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 2010-11-22

00:03:37 Join MethoS- [0] (~clemens@
00:10:08 Nick Ypsy is now known as YPSY (
00:13:57Lloreansaratoga: The beast also charges from USB
00:14:05LloreanBut unlike the F, it doesn't charge from USB when turned off (the F does)
00:14:27saratogado I need to force it? i just plugged it in and it said discharging until i added ac power
00:15:18 Quit ender` (Quit: "I'm the head wizard now. I've only got to give an order and a thousand wizards will ... uh ... disobey, come to think of it, or say 'What?', or start to argue. But they have to take notice." -- Archchancellor (Terry Pratchett: Lords and Ladies))
00:15:19LloreanI generally don't have to.
00:15:27LloreanJust hold Menu while plugging in, and there it goes.
00:15:49LloreanIt doesn't show the same charging display that AC power does, but battery charge does go up over time
00:16:01 Quit dfkt (Ping timeout: 245 seconds)
00:16:25saratogahmm it didn't go into USB mode either, so maybe i should try a different cable
00:20:49saratogaoh you said menu, i was expecting it to be select like on teh sansas
00:20:51saratogathat worked
00:21:12saratogaany reason we have a specific button to hold to get charge mode? it would be easier for users if you could hold any button
00:22:53 Join webguest43 [0] (
00:24:22 Quit webguest43 (Client Quit)
00:26:56saratogaaligning a few structs in libmad speed it up on the beast
00:26:59Lloreansaratoga: I've been saying it should be any button for a while. I haven't really heard anyone argue strongly against it.
00:27:10saratogaany idea where the check is?
00:27:25saratogagrep suggests usb.c or powermangmt.c
00:27:39LloreanNo idea.
00:28:18LloreanI think people have argued against it, but I honestly can't remember anything beyond "some buttons could have unexpected behaviour like fast forward" which doesn't seem much of a real problem / likelihood (I think people will realize holding a button will do whatever that button *does* until they insert the USB cable)
00:30:05 Quit stripwax (Quit:
00:30:29 Join ReimuHakurei [0] (
00:30:43saratogai wonder how you force alignment in a .S file
00:33:41 Quit dys (Ping timeout: 276 seconds)
00:33:54 Join JdGordon| [0] (
00:33:54 Quit JdGordon| (Changing host)
00:33:54 Join JdGordon| [0] (~jonno@rockbox/developer/JdGordon)
00:34:32 Quit JdGord (Quit: Bye)
00:37:59JdGordon|saratoga: the usb connect thing is something that always get discussed but nothing ever happens with
00:38:10saratogado you know where the check is
00:38:21JdGordon|I personally think a menu on connect is the best way to do it but noone will come to an agreement
00:38:42JdGordon|no, gevaerts would probbaly know
00:38:52JdGordon|or search for button_get() in firmware/
00:40:44gevaertssaratoga: usb_power_button() in usb.c
00:41:13soapFWIW I believe I can pick up a Cowon S9 for $175. Not cheap, but cheaper than $250.
00:41:14gevaertsThe button to use is in usb.h
00:41:30saratogawhat header do i need to use MEM_ALIGN
00:41:34LloreanJdGordon|: Well there still needs to be a decision for cases where a menu can't come up, or be used for whatever reason, so menuless behaviour is still good whether or not a menu will go in in the future.
00:41:56saratogait shouldn't check for any particular button, just if (btn)
00:42:01saratogano bitmask
00:42:22JdGordon|that has been one of the common arguments
00:42:25soapahh, nevermind. That was a local price. On ebay one can score a refurb for less than that.
00:42:45*JdGordon| is hoping to score a mini2g for $30 :)
00:43:28saratogawtf adding codec.h is enough for layer3.c, but not for huffman.c
00:43:36saratogacan one macro depend on two headers?
00:45:27saratogaand make clean returns to both not compiling . . .
00:47:44JdGordon|sounds like you got yerself a broken header :p
00:47:58saratogahow the hell do i get system.h into a codec
00:49:53JdGordon|what do you need from it? it might be wrapped in !PLUGIN/CODEC?
00:50:20saratogai just want to use mem_align in a codec
00:50:31saratogaand it randomly works or doesn't work in some files depending on the make -j setting
00:53:01saratogahuh __attribute__((aligned(16))) doesn't work either
00:53:11saratogaare there restrictions on what types of variables can be aligned?
00:53:21JdGordon|is your system stuffed?
00:53:38JdGordon|that all just sounds broken
00:54:13***Saving seen data "./dancer.seen"
00:54:52saratogano i see what it is
00:55:05saratogathe attribute has to go before the =
00:55:34saratogaI was doing "int x = 0 MEM_ALIGN;"
01:03:43saratogaanyone available to test on android?
01:04:23Jonas29is it possible to go on web site with rockbox
01:04:25saratogai get a small speed up on the beast, but I'd be curious what its like on Android where alignment seems more important
01:06:47 Quit markun (Ping timeout: 240 seconds)
01:07:14saratogawhat does USBPOWER_BTN_IGNORE do exactly?
01:07:46Jonas29is it preview to appear its a future version?
01:09:53saratogaoh the sansa USBPOWER_BTN_IGNORE is power, but holding power will just turn off the player
01:09:56saratogawhat is the idea here?
01:10:04saratogaJonas29: not possible
01:11:54 Join dys [0] (
01:12:26Jonas29saratoga: why is it not possible?
01:12:37saratogahardware limitations
01:13:45Jonas29on the ipod touch, there is a hardware limitations
01:15:23saratogawe don't run on the ipod touch
01:16:14saratogaUSBPOWER_BTN_IGNORE was added by amiconn in r7674 specifically for the archos players, it seems people continued to update the macros as new players were added even though they're often nonsensical
01:17:03Jonas29ok it is probably eventually provided
01:17:40saratogaif it is, it'll probably be as an application so you would just use the apple web browser
01:19:13saratogagevaerts: logs say you added the inverted USB connect logic for the H10, do you remember why thats needed?
01:28:51 Join markun [0] (
01:28:51 Quit markun (Changing host)
01:28:51 Join markun [0] (~markun@rockbox/developer/markun)
01:34:21Lloreansaratoga: What button is it on the H10. I seem to remember when that happened and it making sense, but I can't for the life of me remember why so I may just be confused
01:37:18 Quit ReimuHakurei (Ping timeout: 264 seconds)
01:40:31 Join Keripo [0] (
01:48:46saratogabah we use rockbox USB anyway, so that code never runs
01:48:52saratogagonna throw up a patch
01:51:05 Join fdinel [0] (
01:55:50 Quit Llorean (Read error: Connection reset by peer)
02:07:30saratogaFS #11769 - Change USB charge mode detection key
02:07:48saratogai'd appreciate if amiconn or someone else who wrote the code originally for archos could comment on it
02:10:48 Join ReimuHakurei [0] (~reimu@
02:18:55saratogahow hard is it to add a yes/no menu option to a plugin?
02:20:15JdGordon|not at all
02:20:17 Quit BHSPitMonkey (Ping timeout: 265 seconds)
02:21:17 Join Llorean [0] (
02:21:38 Quit Llorean (Client Quit)
02:22:47saratogaJdGordon|: right now, test_codec calls "do_menu" and then checks the result to figure out what to do, what would I need to have an option to set boosting to yes/no without returning from teh do_menu call ?
02:23:00saratogabasically i want to be able to choose any of the options on the menu with or without boosting
02:24:54JdGordon|probably not very much... hang on
02:27:29JdGordon|just add an item to the MENUITEM_STRINGLIST() and check for it, then goto show_menu;
02:30:50saratogaJdGordon|: I have to do a "rb->set_option" if I want the yes/no menu right?
02:31:19JdGordon|sounds right
02:31:32JdGordon|umm, maybe not
02:32:16JdGordon|you could just do "toggle boost" and then a splash saying "Boost on/off"
02:33:40saratogayes/no is probably better though, since eventually i want to have othe roptions
02:34:09saratogain the opt_items structs, what does the second column mean? its usually -1
02:34:10JdGordon|I cant remember if set_option only works with the global_setting settings or not
02:34:13JdGordon|... probably doesnt
02:34:18JdGordon|lang string
02:35:23JdGordon|umm, I meant it probably *does* work how you want it
02:36:40saratogayeah i figured
02:36:47saratogawell compiling now, will bug you in probably 1 minute
02:37:40saratogafinally getting around to fixing up test_codec a little
02:40:30saratogaJdGordon|: ok it draws the menu, but it doesn't actually update teh variable
02:40:52saratogai just did "rb->set_option("Boosting", &boost, INT, boost_settings, 2, NULL);"
02:40:56saratogawhere boost is the variable
02:41:02saratogait apparently stays set to zero
02:41:24JdGordon|did it pop up the set opiton gui?
02:41:52saratogaand it even defaulted to the right value
02:41:58saratogabut changing it didn't seem to do anything
02:42:36JdGordon|I havnt looked at set_option() is AGES, check how other plugins use it
02:44:27saratogai'm using it the same as fireworks.c, so maybe i'm doing something else wrong
02:44:45 Join enginerd [0] (
02:46:21enginerdAnybody have experience with rockbox on a Toshiba Gigabeat S?
02:46:46saratogaa little, just playing with it now
02:47:45enginerdI had 3.6 installed previously with very few issues and I tried to update to 3.7 and am having trouble
02:48:08enginerdwhen I plug the device into the computer it enters the "error" screen
02:48:46enginerdI was able to get it to load but as soon as I reconnected to add music it started acting up
02:49:01saratogadoes rockbox actually work on it?
