp -xassembler-with-cpp in the Makefiles
21:17:29soxamiconn: instead of gcc
21:19:11 Join markun [0] (
21:19:51markunIs anyone here running the X11 uisimulator?
21:20:06amiconnsox: Does gcc -Wp,-xassembler-with-cpp also work?
21:21:03amiconnOf course in addition to the other options
21:21:55soxamiconn: hang on, testing
21:22:23markunuisimulator crashes here when trying to scroll a line..
21:23:24linuxstbmarkun: Yes, I noticed the other day, but I had forgotten about it.
21:23:57amiconnlinuxstb, markun: Linux?
21:24:01linuxstbI can't remember ever seeing it work - I very rarely have any files in my archos directory.
21:24:07linuxstbYes, Linux.
21:24:23markunI was trying to debug my unicode stuff and thought I broke scrolling, but it also fails for a clean checkout.
21:24:24amiconnLemme check with cygwin, perhaps it ddoes also crash...
21:24:31markunamiconn: no, FreeBSD.
21:25:02soxamiconn: no: cc1: error: unrecognized option `-xassembler-with-cpp'
21:25:15markungdb tells me it segfaults in lcd-x11.c:118, but I can't see anything strange about it.
21:27:10 Join stripwax [0] (
21:27:36linuxstbgdb tells me line 381 - when accessing the pf->width array.
21:27:50amiconnhmm, scrolls like mad here... (recorder sim). Building iriver x11 sim...
21:29:09markunlinuxstb: I hope it wasn't anything I changed before comitting Tremor that makes LOW_ACCURACY produce only static..
21:29:59linuxstbamiconn: The segfault I get is in line 381 of drivers/lcd-recorder.c
21:30:18soxamiconn: so gcc -Wp,-xassembler-with-cpp does not work
21:30:26linuxstbmarkun: No, I don't think so. I compared your version and the original, and didn't see any significant changes.
21:32:12amiconniRiver x11 sim also working.. strange
21:33:39markunYes, very strange.. Maybe it's because you use a different X server.
21:34:31amiconnflac2wav doesn't compile here for the sim... undefined reference to `_iramstart' / undefined reference to `_iramcopy' / undefined reference to `_iramend'
21:34:46preglowyes, linus committed those yesterday
21:35:21amiconnYes, I know. However, I wonder why this doesn't work here, while the official builds seem to work
21:37:44soxso now Im out in the cold building the sim: "Unsupported system: Darwin, fix configure and retry"
21:41:06soxI guess I could fix it if I knew the right LDOPTS for Darwin
21:42:09linuxstbDoes anyone know what the circle with a dot inside means in the top-right corner of the screen?
21:42:38linuxstb(I'm using the iRiver X11 simulator)
21:44:51amiconnHmm, I don't get this here, but then I can't use 'make install'
21:45:17amiconnmake install tries to remake missing files, and it fails because of those symbols I mentioned...
21:45:36amiconnThis symbol shouldn't be there, or does the iRiver not have a hd led?
21:46:05preglowit does
21:46:28linuxstbamiconn: are you talking about my symbol?
21:46:34amiconnlinuxstb: yes.
21:47:14amiconnThe Ondio does not have a disk (well: MMC) led, so [IDC]Dragon implemented an activity indicator in the top right corner.
21:47:23linuxstbI've sort of tracked down the crash (but not the reason). When I scroll down slowly (pressing and releasing the down error), then that symbol appears when I get to the penultimate line. If I then press any key, it wil crash in line 381 of lcd-recorder.c
21:50:41amiconn??? lcd-recorder.c !!! I thought you were building the iriver sim...
21:51:29 Quit sox ("This machine just fell asleep")
21:51:37linuxstbI am - I guess that's the place for generic graphical LCDs.
21:52:12amiconnNote quite. For iriver, lcd-h100.c is used instead
21:53:18amiconnlcd-recorder.c is for archos recorders, ondio, and gmini
21:54:13linuxstbStrange. My build/Makefile contains "TARGET=-DIRIVER_H100", but the sim is definitely building and linking drivers/lcd-recorder.c
21:54:47amiconnYes, it is doing the same here, and that is most likely the problem
21:55:14amiconnThere is a maximum number of scrolling lines; I doubled that for iriver some time ago
21:55:39amiconnSo if the sim uses the wrong file, an out-of-boud array access will happen...
21:55:47 Join Aison [0] (
21:56:40hubbleXShocK: yo, are you there?
21:56:48hubbleXShocK: i've got Dma sort of working
21:56:54linuxstbYep, that's the bug. I renamed lcd-h100.c to lcd-recorder.c and the sim now works fine.