02:49:17enginerdit did at first but I can't get it to load now
02:49:23saratogaprobably a screwed up file system
02:49:32enginerdand suggestions
02:50:07saratogaJdGordon|: any idea why test_codec doesn't update teh screen if the backlight is off?
02:50:27saratogait means that if the backlight goes off while the plugin is running, then it finishes, theres no way to get the output
02:50:35saratogai'tll just show the last thing blitted to the screen while it was on forever
02:50:57JdGordon|battery saving, on some targets no updates happen with the backlight off
02:51:52saratogahow do i tell the screen to update the framebuffer if the screen is off?
02:52:01 Join Llorean [0] (
02:52:03JdGordon|I don't know if you can
02:52:21 Quit Llorean (Changing host)
02:52:21 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
02:54:15***Saving seen data "./dancer.seen"
02:54:29saratogaJdGordon|: I think the problem is i'm not telling set_option what to put in the output correctly
02:54:43saratogai assume it just puts the index of the option, but maybe thats not the case?
02:54:55JdGordon|thats what it should be doing
02:55:24JdGordon|whats that last NULL param?
02:56:00saratogai'm not sure, but other things seem to do it
02:58:21saratogacould you take another look at what I have?
02:58:27saratogaboost seems to be always zero
02:58:39saratogait must be something stupid
03:00:10JdGordon|it looks correct :/
03:02:21saratogathe Yes/No are shown in reverse order on the screen
03:02:33saratogaso Yes maps to 0 and No to 1 I bet
03:04:03 Quit froggyman (Quit: Bye)
03:04:20JdGordon|which target?
03:05:28JdGordon|clip+ shouldnt be swapping things
03:05:35saratogayeah I don't get it
03:05:41saratogathe struct has NO, then YES
03:05:46saratogabut on device its YES then NO
03:06:09saratogai'm going to make the menu print the values it returns
03:06:26 Quit Boldfilter (Quit: Boldfilter)
03:06:47enginerdsaratoga: I used the gbs_update_1_2_us.exe to restore the device (which worked as it should). When I run the beastpatcher.exe and it should reboot to "bootloader USB mode" it enters the error number 2
03:06:53enginerdany ideas?
03:07:01saratogano, but check the wiki
03:07:24saratogatheres a lot of documentation there
03:08:55saratogaoh fucking hell how does log_text not recognize \n as new line
03:10:00saratogadoes the beast not have boosting?
03:14:39JdGordon|I think yes it doesnt
03:15:27 Quit DerPapst (Quit: Leaving.)
03:16:02 Join eWill [0] (
03:16:04JdGordon|that wouldnt stop set_option() from working though
03:17:53eWillSay I install RB to a Fuze v2, and all is good. Later I upgrade RB, but some core RB boot file gets corrupted, and RB won't boot. Does this mean I will have to CRACK OPEN my player in order to shot the two pins that allow recovery mode?
03:18:10eWill*shot = short
03:18:57mc2739eWill: no, you can still hold the left button to boot the OF
03:19:13eWillthanks a lot :)
03:20:40saratogaJdGordon|: i think its working now
03:20:46saratoganot really sure what i was doing wrong
03:20:58 Quit enginerd (Quit: Leaving)
03:22:59CIA-7New commit by saratoga (r28637): Add option to run test_codec unboosted on players with boosting.
03:24:50 Join froggyman [0] (~seth@unaffiliated/froggyman)
03:25:01CIA-7r28637 build result: All green
03:28:03eWillJust ordered a re certified Fuze v2 4gb. I'm all excited.
03:28:14CIA-7New commit by saratoga (r28638): Fix mistake on targets without frequency scaling.
03:30:10CIA-7r28638 build result: All green
03:42:24 Quit eWill (Quit: ChatZilla 0.9.86 [Firefox 3.6.12/20101026210630])
03:43:11 Join Xerion_ [0] (
03:45:03 Quit Xerion (Ping timeout: 240 seconds)
03:45:04 Nick Xerion_ is now known as Xerion (
03:45:23 Quit Rob2223 (Read error: Connection reset by peer)
03:45:58 Join Rob2222 [0] (
03:51:15CIA-7New commit by saratoga (r28639): Force alignment of various data structures in libmad. Speeds up Gigabeat S decoding by about 1MHz.
03:52:56CIA-7r28639 build result: All green
03:57:52saratogakugel: i can't get your ruby script to parse the MHz field on my test_codec output
03:57:59saratogai just get 0MHz
04:01:56 Join dys` [0] (
04:03:59saratogawith my mp3 optimizations, mp3 is now faster then vorbis on arm11
04:04:03 Quit bluebrother (Disconnected by services)
04:04:05 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
04:06:49 Quit dys (Ping timeout: 276 seconds)
04:14:13 Join Barahir [0] (
04:14:29 Quit Barahir_ (Read error: Operation timed out)
04:18:42 Quit Keripo (Read error: Connection reset by peer)
04:19:20 Join Keripo [0] (
04:35:45 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
04:45:43 Quit panni_ (Quit: ( :: NoNameScript 3.81 :: ))
04:46:33 Quit Jonas29 (Quit: ChatZilla 0.9.86 [Firefox 3.6.12/20101027154941])
04:47:58 Quit TheSeven (Ping timeout: 240 seconds)
04:53:12 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
04:54:17***Saving seen data "./dancer.seen"
04:54:43 Quit anewuser ()
04:55:37 Quit pixelma (Disconnected by services)
04:55:39 Quit amiconn (Disconnected by services)
04:55:39 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:55:39 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:55:41 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:55:59 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
05:01:53 Join dys`` [0] (
05:03:41 Quit Keripo (Quit: Leaving.)
05:04:01 Quit dys` (Ping timeout: 276 seconds)
05:06:57 Quit amiconn (Quit: No Ping reply in 64 seconds.)
05:07:06 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
05:07:26 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
05:10:37 Join Llorean1 [0] (
05:10:38 Quit Llorean (Read error: Connection reset by peer)
05:13:16 Nick Llorean1 is now known as Llorean (
05:13:21 Quit Llorean (Changing host)
05:13:21 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
05:14:52 Quit saratoga (Quit: Page closed)
05:21:16 Join ReimuHakurei_ [0] (~reimu@
05:25:16 Quit ReimuHakurei (Ping timeout: 255 seconds)
05:34:20 Quit ps-auxw (Ping timeout: 260 seconds)
05:44:05 Quit Horscht (Quit: Verlassend)
05:45:40 Join ps-auxw [0] (
05:46:23 Quit factor (Quit: Leaving)
05:46:35 Quit MethoS- (Remote host closed the connection)
06:02:19 Join KiwiCam [0] (
06:05:25 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
06:12:21 Quit InsDel (Read error: Connection reset by peer)
06:29:02 Join factor [0] (
06:31:01 Part toffe82
06:39:40 Quit KiwiCam (Ping timeout: 260 seconds)
06:50:29 Join Llorean1 [0] (
06:50:43 Quit Llorean (Disconnected by services)
06:50:45 Nick Llorean1 is now known as Llorean (
06:50:47 Quit Llorean (Changing host)
06:50:47 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
06:54:19***Saving seen data "./dancer.seen"
07:11:11 Join xxcv [0] (
07:17:38 Quit JesusFreak316 (Ping timeout: 245 seconds)
07:28:20 Quit JdGordon| (Quit: leaving)
07:32:36 Join Buschel [0] (
07:34:49 Join KiwiCam [0] (
07:35:39 Join JdGord [0] (~jd@
07:36:04 Join Topy [0] (
07:39:43 Quit T44 (Ping timeout: 245 seconds)
08:04:35CIA-7New commit by saratoga (r28640): Align various libwma buffers. Saves about 1 MHz on the Gigabeat S.
08:06:06Buschelsaratoga: stil awake?
08:06:40CIA-7r28640 build result: All green
08:08:23Buschelsaratoga: can you give this a try on your beast? ->
08:10:49 Join saratoga [0] (9803c6dd@gateway/web/freenode/ip.
08:12:13saratogaBuschel: 35.93 with that patch
08:12:50Buschelsaratoga: thanks, and without?
08:13:21Buschelthere is still no atrac3 sample in our test file compilation
08:13:29saratogayes I made one tonight
08:13:42saratogajust have to find a CDR and I can make one from the official test file
08:13:50saratoga35.84 without that patch
08:13:50Buschelperfect :)
08:13:56saratogaso basically no change
08:14:46saratogabut it looks worth having so I say commit
08:14:52saratogamight help on android
08:14:54amiconnsaratoga: Seen ?
08:15:07 Quit KiwiCam (Ping timeout: 252 seconds)
08:15:09saratogaamiconn, no thanks
08:15:43saratogadid you see the patch i put up before about USB charge mode?
08:16:47saratogaBuschel: alignment mostly helps if you do ldm or load doubles to aligned buffers, so if aligning something doesn't help theres probably no multiple loads going on
08:17:01saratogathat might mean theres a chance to further optimize accessing that buffer
08:18:09Buschelyes, seems like the access to qmf_window[] is the main optimization
08:20:21Buschelneed to go to work now, see you later
08:20:23 Quit Buschel (Quit: ChatZilla 0.9.86 [Firefox 3.6.12/20101026210630])
08:22:06saratogaBuschel: weird that outSamples didn't make a difference, its accessed 4 words at a time
08:22:15saratogamaybe it was aligned by chance
08:28:14amiconnsaratoga: The USB power button should always be chosen so that the action its long-press will trigger is not confusing or dangerous
08:28:36Lloreanamiconn: If it's "all buttons" then assuming any button has a safe action, it'll always be available to the user
08:28:46LloreanSince "safe" can vary from screen to screen, or even based on the opinion of the user.