21:57:58hubbleXShocK: I'm using "DCR0 = DMA_INT | DMA_EEXT | DMA_CS | DMA_SINC "
21:58:04 Join Chamois [0] (
21:58:36hubbleXShocK: so it good news, only bad thing is that the pitch is little wrong (too slow) =)
21:58:40linuxstbamiconn: The problem is in firmware/SOURCES - the CONFIG_LCD variable isn't defined for the simulator builds. So it defaults to lcd-recorder.c
21:58:56amiconnI was about to say the same...
21:59:53 Nick Sucka is now known as Sucka`away (
22:00:32hubbleXShocK: think it's because my pcm buffer is in SDRAM.. going to move it to SRAM now
22:02:03amiconnlinuxstb: It's probably the same thing that causes the funny symbol. HAVE_LED is also only defined for the targets (that have one)...
22:02:39amiconnI'm not sure how to handle this :(
22:03:01preglowthey should have bundled a couple of megabytes of sram
22:03:16linuxstbNo, nor am I. I wonder why those config-*.h files don't define the hardware for the simulators.
22:04:25amiconnImho there shouldn't be that #ifndef SIMULATOR...
22:04:49linuxstbamiconn: I've just deleted it from config-h100.h and I will see what happens...
22:06:04linuxstbI now get an error compiling system.c - it is trying to compile some assembly code.
22:07:14amiconnHmm, that's bad.
22:07:46amiconnI guess the current problem also results from the simulator build system change
22:08:27linuxstbIt looks like lots of the code depends on those #ifdefs being defined only for the target. The SIMULATOR variable isn't used.
22:09:16amiconnIf any, only CONFIG_CPU should be protected by #ifdef SIMULATOR, imho
22:09:36amiconn#ifndef SIMULATOR of course
22:09:38 Quit DrRick ()
22:13:36amiconnHmm. Doesn't work either...
22:14:30 Join olivierd [0] (
22:16:32olivierdhi there, trying to compile gcc for coldfire on cygwin
22:17:05olivierdI get an error in the m68040 part, is there a way to only build for the coldfire target?
22:18:16preglowm68040 part?
22:18:33preglowwhy would m68040 be involved?
22:19:31olivierdI use m68k-elf as a target and it seems to build a lot more than coldfire
22:23:57preglowit probably builds for all 68k, yes
22:23:58linuxstbamiconn: The only variable I can see to use is the CONFIG_KEYPAD variable. It's not the right thing to do, but it works.
22:24:05preglowbut i've never seen an error
22:24:10preglowwhat gcc version?
22:24:57amiconnlinuxstb: Okay. It's ugly, but it works for now. You could commit this, with a comment stating FIXME or such...
22:26:28linuxstbOK, I'll commit it.
22:29:04 Quit olivierd ("CGI:IRC (EOF)")
22:29:53 Join olivierd [0] (
22:32:01 Join Renko [0] (
22:33:25 Quit olivierd (Client Quit)
22:36:18XShocKhubble: i am here finally :)
22:36:34 Join olivierd [0] (
22:37:11XShocKhubble: cool! :)
22:37:15XShocKcan you send it to me?
22:37:30olivierdpreglow: sorry, got caught up in a fight with my IRC client
22:37:45preglowolivierd: didn't have much to add anyway, what gcc version?
22:37:52ChamoisXshock : when WAV will be commit on CVS ?
22:37:53preglowolivierd: are you trying to build, that is
22:38:11preglowolivierd: weird, works fine here. binutils?
22:38:29olivierdcvs as instructed in the wiki
22:38:58XShocKchamois: i would guess when kind of sound framework will be done.
22:39:11preglowolivierd: weird, i've never had that not work
22:39:42XShocKor of course it can be added somewhere near, just giving a basic sound API
22:40:06 Join Patr3ck [0] (
22:40:20Chamoissound is more than the 10% we can see on the wiki ?
22:40:44Chamoishubble and you made lot of progress ? no ?
22:40:48XShocKchamois: hubble did DMA now, so i guess it is somewhere near 50. :)
22:41:02Chamoisok ! good !
22:41:06preglowwith recording being the remaining 50%?
22:41:14stripwaxheh :-)
22:41:19olivierdwell... I've been trying to get it to work for a few days. gets operands mismatch in .../src/gcc-3.4.3/gcc/libgcc2.c -o libgcc/m68040/_fixunsdfsi.o
22:41:23hubbleum.. how big is the SRAM of iriver ?
22:41:25preglowXShocK: like it works perfectly?
22:41:30preglowhubble: 96kbyte
22:41:34preglowhubble: try not to use too much of it
22:41:38stripwaxhubble - 64KB can be used by DMA
22:41:50preglowhubble: yes, it needs to be in the uppers 64kb
22:41:59hubbleStrath: um.. what?
22:42:03hubblestripwax: what?