08:29:07LloreanI consider context menu generally safer than quickscreen, for example
08:29:10LloreanBut not everyone agrees.
08:29:11amiconnThe commit message for 7674 says why USBPOWER_BTN_IGNORE was added.
08:30:41amiconnThe original ones (MODE for ondios, F1 for recorder fm/v2) are the menu buttons on their respective targets
08:31:00*Llorean doesn't see how "all buttons" introduces a new danger.
08:31:17LloreanIs the theory that users will assume that not a single button on the player will do anything other than allow USB mode when held?
08:31:29amiconnIt does - for the very same reason why USBPOWER_BTN_IGNORE was introduced
08:31:52LloreanCould you explain that simply?
08:31:55amiconnThat would cause devices with flashed rockbox to *always* enter usb power mode on startup
08:32:36LloreanSo boot and then insert the cable?
08:32:40amiconnI.e. the button which is used to power up the device *must* be ignored
08:33:01amiconnWhat if you can't boot without power from USB?
08:33:02LloreanOr put in a special condition that the BTN_IGNORE button is ignore until boot "completes" safely
08:33:24 Quit BHSPitMonkey (Remote host closed the connection)
08:33:28amiconnCompletion is too quick for this
08:33:31LloreanIf you can't boot without power from USB, isn't going into USB charging mode acceptable? You could then replug if you'd like UMS much like you have to if you're charging anyway and decide you want UMS access.
08:33:47LloreanThe boot is too fast to give the user time to let go of the button?
08:34:14amiconnThe Ondio can be used from USB power with no batteries inserted. If any button enters USB power mode, you would always end up there, and be unable to enter mass storage mode
08:34:48LloreanSo the boot is too fast for a user to be able to let go of the button before it completes?
08:35:17 Join kish_ [0] (~o2@unaffiliated/spice)
08:35:19amiconnYes, because you need to hold it for a short while, until you can be sure that it boots
08:35:34amiconnOtherwise USBPOWER_BTN_IGNORE wouldn't have been introduced
08:36:04LloreanThings are introduced that aren't strictly speaking necessary, so I wasn't willing to make that assumption
08:36:45LloreanEven then, why not just ignore all "power" buttons? Since it's nonsense to hold down the button that shuts down the player to try to insert USB. If you're even a little slow you don't get anything
08:37:27amiconnYes, that's what is necessary.
08:37:27 Join Guest34717 [0] (
08:37:27 Quit Guest34717 (Changing host)
08:37:27 Join Guest34717 [0] (~bjst@rockbox/developer/Zagor)
08:37:39 Nick Guest34717 is now known as Zagor (~bjst@rockbox/developer/Zagor)
08:37:55LloreanOr allow a button on the player to disconnect USB (if HID is not enabled) and a menu entry to enter MSC if charging from USB (these might be nice anyway)
08:38:26LloreanI'd really like a button to drop out of MSC, honestly, for my car where trying to hold down things is impractical, or where the player is waked by USB power and then enters MSC and replugging the cable is potentially unsafe while driving.
08:38:59 Join esperegu [0] (~quassel@
08:39:18amiconnYes, that might be useful. It also seems that some users would prefern an option what to do on usb insert instead of that button-hold mechanism
08:39:56LloreanMy ideal would be an option to decide if the default is connect or charge, every button (or nearly every button) causing it to perform the other action, and a button that gets me out of MSC if I'm accidentally in it without having to replug.
08:40:32LloreanI think a menu would slow things down, and you'd still want it to timeout to one or the other (and which it would timeout to being probably user selected) anyway
08:40:34amiconnSomething that resemples "Stop" would be the proper button for this I think
08:40:38LloreanProbably, yes.
08:41:12amiconnI'm not talking about a menu at insert (which might be impossible in the current screen)
08:41:58LloreanAh, well some people want a menu at insert. Not me, though.
08:42:21LloreanI want a menu where I can decide what insert will do by default, then if I hold a button it does the opposite of what I've chosen.
08:42:47LloreanSince I'd much prefer the default on my in-car player to be charge, while the default on my Beast to be MSC
08:44:43amiconn-> an option
08:44:58*amiconn wonders why binsize on the Player increased so much lately
08:45:08amiconnrasher: Are you still running the binsize graphs?
08:45:22 Quit xxcv (Ping timeout: 245 seconds)
08:46:44 Quit JdGord (Quit: Bye)
08:48:47 Join xavieran [0] (
08:48:48rasheramiconn: I am. It'll often be days between that computer is turned on though
08:51:34amiconnPlayer has hit the 200KB loader limit, causing it to use the self-extractor now
08:54:23***Saving seen data "./dancer.seen"
08:57:14 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
09:03:45 Join ender` [0] (
09:04:16 Join DerPapst [0] (
09:24:39 Join Rob2223 [0] (
09:25:45 Join petur [0] (d408b802@rockbox/developer/petur)
09:25:49 Quit S00row (Read error: Connection reset by peer)
09:26:46 Join S00row [0] (
09:28:27 Quit Rob2222 (Ping timeout: 265 seconds)
09:31:07 Join kramer3d [0] (
09:31:07 Quit kramer3d (Changing host)
09:31:07 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
09:31:08 Quit S00row (Read error: Connection reset by peer)
09:31:44 Join S00row [0] (
09:32:27 Join utanapischti [0] (
09:34:42 Join LinusN [0] (
09:34:42 Quit LinusN (Changing host)
09:34:42 Join LinusN [0] (~linus@rockbox/developer/LinusN)
09:36:53 Quit sasquatch (Ping timeout: 276 seconds)
09:38:53 Quit liar (Read error: No route to host)
09:39:08 Quit user890104 (Ping timeout: 272 seconds)
09:40:34 Join liar [0] (
09:49:18 Join swilde [0] (
09:52:22 Join user890104 [0] (
09:57:59 Quit DerPapst (Quit: Leaving.)
09:59:06 Join JdGord [0] (~jd@
10:10:14 Quit xavieran (Ping timeout: 265 seconds)
10:15:32 Quit S00row (Read error: Connection reset by peer)
10:17:08 Join S00row [0] (
10:17:57 Quit factor (Read error: Connection reset by peer)
10:22:07gevaertssaratoga: that code isn't originally mine (although I may have accidentally removed it at some point and then readded it), but the reason for it is that (at least on MTP H10s) you have to hold the right button to get to the ROM USB mode, so in practice you have to hold the right button already before the reboot
10:25:14gevaertsProbably irrelevant now that we don't reboot for USB any more of course
10:30:56 Nick dys`` is now known as dys (
10:30:59 Part LinusN
10:32:22 Join u42p [0] (
10:33:18u42p(how) can i make rockbox start when i plug my sansa clip+ into usb? currently the original firmware loads. i am using
10:33:31 Join factor [0] (
10:35:20 Join DerPapst [0] (
10:45:32 Join LinusN [0] (~linus@rockbox/developer/LinusN)
10:47:56 Quit GeekShadow (Ping timeout: 255 seconds)
10:48:10TheSevenu42p: you'll need to use a bootloader that knows about the fact that rockbox can handle USB, but I have no idea how to do that on a sansa or if there are precompiled ones available
10:51:06 Join xxcv [0] (
10:52:49u42pah, i was guessing correctly yseterday then. well, gonna wait until that is officially done then
10:54:25***Saving seen data "./dancer.seen"
10:58:24gevaertsu42p: such a bootloader will be released as soon as rockbox handles USB itself
10:58:57 Quit kramer3d (Ping timeout: 272 seconds)
11:08:18 Join einhirn [0] (
11:28:17 Join casainho [0] (
11:30:55 Quit xxcv ()
11:50:58 Join xxcv [0] (
11:52:15 Quit xxcv (Client Quit)
11:56:05 Join Llorean1 [0] (
11:57:38 Quit Llorean (Ping timeout: 240 seconds)
11:57:45 Join Llorean [0] (
11:57:47 Quit Llorean (Changing host)
11:57:47 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
11:59:41 Join n1s [0] (
11:59:41 Quit n1s (Changing host)
11:59:41 Join n1s [0] (~n1s@rockbox/developer/n1s)
12:01:12 Quit Llorean1 (Ping timeout: 276 seconds)
12:05:38 Quit Dreamxtreme (Read error: Connection reset by peer)
12:06:19 Join Keripo [0] (
12:08:28 Quit Keripo (Client Quit)
12:15:39gevaertsAlexP: did you follow commits to trunk to check for stable-suitability?
12:15:56gevaertsI'm thinking it may be time for a point release
12:17:40*TheSeven has a question about "const"
12:19:02TheSevenif i have a register that needs to be set to an address, and the data at that address won't be affected by the address being written to that register, how do i declare that?
12:19:42TheSeven#define DMAC0C0SRCADDR (*((volatile void**)(0x38200100))) will complain that the const qualifier (on the argument passed to the function that writes the address to that register) is discarded
12:20:03TheSeven#define DMAC0C0SRCADDR (*((volatile const void**)(0x38200100))) seems to tell the compiler that the register would be read-only
12:20:16TheSevenis there even a way to do that?
12:24:20TheSevencome on, there must have been similar cases before...
12:24:36 Join Dreamxtreme [0] (~Dre@
12:24:39TheSevenor didn't we bother about register types at all and just casted everything to an int when writing to a reg?
12:25:16n1sTheSeven: i'm not sure i understand what you mean but i think you want volatile void* const* if i remember the syntax correctly
12:26:21n1sie pointer to a constant pointer to volatile data
12:27:57Torneisn't it, in fact, void * volatile const *
12:28:01Torneor something similarly insane
12:28:09Tornesince it's the pointer which is volatile, not the target
12:28:27n1svolatile const is nice :)
12:29:24Tornepersonally i vote for casting it to int
12:29:43Torneand having the registers all be the same type
12:31:34TheSevenwhat i actually want is a pointer to a volatile pointer to constant data
12:32:16n1sconst void* volatile* foo
12:32:31*n1s recommends cdecl
12:33:01n1sbut basically just write it backwards
12:34:08TheSevenwhat does *((volatile const void**)whatever) actually do then?