22:42:27stripwaxhubble? DMA controller can only use the 64KB SRAM bank not the 32KB bank
22:42:29preglowhubble: only 64kb out of the 96 can be accessed by dma
22:42:38XShocKit sounds perfect even without DMA. with AUDIOTICk it tend to get more CPU, but still it sounds very good
22:43:06preglowXShocK: how about an example rockbox.iriver? :)
22:43:07hubbleum.. well.. i just tried to use DMA to transfer from SDRAM and that was no problem
22:43:08stripwaxhubble what did you use to measure that it plays slowly without SRAM?
22:43:15XShocKhubble: i think there would be problems, i also read somewhere that SRAM will be problem
22:43:23stripwaxhubble you were talking about SRAM not SDRAM, no?
22:43:28XShocKsince it is not addressable by DMA
22:43:42preglowsram is addressable by dma
22:43:44preglowjust not all of it
22:43:50XShocKpreglow: ok. :)
22:44:00hubblehm.. oh, what.. i used a buffer allocated by buffer_alloc
22:44:04preglowyou have access to 64kb of it, but don't even think about using that much, heh
22:44:29stripwaxCouldn't it all be used for PCM buffer?
22:44:34preglowhell no
22:44:38stripwaxWhy not
22:44:41preglowwe need space for performance critical code and data
22:44:46stripwaxthe other bank?
22:44:47XShocKhubble: aah, then ok. but still.... i think SRAM is not very needed,
22:44:48preglowlots of space
22:45:36hubbleXShocK: you dont? it didn't run at normal pitch, but maybe that's because of cpu speed?
22:45:42Strathah... username autocomplete :)
22:45:49XShocKhubble: since it even worked with PIO without SRAM, i don't see why wouldn't it work with DMA
22:46:11XShocKmaybe there is another prblem?
22:46:16hubbleXShocK: yes, maybe
22:46:36preglowhow do you know it doesn't run at normal pitch? what's the reference
22:46:58stripwaxIs this the output of the sine or actually a WAV output?
22:47:09preglowyou do ofcourse set the sample frequency? :)
22:47:18hubblestripwax: so, um.. where's the upper 64kb of sram? 0x10010000 -> 0x100020000 ?
22:47:49hubble0x10010000 -> 0x10020000 i mean
22:48:18stripwaxWell it can be moved by MBAR
22:48:20hubbleiriver's dma buffer starts at 10000810
22:48:49preglow /* 64K DMA-capable SRAM at 0x10000000
22:49:31hubblepreglow: oki.. so it's 0x10000000 -> 0x10010000
22:49:38 Quit Chamois ("CGI:IRC (EOF)")
22:49:43preglowhubble: well, it would pretty much have to be
22:49:57preglowthat's where the non-dma sram starts, at least
22:50:03preglowit's set up in crt0.s in firmware/
22:50:39stripwaxoh so only the 32KB can be used by DMA?
22:51:02stripwaxscratch that, misread
22:56:22stripwaxMeant RAMBAR not MBAR, obviously. Both SRAM banks can be moved to any 16KB boundary, independently
22:56:33XShocKi think SRAM is not very needed, since if we use it then we anyway need another bigger PCM buffer. I am sure that I am making some big mistake. But why anyone would want to copy data from PCM buffer to SRAM, since it would take even more CPU cycles than if we just make DMA copy from PCM buffer. I think bigger PCM buffer is needed since if we use crossfading between tracks, it is more than SRAM would give space for.
22:56:54XShocKPlease correct me if i am wrong, since I am learning. :)
22:56:56stripwaxby the way, there's theoretically power management benefits from marking either bank as only holding data, or code.
22:57:42stripwaxCouldn't the decoder decode directly into SRAM (i.e. PCM buffer IS in SRAM) or would that just not work?
22:57:54 Nick Sucka`away is now known as Sucka (
22:58:08XShocKyes, it will, but corssfading between tracks is not possible,
22:58:39XShocKor it is possible, but will make crossfading much harder for processor.
22:58:59stripwaxAs for crossfading, probably need two PCM buffers in SDRAM and generated the crossfaded buffer into SRAM so again no copying
22:59:23***Saving seen data "./dancer.seen"
23:00:18stripwaxHow were you planning to implement crossfading?
23:01:12linuxstbEven if the whole 64K of DMA'able SRAM was used for PCM data, it would only hold 16000 stereo samples - which is only about a third of a second. I think we would need a bigger buffer than that, even without cross-fading.
23:01:53stripwaxReally? What for?
23:02:15amiconnImho the pcm buffer should only hold a few frames worth of samples
23:02:15linuxstbI'm thinking mainly for safety - we don't want the buffer to ever become empty.
23:02:42stripwaxAgreed - what would cause it to starve?