12:34:28TheSevenor even &(*((volatile const void**)(whatever)))
12:35:47n1svolatile const void** is "pointer to pointer to volatile const void" so you cast whatever to that and dereference the first pointer so you get a "pointer to volatile const"
12:36:03 Quit JdGord (Quit: Bye)
12:36:24TheSeven(*((volatile const void**)(0x38200100))) = &(*((volatile uint32_t*)(0x38300040)));
12:36:26TheSevenfunny, eh? :D
12:37:03n1s&(*(...)) seems pointless
12:40:54Stummibut not pointerless :)
12:41:19Stummisry for lame joke :p
12:42:52 Quit Gnea (Ping timeout: 245 seconds)
12:42:54TheSevenn1s: it isn't if the * is part of a define, but the & is outside
12:43:07TheSeventhis is basically DMAC0C0DESTADDR = &LCDWDATA;
12:44:01TheSevenif "cast x into pointer to volatile pointer to const void" is what i want, that would be "(const void * volatile *)x"
12:45:20TheSevennow look at
12:45:41TheSevendoesn't this mean that REG32_PTR_T should actually be "uint32_t volatile*"
12:46:15Torneno, because the relative order of a (normal) typename and const/volatile is meaningless
12:46:24Torne"volatile int" and "int volatile" are the same type
12:46:43TheSevenok, but as soon as this is void* instead of uint32_t, it isn't?
12:46:46Torneit gets complicated if the typename is actually a typedef for a composite
12:46:57Tornethis is just one of those things in C
12:47:31Torneit accepts both, and people tend to write volatile int, but as soon as the type isn't a primitive it may matter which you write
12:49:38Torneit would be much more sensible if the language required qualifiers to follow types always, because then the read-right-to-left rule would mostly be sufficient :)
12:49:45Tornetypedefs can still fuck it up if you try hard enough
12:51:49 Join Jerom [0] (~jerome@
12:52:41TheSevenyeah, those "volatile int"s were what made me think they would be generally written left-to-right
12:53:41Torneyup, but it's tough because people have been writing volatile int for too long to change their habits ;)
12:53:59Tornethe cases like this are too rare
12:54:29***Saving seen data "./dancer.seen"
12:55:02TheSevennow why does (*((void* volatile*)(0x38200104))) = &(*((uint32_t volatile*)(0x38300040))); say "assignment discards qualifiers from pointer target type"?
12:55:48TheSeventhe same happens for void* test = &(*((uint32_t volatile*)(0x38300040)));
12:56:21TheSevendoes it think the DMA controller might not know about its volatility?
12:56:29Tornebecause dereferencing it and then taking its address again eliminates the volatile
12:56:34Tornei think
12:56:47TheSevenany way to suppress that? (it's intentional)
12:57:00n1smore casting :)
12:57:04Torneno. in fact, i think that expressoin might even be required to have side effects
12:57:23Torne*((uint32_t volatile*)(address)) is supposed to issue a read
12:57:26Tornebecause of the volatile
12:57:30TheSeveni still want writes to *((uint32_t volatile*)(0x38300040)) (which is used as a define in there) to be volatile
12:57:40TorneI don't think you can do it
12:58:05n1susing an intermediate var to store the adress would work, no?
12:58:16Torneonly with *another* cast
12:58:47TheSevenis there some way to get to the address inside that define without triggering that warning or issuing a read to it?
12:58:51 Quit user890104 ()
12:58:54Torneanyway, i think you'll find the type of your RHS there is uint32_t *
12:58:57TheSeven(even if it means some more casts :) )
12:59:09Torneand the type of the LHS is void * volatile
12:59:16Torneso it is indeed discarding qualifiers
12:59:39TheSevenwhat's bad about (void* volatile) = (uint32_t*)?
13:00:06TheSevendoesn't this just mean that the LHS must be written to immediately?
13:00:16TheSeventhe type cast to void shouldn't matter IIUC
13:00:16*Torne repeats his earlier suggestion that you cast the address to an int
13:00:48TheSevenlet's say we have #define LCDWDATA (*((uint32_t volatile*)(0x38300040)))
13:01:09TheSevenint whatever = (int)&LCDWDATA; won't help, right?
13:01:21Tornewhy not>
13:01:27TheSevenso there's nothing I can do outside of that define to get at its address?
13:01:47TheSevenIIUC the "&LCDWDATA" part is sufficient to trigger the warning
13:01:58Tornenot on its own
13:02:06Torneit's an assignment warning
13:02:12*TheSeven has misunderstood something then :)
13:02:15Tornean expression can't generate that without an assignment in it
13:02:39TheSevenso basically DMAC0C0DESTADDR = (void*)((int)LCDWDATA); might work?
13:03:04Torne(not without an &)
13:03:07Torne(and who knows)
13:03:19*Torne suggests just defining the LHS register as volatile int *
13:03:30Torneand abandoning any pretence of type safety
13:03:52Torneaddresses are uint32_t. it doesn't need to actually be a pointer :)
13:04:10amiconnTheSeven: Why care about register types? So far we didn't
13:04:43TheSevenbecause I personally don't like casting addresses to integers all the time for DMA
13:04:55TheSevenwell, pointers :)
13:05:16TheSeventhe code would be full of REG = (uint32_t)thing;
13:05:17Torneactual type safety is impossible in C anyway, just give up :)
13:06:04amiconnWhy? The cast is implicit anyway
13:06:22TheSevenREG = thing; triggers a warning
13:07:19amiconnLook at mcf5249.h, pp5020.h etc - registers are defined as (*(volatile unsigned long*)(0xADDRESS)) (or short resp. char) all over
13:11:04 Join T44 [0] (
13:11:29 Quit Topy (Read error: Connection reset by peer)
13:14:12 Join Gnea [0] (~gnea@2610:130:112:400:225:d3ff:fe72:4512)
13:14:19 Quit Gnea (Changing host)
13:14:19 Join Gnea [0] (~gnea@unaffiliated/gnea)
13:15:53 Join xxcv [0] (
13:17:00Torneamiconn: you can't implicitly cast from void * to uint32_t
13:21:56 Quit Jerom (Quit: Leaving.)
13:22:19 Join dfkt [0] (dfkt@unaffiliated/dfkt)
13:26:48amiconnTorne: Well you can, but it will throw a warning
13:27:32amiconnI checked some places which I know use addresses in conjunction with registers. Casts are used where necessary, but there aren't that many places
13:28:21amiconnIt's needed for public variables which need to have the correct type (e.g. lcd_framebuffer); for private variables some drivers just use an int32 even if it holds an address -> no cast necessary
13:34:59Tornewell, yah, i count "with a warning" as "can't" :)
14:08:02 Quit xxcv ()
14:18:30n1shmm, seems like the coldfire tremor slowdown with gcc 4.4 is mainly in the fft stuff, the profiling threw me off at first since gcc 3.4 basically disables inlining when profiling while 4.4 does not
14:20:03n1smaybe asming fft4/8/16 will help
14:32:34 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
14:34:53 Nick benedikt93 is now known as benedikt93|AFK (~benedikt9@unaffiliated/benedikt93)
14:38:56 Join kkit|sh [0] (
14:39:19 Quit krazykit (Quit: awe yeeeeeee)
14:42:11 Join saratoga_ [0] (42391dc9@gateway/web/freenode/ip.
14:43:46saratoga_n1s: ive hoped someone would look at fft8 on cf for a while
14:44:09saratoga_it accounts for a large fraction of the entire fft
14:44:23saratoga_and its very short
14:44:35n1ssaratoga_: yes, i'm taking a look now, starting with fft4 now, gcc4.4 produces significantly worse code for this than 3.4 :/
14:45:01saratoga_yeah it screwed up on arm too
14:45:11 Quit Zambezi (Remote host closed the connection)
14:45:46saratoga_iirc fft4 becomes a few loads adds and stores in fft8 and gcc has no idea how to do it
14:47:56saratoga_is there any advantage to sequential loads to iram on cf,
14:48:30n1snot much, should be slightly faster
14:49:24saratoga_i want to special case the 2048/512 fft case for targets with lots of iram or only dram
14:49:44saratoga_rearrange the trig values to be sequential
14:50:19saratoga_probably help a lot on arm11, no idea if its useful on cf targets with 128k iram
14:50:42saratoga_would waste probably 8 kb of ram
14:52:25 Join evilnick_B [0] (0c140464@rockbox/staff/evilnick)
14:53:02n1sspecial lookup tables you mean?
14:53:28 Join kevku [0] (~kevku@2001:7d0:0:f000::135d)
14:53:58 Quit Gnea (Ping timeout: 240 seconds)
14:54:31***Saving seen data "./dancer.seen"
14:55:35n1shmm, in fact load multiple should save cpu cycles when done from iram too, according to the cpu user manual
14:55:59n1smovem takes 1+n cycles while regular loads take 2 cycles
14:56:18saratoga_same as arm7 then
14:57:03saratoga_the current mdct is extremely clever about trig values thanks to stripwx, savesa ton of memory
14:57:18saratoga_but the access patterns are a bit weird
14:57:18n1ssaratoga: and even if not using movem, sequential loads usually mean autoinc adressing modes can be used which is faster than constant offsetts
15:00:02saratoga_most codecs only use the 2048 block size 95 percent of the time, so we could halve the size of the trig tables, the add a separate table for the common case of the 2048 transform
15:00:20saratoga_then sort it for sequential acess
15:01:50n1sthat would probably help cf a bit if it could be in iram, but to make full use of sequential accesses we need to write asm...