23:03:14XShocKone idea is that if we get decoder decode at somewhere near 105-110%(throwing info, did not calculate exactly), then before the track finished playing, the decoder already start to decode second track(several seconds before) and mix the info to the end of the first track. of course it require tracks to be a normal length( but i doubt that there are mant tracks shorter that one minute)
23:03:15hubblethe wierd thing about the original iRiver firmware is they only transfer 128 bytes each DMA transfer (and that buffer is 128), although they have 2 x ~9000 byte buffers in SRAM also.. I have not found out how they fill that 128 byte buffer
23:03:58hubbleerr, 256 not 128 =)
23:04:01 Join midk [0] (
23:04:15XShocK0x80... 128 i think
23:05:06 Quit Strath ("Client closed")
23:06:06hubbleXShocK: but the buffer seams to be 256 bytes if you look at 10000810 (the next buffer starts at 10000910)
23:08:03preglowhubble: have you got an ida setup for the firmware?
23:08:42XShocKoops.. question not to me.. :)
23:08:53hubblepreglow: you want?
23:08:57preglowi don't care who answers, what i want to know is, may i have it? ;)
23:09:01stripwaxanyone ever make any modifications to the coldfire emulator? I never got a chance to do much but it seemed like it's not really necessary?
23:09:13hubblepreglow: sure
23:09:27preglowstripwax: depends, it sure would be nice, but a gdb stub just might be easier and better
23:09:42preglowbut would require a hardware mod, i guess
23:09:43stripwaxhubble does your ida setup handle the CF5249 instructions?
23:10:02hubblestripwax: no, i have to manually fix some things
23:10:05XShocKhubble: where does it set the buffer to be 0x100, since i see only 0x80 even in the handler.
23:10:32stripwaxpreglow - a gdb simulated target might be better but then would probably just be a port of the coldfire emulator to the gdb sim framework :-(
23:10:45hubbleXShocK: it doesnt, but the buffer itself seams to be 0x100
23:10:47stripwaxgdb hardware mod should be easy enough however
23:10:56XShocKaahh, ok :)
23:11:00stripwaxoops, ^gdb^serial
23:11:52hubblepreglow: i use the rule that better to comment stuff (even if it might be wrong first) and fix later.. DCC or mail?
23:12:27preglowhubble:, please
23:18:34 Quit preglow ("reboot")
23:20:24 Join preglow [0] (
23:21:24preglowhubble: why 1.40, btw?
23:21:24HClpreglow: the iriver has a fairly accessable serial port, the gdb stub can use that, all you have to do is solder some wires to it to make it work, i think
23:21:42preglowHCl: yeah, know, just have to find a suitable place on the player for wires to be sticking out :P
23:21:53preglowHCl: you'd also need a logic level converter
23:21:58HClwell, we could write a gdb stub that works on audio signals :P
23:22:13preglowserial over s/pdif
23:22:34hubblepreglow: i started looking at the iriver_memory_map wiki page and tried to follow.. had lots of problems before i understood things =)
23:22:44*HCl goes to sleep..
23:27:16stripwaxXShocK/hubble are you using bdm or not?
23:27:52hubblestripwax: no, just the bootloader
23:28:09*HCl sighs and wishes he knew why rockboy crashed :/
23:28:58stripwaxyeah.. I bought a BDM wiggler but haven't got around to opening the iRiver yet. I can't really afford to just donate it to Rockbox but I'll consider selling it at a discount if anyone needs it?
23:29:06Renkohas any1 built the bdm from those plans?
23:29:20XShocKall i have is USB cable and the player itself. :))
23:30:57preglowhubble: what ida version did you say you used?
23:31:40hubblepreglow: 4.7
23:32:19stripwaxWhat's diff over 4.3 or 4.5?
23:33:04hubblestripwax: think 4.7 is the first with the integrated debugger (for x86).. hum.. other than that i dont know
23:33:42 Quit olivierd ("Chatzilla 0.9.67 [Firefox 1.0/20041107]")
23:35:15*rasher punches iRiver's m3u support
23:35:25XShocKmine is also 4.7, and i couldn't open the hubble's base with neither 4.15 or 4.3
23:38:05rasherWhat is the requirements of the m3us?
23:38:14rasherFor iRiver to be able to load it
23:41:14*HCl hasn't had any problems with m3u support..
23:44:20 Quit methangas (" HydraIRC -> <- 100,000+ downloads can't be wrong")
23:44:45stripwaxrasher - no absolute drive letters for one..
23:45:10stripwaxI don't use them however
23:45:35stripwaxoh, and the direction of the slash must be correct (tho can't remember if it's forward slash or backslash..)
23:56:36hubblepreglow, stripwax: do you know if you have to use DMA1 to transfer from SRAM (cause iriver does that and when I try to use DMA0 it works from SDRAM but not from SRAM) ?
23:58:38rasherstripwax: Thanks
23:58:42preglowhubble: no idea
23:58:59stripwaxhubble, what preglow said

Previous day | Next day