15:04:18saratoga_well yeah but thats not my problem :)
15:09:03 Join InsDel [0] (
15:12:48n1smaybe we can let the codecs optionally set up the tables (copying the one(s) it wants to use to iram)
15:13:35saratoga_probably better to just make it a compile time option
15:14:11saratoga_there not much use in varying it per codec
15:14:51n1sif the codec or file uses a different block size
15:15:08n1s(tremor does this with its dewindowing tables already)
15:15:15saratoga_right now theres one table used for all sizes
15:15:37saratoga_its really quite clever
15:16:18n1syes but if we have to chose between having the special case table or the generic table in iram, one compile time choice may not fit all codecs and files
15:16:49 Quit JdGordon (Ping timeout: 245 seconds)
15:18:03saratoga_are there codecs where we couldnt spare 8kb on cf?
15:18:51n1sseveral i think, at least for the targets with 96kB iram
15:19:14saratoga_i dislike the idea of having state in the fft
15:19:22 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
15:19:26n1syes, the whole iram thing is a mess
15:19:41saratoga_since it would make it ackward to have a plugin and codec use the fft at once
15:20:16n1sthey would have their own instance, no?
15:20:27 Join komputes [0] (~komputes@ubuntu/member/komputes)
15:20:30n1scodeclib is linked statically to each codec
15:20:45 Quit benedikt93|AFK (Ping timeout: 276 seconds)
15:24:23 Quit saratoga_ (Ping timeout: 265 seconds)
15:40:30n1sthe fft seems to have been the cause of the slowdown indeed, a brain dead/naive fft4 asm implementation saved 0.3MHz
15:40:58n1slet's see it it can be made a little faster
15:54:50 Join kugel [0] (~kugel@
15:54:52 Quit kugel (Changing host)
15:54:52 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:59:41 Quit InsDel (Changing host)
15:59:42 Join InsDel [0] (
15:59:50kugelamiconn: i'd use intptr_t, IIUC it even enables implicit casts
16:00:59 Quit CaptainKewl (Ping timeout: 252 seconds)
16:01:03 Join timccc [0] (~timccc@
16:06:39 Quit kugel (Read error: Connection reset by peer)
16:06:51 Join MethoS- [0] (~clemens@
16:08:01 Join Qbe [0] (
16:08:14 Quit MethoS- (Remote host closed the connection)
16:08:50 Quit Kitr88 ()
16:10:08 Quit tchan (Quit: WeeChat 0.3.3-dev)
16:10:41QbeHi all, appears to be unreachable this morning−−for me, anyway. Is anyone else having this problem?
16:11:39ZagorQbe: indeed. use for now.
16:12:20QbeThanks very much. It's time to get something better on my e260.
16:13:51 Join MethoS- [0] (~clemens@
16:13:59 Join tchan [0] (~tchan@lunar-linux/developer/tchan) worked like a charm. Thanks again for the help.
16:18:21Zagoryou're welcome
16:18:23 Join Kitar|st [0] (
16:18:34 Quit Qbe (Quit: CGI:IRC 0.5.9 (2006/06/06))
16:18:48 Quit evilnick_B (Quit: Page closed)
16:19:53 Join ReimuHakurei__ [0] (~reimu@
16:19:53 Quit ReimuHakurei_ (Read error: Connection reset by peer)
16:27:09 Join evilnick_B [0] (0c140464@rockbox/staff/evilnick)
16:38:26 Join toffe82 [0] (
16:39:26saratogai've seen a lot of people complain about building rockbox on cygwin
16:39:46saratogais our cygwin environment broken or does it just not work with current rockbox svn?
16:41:20pixelmalast I tried it worked (except −−non-eabi on arm targets)
16:42:32 Join JesusFreak316 [0] (
16:43:07 Join Zambezi [0] (Zulu@
16:47:14 Quit benedikt93 (Quit: Bye ;))
16:48:05 Quit JesusFreak316 (Ping timeout: 245 seconds)
16:48:46saratogai got 2 extra hours with the Fuzev1 and the lowered max clock
16:48:48saratoganot bad
16:49:52Zagorsaratoga: nice
16:50:07 Join tchan1 [0] (
16:50:11saratogaprobably would get more with the lowered max voltage too
16:50:45n1sasm fft4 closed the gap between gcc 4.4 and 3.4 by 0.5MHz as it makes no speed diff on the old gcc...
16:50:58n1sso i'll probably take a stab at fft8 later
16:51:17 Quit tchan1 (Read error: Connection reset by peer)
16:52:07 Join tchan1 [0] (
16:52:15 Quit tchan (Ping timeout: 250 seconds)
16:54:33***Saving seen data "./dancer.seen"
16:54:58 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
16:55:57 Part Zagor
17:00:19n1shmm, this is more complicated than i thought, fft8 loads some of the values fft4 stores again, maybe that can be exploited...
17:03:35 Part LinusN
17:12:03saratogaZagor, Bagder: could one of you upload this file to the test_codecs folder on the download server:
17:12:27saratogaby the way if anyone else was thinking of trying to find the sony software needed to create atrac files, i suggest that they not do so
17:12:31saratoganothing is worth that
17:19:48 Join Topy [0] (
17:20:35 Quit T44 (Ping timeout: 245 seconds)
17:26:37 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
17:27:20 Quit tchan (Read error: Connection reset by peer)
17:28:09 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
17:28:43 Quit tchan1 (Ping timeout: 245 seconds)
17:31:25 Join JesusFreak316 [0] (
17:35:20 Quit tchan (Read error: Connection reset by peer)
17:36:10 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
17:38:26 Join Topy44 [0] (
17:40:35 Quit Topy (Ping timeout: 245 seconds)
17:41:16Topy44hm... it would be really cool to have the "compressor" available on recording
17:41:51Topy44for recording interviews and stuff like that, where the volume is somewhat changing and unpredictable
17:43:13 Quit JesusFreak316 (Remote host closed the connection)
17:43:39 Join JesusFreak316 [0] (
17:44:39 Quit petur (Quit: Page closed)
17:45:19gevaertsDoesn't one of the gain options help with that?
17:53:24evilnick_BAuto-Gain Control
17:54:14pixelmait's not available on all recording targets
17:55:03pixelmaI even believe it's *only* there on the H100/300s, not sure though
17:57:19n1spixelma: yes, only iriver h100/300
18:01:45pixelmasaratoga: FWIW, I just successfully built r28640 for all 3 of my targets under cygwin (OndioFM, M5, c200v1)
18:05:57 Nick jfc^3 is now known as jfc^2 (
18:12:14AlexPgevaerts: yes, I've been trying to keep an eye out
18:12:18AlexPone mo
18:14:59 Part u42p ("Leaving")
18:21:57 Join bertrik [0] (
18:21:57 Quit bertrik (Changing host)
18:21:57 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
18:24:44 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
18:37:31 Quit DerPapst (Quit: Leaving.)
18:40:23 Join eWill [0] (
18:40:55eWillHas any DAP producer ever officially acknowledged Rockbox?
18:41:23n1sdon't think so, why?
18:41:46gevaertsWhat does "officially acknowledged" actually mean?
18:42:39eWillJust curious. I was going through some Fuze forums, and keep seeping people requesting things on the OF, that RB already does.
18:42:52eWillby "Officially acknowledged"
18:43:19eWillI mean like symbolic head-nod or something.
18:43:37*gevaerts has the same question about that :)
18:43:59eWillor suggested to some customer that they should install RB.
18:46:01bertrikWe got some datasheets and a development model of some player I think
18:46:44gevaertsI don't think we got datasheets
18:47:58eWill"development model" −− does that mean the company sent RB a free player so you could start working on it? I'd love to know the company −− sounds like someone worth supporting.
18:48:18 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
18:48:34gevaertseWill: does sending someone $50 worth of gifts really mean *that* much?
18:48:42n1sthe cost of a player is rarely the roadblock preventing a port
18:50:03eWillI guess maybe not, it just sounded like "they want me to have what I want", but after thought, I guess maybe it could be an effort simply to make more money...?
18:50:37eWill"Rockbox works on our player! −− buy it!"
18:50:47n1sdatasheets would be very welcome as they would make porting much easier and much less painful
18:51:24alexbobPeWill: still, I'd readily support a company that embraced rockbox on their product
18:51:26eWillBut they are under contract not to disclose the datasheets −− right?
18:51:28alexbobPand actually provided useful information
18:51:47eWillNot just being difficult −− right?
18:52:32n1sthe player manuafacturers are, sometimes, but the chip manufacturers are usually able to give out datasheets but they don't want to
18:52:36bertrikI'm pretty sure we got datasheets from AMS
18:52:48n1syes, ams is an exception
18:53:30n1ssome datasheets are already freely available like for the hardware in the beast
18:53:49eWillIf a chip manufacturer published their datasheets, would that make it easier for competitors to steal ideas?
18:54:08saratogaprobably not
18:54:36***Saving seen data "./dancer.seen"
18:55:54 Quit JesusFreak316 (Ping timeout: 252 seconds)
19:00:04n1seWill: please stop that
19:00:29eWilloops sorry i have two tabs open. thought i was in the other one.
19:01:00 Join Strife89 [0] (a80dbf52@gateway/web/freenode/ip.
19:02:03pamaurybertrik: I had a look at the elftosb2 program
19:02:48pamaurysome strings mention Rijndael CBC MAC
19:03:44pamaurysome it could be some sort of SHA-128 but I tried with zero key on some random program and it does not seem to be plain SHA-128
19:08:48bertriktoo bad, nice try anyway
19:09:24pamauryI currently trying to see if the executable section has exactly the same size as the binary. I alrzady noted that elftosb2 strips some part of the elf header
19:09:48 Join KiwiCam [0] (
19:09:58pamaurybut it could also add some hashing at regular positions, which would make it even harder to understand to algorithm :)
19:10:38 Join DerPapst [0] (
19:10:55bertrikso you've already actually added something to a .sb with elftosb?
19:11:13bertrikI tried it with some chumby binaries
19:12:43pamauryI just built a dummy executable, then add it to the command file using load and indeed, it adds the binary
19:13:09pamaurywhen there is no encryption, it just creates a ____ section with the content but with part of elf header stripped.
19:13:11bertrikJust some ideas: make time() always return the same value with some preload magic, stub /dev/urandom
19:14:40pamauryI also decided that it was way to difficult to reverse engineer elftosb2, it's written in C++ and even though there lots of strings out there, it would probably be a nightmare
19:14:45bertrikAnd with "zero-key" encryption the binary is still undecryptable?
19:15:11pamaurybut I still have lots of ideas before giving up
19:15:53bertrikUnless sandisk used some kind of default key, I think it's nearly impossible to make it accept a custom image. The crypto seems solid.
19:16:55*TheSeven guesses that D1671/D1759 might be part of the chip number of the nano3g/nano4g PMU. Any ideas what this thing could be?
19:17:02pamaurywell, sha-128 is quite solid indeed :( But we have to make sure first that they didn't use the default key
19:17:45alexbobPbertrik: sandisk uses firmware signing?
19:17:52alexbobPon what device?
19:18:07bertrikalexbobP, I think so, not completely sure, on the fuze+
19:18:48eWillAre there any open source voice recognition programs (wondering if a RB voice recognition would be possible).
19:19:11 Join Jerom [0] (~jerome@
19:23:43alexbobPeWill: there are, but rockbox voice recognition would probably be way too hard for embedded hardware to handle
19:23:54alexbobPbertrik: ah. well that's ironic then, because the fuze+ isn't *worth* porting to :P
19:23:57alexbobPit's just shit hardware
19:26:49 Join Horscht [0] (
19:26:49 Quit Horscht (Changing host)
19:26:49 Join Horscht [0] (~Horschti@xbmc/user/horscht)
19:31:34amiconnn1s: On coldfire, loading with constant offset, no offset, autoinc and autodec are equally fast
19:31:47n1samiconn: not in my experience
19:31:51amiconnOnly indexed addressing is slower
19:32:15amiconnn1s: They all take two cycles for 32 bit, 3 cycles for 16 bit and 8 bit
19:32:23n1swhat i've seen si that constant offsetts are slower
19:32:41n1s(i'd guess because they make the instructions longer)
19:32:52 Join panni_ [0] (
19:33:15n1ssame with using 32 bit immediate values
19:33:28amiconnHere we're talking 16 bit offsets though
19:34:03amiconnIirc the pipeline is able to fetch 32 bits per cycle, so even back-to-back loads with 16 bit offsets shouldn't starve the pipeline
19:34:39amiconnYou may see aliasing effects caused by the icache though. These are tricky
19:34:54n1sright, not sure if i've tested with move.l but emac paralell loading was measurably faster with sequential access and postinc rather that offsetts when i made a test for Buschel on mpc
19:35:14n1sthose becone 48 bit instrs when using offsetts thogh
19:35:20amiconnWell, emac parallel load is of course faster
19:35:30eWillHow many revisions back are on svn?
19:35:49n1si mean that emac parallell load with postinc is faster than emac paralell load with offsett
19:35:55amiconnAnd instructions using two extension words will starve if issued back-to-back
19:35:59gevaertseWill: back to r1
19:36:21n1samiconn: ok, so that is maybe only true for emac paralell loads then
19:36:47amiconnIt only is if you're also issuing them back-to-back
19:37:24n1s(difference in that mpc filter loop was 0.9MHz with the dct reordered so that the paralell loads were sequential)
19:37:29amiconnActually even then it shouldn't, because emac with parallel load needs two cycles anyway
19:37:43eWillwow cool. I've got to do the "find a good revision −− then go half way between that an the bad one −− then go half way between that and the bad one...." thing. What did you call that again?
19:38:04n1seWill: bisection
19:38:22amiconnThat may be simple aliasing. Reordering functions will change speed, often by several percent, in an unpredictable way. Slower or faster...
19:39:00amiconn(unless the code involved is in iram, of course)
19:40:09n1si know, but that loop is in iram (as it's by far the hottest in the codec AFAIU)
19:41:57amiconnMake sure it actually is... when specifying both 'inline' and ICODE_ATTR, and the parent function is not in iram, the result may be different from what's intended
19:42:23amiconnIt should be obvious, but nevertheless I've seen cases of that in rockbox
19:43:02n1sthe one in musepack is a real asm function so it will never be inlined
19:43:25CIA-7New commit by Buschel (r28641): Fix typo in comment.
19:43:32n1sbtw i wonder why gcc doesn't warn if a function is inlined so it ends up in another section when you explicitley specify one
19:43:49 Join Buschel [0] (
19:44:14amiconnBecause it's gcc? </sarcasm>
19:45:08n1strue that
19:45:27CIA-7r28641 build result: All green
19:45:47n1sspotted another bug in it yesterday that causes our workaround that makes profiling work with gcc3.4 to break in 4.4 :)
19:46:12n1salthough the workaround isn't needed in 4.4 so no big problem
19:47:08 Join krabador [0] (
19:48:41 Quit GeekShadow (Read error: Connection reset by peer)
19:49:04 Join GeekShad0w [0] (
19:49:16 Join piggz [0] (~piggz@
19:49:50piggzreagrding yesterdays discussion on sleep/shutdown and startup times...the quoted 3 seconds isnt right...more like 27-30 seconds! :)
19:49:50 Join JesusFreak316 [0] (
19:51:25gevaertspiggz: that sounds long
19:51:34n1spiggz: what target, which version of rockbox, what settings and did you actually time it?
19:54:11piggzipodvideo 5g
19:54:34piggzim not aware of any special settings i have
19:55:20 Quit MethoS- (Remote host closed the connection)
19:56:53 Quit Strife89 (Quit: Let's see what happens when I run from the hard drive.)
19:58:45gevaertsI know I once measured 7 seconds from pressing the power button to having playing audio on my gigabeat F
19:59:06gevaertsflash targets are a lot faster still
19:59:56gevaertsIIRC there are some unexplained slowdowns with that disk
20:01:17piggzprobably why apple invented sleep mode ;)
20:02:46pixelmado you have things like dircache enabled or "load to RAM" or "auto-update" with the database?
20:03:13 Join Strife89 [0] (a80dbf52@gateway/web/freenode/ip.
20:05:11piggzpixelma: maybe, what should i disable
20:06:37gevaertsTry enabling dircache
20:07:25piggzi will when i find it :)
20:07:47pixelmadepends on your way of using it. I just remember reports about slowdown when all three is enabled (not sure if this is still the case). Some of these could also help decreasing boot time when enabled (as gevaerts mentioned)
20:08:04gevaertspixelma: that one *should* be fixed
20:08:14pixelmapiggz: we have a manual which probably helps you finding these options... ;)
20:08:17piggzit already is enabled
20:08:27n1spiggz: is it a recent version of rockbox?
20:08:45 Join phanboy4 [0] (
20:09:39piggzmanuals are for wimps :)
20:10:17phanboy4Is anyone here with a Fuze v1/v2 and Rockbox 3.7 that's not seeing this?:
20:11:55n1sphanboy4: fuzev1 here, i can insert several albums with no problem, more recent build than 3.7 tho
20:12:26phanboy4I'll try todays build
20:12:32 Join freddyb [0] (
20:13:06freddybDoes we have a datasheet for the FuzeV2?
20:15:43krabadorhi RB devs, are some of you planned to improve Fuzev1 battery duration?
20:16:36phanboy4n1s, Aha, seems to have fixed it, works now
20:17:20bertrikkrabador, yes I think there are some ideas for that
20:18:06 Quit casainho (Remote host closed the connection)
20:22:27pixelmapiggz: no, manuals are for us to point people to it so we don't have to answer the same questions again and again...
20:22:56krabadorbertrik, great, i've fear that developers are thinking to discontinue fuzev1 rockbox development
20:23:54gevaertskrabador: so what you're saying is that someone should decide to get a great idea on how to improve fuzev1 battery life tomorrow at 17:45 or something like that?
20:24:38gevaertsphanboy4: any chance of finding out which commit fixed it?
20:24:56gevaertsIf it's broken in 3.7, we might want to fix that in 3.7.1
20:26:50krabadorgevaerts, certainly not :) i'm only asking if fuzev1 developer are continuing
20:27:25gevaertskrabador: well, you keep asking this every few days, so I thought you had something specific in mind
20:29:30 Join aevin [0] (eivindsy@unaffiliated/aevin)
20:30:04 Quit freddyb (Ping timeout: 264 seconds)
20:30:10 Quit KiwiCam (Ping timeout: 245 seconds)
20:31:07 Join stoffel [0] (~quassel@
20:31:38krabadorgevaerts, rockbox totally captured my interest in portable music, and by the day i rockboxed my fuzev1 , i'm only really interested with deveopment on it. I really would contribute, but i'm not a developer
20:32:15gevaertsSo what you're asking is that people drop everything and concentrate on the fuzev1 at the expense of all other targets?
20:32:38krabadorgevaerts, certainly not
20:33:38krabadormany new ports and ideas i look, are really great
20:33:55 Join Kupop [0] (
20:34:24phanboy4gevaerts, I'll look
20:35:00 Quit Strife89 (Quit: Heading home.)
20:36:31eWillis the Fuze V1 RB batt life really that much different than the OF batt life?
20:37:15krabadoreWill, yes, RB battery life are almost the half of OF
20:42:15 Quit chattr (Ping timeout: 245 seconds)
20:44:16saratogafreddyb: no datasheet, but its pretty similar to the as3525
20:44:25saratogawith some stuff from the as3543
20:46:06 Join chattr [0] (
20:49:39 Join Nerdy3_14159265 [0] (
20:50:37Nerdy3_14159265Hey guys, I'm having trouble with Rockbox on my Sansa Fuze v2
20:51:44pixelmayou need to be a bit more descriptive ;)
20:53:04eWillI have to do some bisecting. Can I just make a copy of my RB development dir, and do it all in there? I'm trying, but "svn status" result is −−−− svn: warning: '.' is not a working copy
20:53:16Nerdy3_14159265I was just waiting to see someone respond to continue
20:53:56Nerdy3_14159265I was outside one day so I cranked the brightness on the Fuze, ever since when it starts it shows the rockbox logo then goes to a black screen and won't display anything
20:54:02alexbobP13:25:02 < gevaerts> krabador: so what you're saying is that someone should decide to get a great idea on how to improve fuzev1 battery life tomorrow at 17:45 or something like that?
20:54:06alexbobPgevaerts: you know what? I like that idea
20:54:09alexbobPgevaerts: someone should go ahead and do that
20:54:19gevaertsalexbobP: don't let me stop you
20:54:39***Saving seen data "./dancer.seen"
20:54:52alexbobPgevaerts: meh, my fix is to carry a laptop and charge my mp3 player with it
20:55:06alexbobPcharging the player's whole battery is a drop in the bucket to a sleeping laptop
20:55:06pixelmaNerdy3_14159265: is it possible you set the brightness to the lowest setting (wrapped around the list)?
20:55:31pixelmaNerdy3_14159265: you could try resetting your settings or have a look at it from the PC
20:55:54alexbobPgevaerts: also, on Thursday someone should go ahead and make rockbox give the clip+ a color screen, because I miss the days when sandisk made color mp3 players
20:56:31alexbobPkrabador: you should become a developer!
20:56:46Nerdy3_14159265pixelma: It shouldn't be that it looped around because I played on it for a bit before the screen started just going to black, also I have no idea how to reset the settings without being able to see the screen and I have no cord atm to hook it to my pc
20:59:06pixelmasome players let you reset settings during boot by holding a specific button. I'm not sure if this is implemented on the Fuze and which button it would be if it is, you could search around in our manual
21:00:36 Quit Jerom (Quit: Leaving.)
21:01:50Nerdy3_14159265pixelma: I'm pretty sure that it's the brightness at max, I think I managed to navigate to brightness but changing it didn't affecct anything, I was using the rockboxed fuze simulator
21:02:00Nerdy3_14159265pixelma: as a reference
21:02:01 Join freddyb [0] (
21:02:25 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
21:02:31pixelmathat's the only idea I could contribute as I don't have a Fuze I don't know what else could be going on. If you stick around for a bit maybe someone else could help you
21:02:42 Quit freddyb (Remote host closed the connection)
21:02:49Nerdy3_14159265pixelma: BTW where would that reset button be listed in the manual
21:03:20 Quit jae (Ping timeout: 276 seconds)
21:05:56 Join freddyb [0] (
21:06:54pixelmahmm, I need to look it up myself
21:13:06pixelmait's not in the Fuze (v2) manual where it would be (in "Advanced Settings > Managing Settings", a not under the description of the "Reset Settings" menu item). So it's either not implemented yet or an oversight in the manual source
21:15:17 Quit krabador (Read error: Connection reset by peer)
21:15:34pixelmaI meant a note
21:17:16freddybSaratoga, thanks.
21:28:38 Join edboyer93 [0] (
21:29:05Nerdy3_14159265pixelma: Well I think this max brightness must be a glitch then freaking out the display. I'll have to attempt resetting when I get home
21:32:04 Quit Nerdy3_14159265 (Quit: ChatZilla 0.9.86 [Firefox 3.6.12/20101026210630])
21:43:31bertrikI think I've read of other people having problems with the fuze v2 on max brightness
21:50:40*pixelma is reminded of the c200v1's contrast settings :\\
21:53:14eWillIs there a longer commit list than ?
21:53:55bertrikpixelma, what's wrong with it?
21:54:19bertrikI only know the bug that the screen goes off under some circumstances when changing the brightness
21:54:38 Join krabador [0] (
21:54:58eWillNerdy3_14159265: did this happen in rockbox or the OF?
21:55:59pixelmabertrik: it goes from 0 (total black) to 255 (total white) but on mine at least it should actually only be 0 to 127 (127 is total white too and 128 is total black again), It is as if the range is there twice but I remember someone looking into some datasheet and saying 256 steps would be correct...
21:56:02bertrikeWill, you could check out the source and use svn...
21:56:30bertrikpixelma, I just tried that and it behaves the same for me (looks like a wrap at 128)
21:57:10bertrikI'll try to check the datasheet again
21:57:11eWillbertrik: That's what I'm doing, but in between builds, I could be manually searching the list (now that I'm under 100 possible commits)
21:57:15pixelmaeWill: he left
21:57:42pixelmabut I think (hope) he was talking about Rockbox
21:57:46eWillpixelma: I just ordered a Fuze v2. Do you know if he screwed it up in RB mode or in the OF?
21:58:00bertrikpixelma, have you ever heard of people where 0-255 worked as described in the datasheet?
21:58:47pixelmanot that I remember but I doubt anyone using settings coming close or changing it so much (despite testing as I do)
21:59:11 Join mirak [0] (
22:01:48 Quit krabador (Quit: Sto andando via)
22:03:09bertrikhm, the datasheet has some confusing statements about the contrast setting
22:03:33*TheSeven wonders at which point he should start to port rockbox to the ipod classic
22:04:50n1sTheSeven: can you access the storage and display?
22:05:00TheSevendisplay: yes, storage: no
22:05:03TheSevenusb: yes
22:05:10TheSevenNOR flash: yes
22:05:19TheSevenclickwheel: no
22:05:31n1si imagine rockbox will be painfull untill it can read the storage
22:05:44TheSevenramdisk :P
22:06:06TheSevenand rockbox boots fine without a storage driver
22:06:25n1soh, didn't think it would
22:06:37TheSevenit will fall back to the compiled-in versions of everything
22:06:44TheSevenjust make each storage call return -1
22:07:20eWillProbably a really stupid question, but can the simulator simulate the OF on any targets?
22:07:41 Quit freddyb (Quit: ChatZilla 0.9.86 [Firefox 3.6.8/20100723085406])
22:07:51bertrikpixelma, 255 (or even 127) seems like a slightly ridiculous number of steps anyway
22:07:54gevaertseWill: it's not an emulator
22:08:24n1sTheSeven: then i don't see why not start at porting rockbox straight away
22:08:43pixelmabertrik: I agree but maybe fine-grained adjustment doesn't hurt on this display ;)
22:08:46TheSevenit's mainly a question of when it becomes useful
22:09:18 Join jae [0] (
22:09:21n1si'd say whaen you have at least storage and display
22:10:19 Quit GeekShad0w (Ping timeout: 272 seconds)
22:14:49pixelmabertrik: the most annoying fact of this is the wrap from "black" to "white" in the middle of this to me
22:15:08 Quit mystica555_ (Ping timeout: 245 seconds)
22:15:29 Quit Llorean (Quit: Leaving.)
22:16:26 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
22:18:33bertrikpixelma, I'm trying to figure out if the full range is perhaps possible by configuring some other part of the display. There is a note in the ds that says that under some circumstances the range is less. If that's not possible, I'm inclined to just reduce the range to 0-127 to avoid the wrap at 128
22:23:14bertrikranmachan, do you see the same effect w.r.t. contrast on the c200v2?
22:24:14pixelmaI'd be fine with 128
22:24:43 Quit jae (Ping timeout: 245 seconds)
22:24:51 Quit piggz (Remote host closed the connection)
22:25:34 Quit mirak (Quit: Ex-Chat)
22:30:08eWillOK I've found the commit that broke a function for me (perhaps it's a "fix"?). The way it used to work on my e200v1: Location: WPS > Context Menu > Playlist Menu > Playlist > View Current Playlist. Function: Press Play/Pause button ("up" button) returned you to WPS. Since r28139 or r28140 (same author/file) this button now does nothing. Anyone care to take a look? http://svn.rockbo
22:30:15 Join mirak [0] (
22:32:33 Join jae [0] (
22:32:47gevaertseWill: the logs don't say anything about an intended change in behaviour there
22:32:57gevaertsMaybe submit a bug report?
22:33:36 Join MethoS- [0] (~clemens@
22:34:14eWillr28138 works, r28139 won't compile, r28140 broke me. Ok, bug report. My first one.
22:35:08 Quit liar (Quit: Leaving)
22:36:08 Join bmbl [0] (
22:36:08 Quit bmbl (Changing host)
22:36:08 Join bmbl [0] (~bmbl@unaffiliated/bmbl)
22:36:46 Quit chattr (Ping timeout: 240 seconds)
22:37:45Buschelsaratoga: you there?
22:38:00saratogaBuschel: yeah
22:38:20ranmachanbertrik: Yeah, it wraps around, 0-127 alone is enough...
22:38:58Buscheljust came by the fact that r28549 broke atrac3 playback of a testfile
22:39:03Buschelin sim
22:39:33 Join liar [0] (
22:40:06saratogaBuschel: do you have a link to that file?
22:40:10BuschelI am also wondering how atrac3_iqmf_dewindowing() is defined for ARM
22:40:15bertrikranmachan, ok thanks for confirming
22:40:22 Join chattr [0] (
22:40:33 Join GeekShadow [0] (
22:40:33 Quit GeekShadow (Changing host)
22:40:33 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
22:41:42 Quit esperegu (Read error: Connection reset by peer)
22:43:41Buschelsaratoga: atrac3_iqmf_dewindowing_armv5e() is declared, but not used for AMR_ARCH>=5
22:44:16 Quit edboyer93 ()
22:44:26saratogaBuschel: ?
22:44:38saratogaline 137: atrac3_iqmf_dewindowing_armv5e(out, in, win, nIn);
22:44:59merbananlove sound of atrac in the morning
22:45:05merbananthe sound
22:45:13Buschelsaratoga: atrac3_iqmf_dewindowing() uses int16_t *win for ARM_ARCH<4. shouldn't this still be using int32_t?
22:45:16merbanandid you port atrac1 yet ?
22:45:24saratogaBuschel: is that an atrac3 file?
22:45:37saratogathe bitrate is very high, i thought atrac3 only support 132k and lower
22:45:44saratogano atrac1 yet
22:46:10saratoganot much interest since theres almost no PC accessible atrac1 files
22:46:10merbanansaratoga: check with ffmpeg -i to be sure
22:46:17 Quit factor (Read error: Connection reset by peer)
22:46:25merbanansaratoga: but there is
22:46:34 Join anewuser [0] (anewuser@unaffiliated/anewuser)
22:46:36saratogayou have to rip them from a player right?
22:47:39Buschelsaratoga: damn, forgt to sync to r28550. forget about the iqmf functions
22:47:58saratogamy bad for committing the commented out version
22:48:04saratogawas benchmarking and forgot to fix it
22:48:17Buschelwell, it was faster on ARMv5e and aboce :)
22:48:30 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
22:50:04 Join Xerion_ [0] (
22:50:25merbanansaratoga: yes, but that is no excuse for not supporting it
22:50:33saratogawell you're welcome to port it :)
22:50:38saratogai'll optimize it
22:50:47saratogathat Goodbye track plays for a second for me, then crashes
22:50:53saratogaor gives codec failure
22:51:06saratogathe bitrate is 176k which isn't valid for atrac3 afaik
22:51:09saratogaso i don't know what it is
22:51:09gevaertssaratoga: that's somehow appropriate in a twisted way :)
22:51:17merbananso it's not enough for me to have written a ffmpeg decoder for it ?
22:51:27merbananI should write a rockbox version also
22:51:40saratogawrite an atrac3plus decoder and i'll port that
22:51:51Buschelsaratoga: yes. it plays before r28549
22:53:22saratoga"The freely downloadable RealSystem Producer Basic (available for Linux, Mac and PC) can encode ATRAC3 audio at 352, 264, 176, 132, 105, 96, 64, 44, 32 and 20kbps. "
22:53:43saratogai guess thats where this 176k file came from
22:54:01 Quit Xerion (Ping timeout: 272 seconds)
22:54:01 Nick Xerion_ is now known as Xerion (
22:54:40***Saving seen data "./dancer.seen"
22:55:23saratogaBuschel: it breaks on armv4 for me and I didn't change anything for that
22:55:26Buschelsaratoga: interesting. if I revert atrac3.h only, it works again. seems to be connected to the memory layout. some boundary stuff?
22:55:30saratogaare you sure it was my commit that broke it?
22:56:02Buschelmoving back outSamples[] to its former position within ATRAC3Context
22:57:09saratogaBuschel: sounds like a buffler overflow
22:57:23saratogamerbanan: are you working on rockbox?
22:57:49saratogai wonder if this real encoder uses some combination the other encoders do not
22:58:11merbanansaratoga: hahaha, you know we have an atrac3+ decoder in the works
22:58:38saratogamerbanan: awesome
22:58:42saratogai've been curious how that thing works
22:59:02saratogathe atrac decoders have been ... interesting to look at
22:59:21saratogahave you figured out the filterbank for it?
23:01:11merbananshort version
23:01:12saratogaBuschel: making that buffer bigger fixes the crash
23:01:14merbananit's a bitch
23:01:23saratogaso its an overflow bug . . .
23:01:32 Quit komputes (Remote host closed the connection)
23:01:35saratogaoh wow, didn't notice that had been updated
23:01:40saratoganeed to watch ffmpeg more closely
23:02:14Buschelsaratoga: just did the same :o) yes, it fixes the bug
23:02:22merbanansaratoga: yes the real encoder uses higher bitrates then the sony one
23:02:26saratogagoing to try it with just 32 extra bytes
23:02:42Buschelsaratoga: I just doubled it
23:03:04merbanansaratoga: rockbox port is going to be alot of work
23:03:42saratogabut we've got an efficient iQMF now!!!
23:03:54saratoga"The sample-frame size is 2048 samples per channe"
23:04:08saratogaoh i had heard it used 4096, was interested to see how that would work
23:04:09 Join factor [0] (
23:04:27saratoga"gain control for reducing pre-echo artifacts"
23:04:35saratogahaha anything sony can do to avoid having different mdct block sizes
23:05:00saratogathe mpeg people must have patented that in conjunction with iqmf or something
23:08:55saratogaBuschel: just adding 32 bytes "fixes" the crash
23:08:59 Quit Buschel (Ping timeout: 255 seconds)
23:10:14saratogaBuschel: adding 4 bytes fixes the crash
23:10:19saratogaso its an off by one error somewhere
23:10:32eWillAnyone remember the bug I told you guys about 20 minutes ago? (Play/Pause button doesn't return to WPS from Playlist) Would the Flyspray catagory for that be "User Interface"?
23:10:46 Join Buschel [0] (
23:10:54bertrikbah, ubuntu nautilus crashes when unmounting the c200
23:10:55gevaertseWill: that would work I think
23:11:09merbanansaratoga: iQMF isn't used any more
23:11:23merbananIIRC, they use something else
23:11:23JdGordoneWill: pretty sure there is already a bug and a patch for that
23:11:50*eWill is looking again
23:11:52saratoga"Finally the QMF synthesis filter will be applied in order to sum all subbands together and reconstruct the encoded audio signal"
23:11:58saratogasounds like atrac3?
23:12:20saratogai hope its iQMF, i just spent an afternoon writing an optimal one in armv5 assembly
23:12:57 Quit n1s (Quit: Lämnar)
23:13:49saratogamerbanan: do you know if anyone has looked at wma lossless? i'd like to complete our wma support some day
23:16:55Buschelsaratoga: +1 byte fixes the crash, but the audio is broken (as if the bitstream is broken)
23:17:04Buschelsaratoga: same for +4 byte
23:17:20saratogaoh 4 bytes sounded ok to me
23:18:16bertrikpixelma, is there a flyspray entry for the c200 contrast issue?
23:18:38pixelmanot that I know of
23:19:08saratogaffmpeg plays it fine, so its something we broke
23:19:41bertrikI looked a bit more into it, but can't make it work for 0-255.
23:20:17saratogaalthough i suppose they could just get lucky like we did
23:20:26saratogabut i'm so lazy and don't want to figure out how to compile ffmpeg
23:23:25Buschelarray size +4 => +16 bytes seems fine
23:23:47 Quit Llorean (Ping timeout: 272 seconds)
23:23:52 Join Llorean [0] (
23:23:58 Quit Llorean (Changing host)
23:23:58 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
23:24:18pamaurybertrik: I changed my mind, elftosb2 does something strange with elf files. It modifies them, or reorder the sections, I don't really know what but it does something !
23:24:56 Quit mirak (Quit: Ex-Chat)
23:27:01pamauryperhaps it goes from elf to plain binary or put instructions about how to write things in memory
23:28:57eWillJdGordon: You said you think there is already a patch for the Play/Pause button not returning to WPS from "View Current Playlist". I can't find it. Were you here ~45 minutes ago when I was talking about the revision that broke me? (r28139 or 28140) Those were committed on Sept. 22, so the bug/patch you were thinking of −− is it that young?
23:29:56*eWill also searched bugs
23:30:41saratogaBuschel: ffmpeg sounds ok to me even with the moved around buffers, so we probably screwed up something
23:33:00Buschelhmm, will now check where it is screwed
23:39:36 Quit bmbl (Quit: Verlassend)
23:46:18TheSevenurgh! they only allow NOR bootloader images that are either encrypted with a device-dependent key, or digitally signed
23:46:39 Quit evilnick_B (Quit: Page closed)
23:48:02 Quit stoffel (Remote host closed the connection)
23:49:18Buschelsaratoga: it seems like decodeChannelSoundUnit() is sometimes writing beyond outSamples[] in the case of normal stereo mode or mono
23:49:37 Quit kevku (Quit: KVIrc 4.0.2 Insomnia
23:49:55*TheSeven just spun up the HDD of his ipod classic :)
23:50:14saratogaBuschel: do you have those benchmarks of libmad that you pastebin'ed the other day?
23:50:45Buschelsaratoga: the profiling?
23:51:17saratogai can't remember how fast the dct32 was
23:51:53 Quit Kupop (Ping timeout: 255 seconds)
23:54:20 Quit bertrik (Ping timeout: 250 seconds)
23:57:01saratogachanging the MUL macro in the dct32 to use SMMUL instead of SMULL and a shift would speed up the dct32 a good bit on arm11
23:57:37Buschelsaratoga: overwrite happens somewhere in gainCompensateAndOverlap()

Previous day | Next day