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

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

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

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

#rockbox log for 2009-10-12

00:00:56kugelamiconn: the debug code would exit/scroll if an action comes in, therefore (if not all buttons are known) only known buttons could change the menu. without buttons it would be plain "view only", not different from what we have now
00:01:49amiconnSometimes the keymap exists before the target code to read the buttons
00:02:38kugelthat doesn't lead to button presses
00:03:35kugelapps code can only react on actions if the button code is able to read buttons (i.e. able to send the actions in the first place)
00:04:27amiconnYes, but you may have half-working code, and want to monitor ports while mashing random buttons
00:05:10*kugel doesn't see how that makes a difference
00:05:22kugelonly known buttons can ever result in something
00:06:08amiconnSure. Some buttons may already be known, and you want to figure out interdsependencies?
00:06:25kugelon the other hands, targets like the clip would take advantage; we know have a home-brewed scrolling like mechanism in the target debug function
00:06:43kugelthat also requires pressing a button
00:06:57amiconnReally, this discussion is not new
00:07:13amiconnBelieve me - using standard UI mechanisms for ports debug is highly annoying
00:07:41 Join bubsy_ [0] (n=bubsy@
00:07:44 Quit bubsy_ (Read error: 54 (Connection reset by peer))
00:07:46kugelit was just an idea, my patch isn't about that
00:09:37 Quit bmbl ("Bye!")
00:09:38 Nick |D|Strife89 is now known as |H|Strife89 (
00:11:48 Join kyle6513 [0] (n=kyle6513@
00:12:20 Quit kugel ("Ex-Chat")
00:19:41 Join shotofadds [0] (i=rob@rockbox/developer/shotofadds)
00:23:38***Saving seen data "./dancer.seen"
00:23:49 Join shot0fadds [0] (
00:24:19 Quit shot0fadds (Read error: 104 (Connection reset by peer))
00:24:40 Join shot0fadds [0] (
00:26:05 Quit shot0fadds (Client Quit)
00:28:39TheSevenlinuxstb: i think i know what's going in with the wheel
00:29:08linuxstbTheSeven: Nice. Does that mean you know how to fix it?
00:29:21TheSevenpossibly, still need to try around
00:29:52TheSevenin fact, the wheel seems to be working, it's rather the SoC's SPI controller that's going nuts when the wheel is powering up
00:30:15TheSeveni just dumped what's going on there, and am seeing bit-shifted valid clickwheel packets
00:30:30TheSevenso let's see how to reset that beast
00:32:11 Quit shotofadds (Read error: 145 (Connection timed out))
00:32:36TheSevenbah. the datasheet doesn't match
00:34:25linuxstbNothing new then...
00:34:32saratogalinuxstb: for what its worth I don't see much difference between "Unusable" and "early development"
00:34:42saratogaif a target doesn't have playback I'd say its both
00:35:03saratogabut I do like the idea of moving some of our targets around (ipods without charging for instance)
00:35:34linuxstbsaratoga: "Unusable" would mean that Rockbox builds and runs but has problems. "early development" are things that might not even build any more.
00:35:52linuxstbThe idea is basically to spread out the targets in more categories.
00:36:15saratogais it really useful to do that though? either way such targets are developers only, and developers should know that already
00:36:46linuxstbI just don't think that a target like the beast deserves to be in the same category as something like the Elio
00:37:10saratogathe beast is just waiting for someone to clean up the install directions
00:37:49saratogaaside from the fact that its nearly impossible to install without asking on IRC, the beast would be in the middle teer already
00:38:25linuxstbYes, the beast is probably a bad example, as the port itself is so far more advanced than the install procedure/instructions.
00:39:04linuxstbBut what about something like the clip? IIUC, Rockbox runs OK on it, but buggy. That's better than the targets where Rockbox doesn't even build.
00:39:26saratogathe clip really isn't useful to users at this point though given that it crashes so much
00:39:40saratogaimo they shouldn't worry about the clip except if they want to watch its development status
00:40:23saratogabasically I dislike the idea of catagorizing ports that users shouldn't be looking at . . .
00:40:49saratogaboth because I think its misleading and because I think most developers won't bother to properly promote their ports
00:40:54linuxstbI see your point, but I don't think that list has to be just for users though. As a developer, I would like to know the status of all those targets a little more - I haven't a clue about them...
00:41:15saratogawe have the targetstatus wiki page
00:41:31saratogaIMO its worked extremely well
00:41:34linuxstbWhy is it misleading? They're called "unusable". I guess I'm just adding a new category underneath there.
00:41:59saratogai suppose
00:42:18linuxstbBut forgetting that for the moment, you agree with renaming "Unstable" to "Usable", and shuffling some devices around?
00:42:36saratogayes that seems sensible
00:42:45saratogawhat is the top teer called?
00:42:47 Quit ender` (" Washing your car to make it rain doesn't work.")
00:43:11linuxstbIn my email I kept it as "Stable". But we could choose something else - I had considered "Mature"
00:43:56saratogasomething to suggest that usable ports are potientially a lot less friendly to use
00:44:45saratogaactually I rather liked unstable just because it discouraged people who were likely to screw up with the command line installer from even trying
00:45:24linuxstbPorts like that should be in "Unusable".
00:45:42linuxstbSo I guess I may be suggesting that some current "Unstable" ports go down a category.
00:45:47linuxstb(and hence my fourth category)
00:46:25saratogalinuxstb: rockbox on the fuze is quite usuable, you just need to be able to copy and paste and a single command from a web page into a command line window
00:46:31saratogasome people have a lot of trouble with that
00:46:32linuxstbi.e. "Stable" and "Usable" should be more or less what the current "Stable" catgeory is. Then "Unstable" becomes "Unusable" and then early development undernearth.
00:47:05linuxstbsaratoga: As long as it's documented, I don't see a problem with that. But doesn't rbutil handle it now?
00:48:06 Quit robin0800 (Remote closed the connection)
00:48:07saratogait does but lots of middle teer targets may not have rbutil support
00:48:46saratogai recall one person sending me about 5 angry PMs because he couldn't figure out how to copy and paste that one line into a windows command line
00:48:59saratogaand was furious that we expected him to be able to do that
00:49:04linuxstbYes, but I would still call them "usable". Some users will of course struggle with any command-line install procedure though, and we should aim to get things in rbutil (or an easy-to-use command-line util like sansapatcher or ipodpatcher) asap.
00:49:10saratogaclearly he needed to be encouraged to wait for top teer support
00:49:54linuxstbDon't let one person cloud your view of our users - most people can handle it, it's just the vocal idiots that cause problems.
00:50:13kyle6513if i may butt in here, i believe all they need is a simple batch file download, that should fix the problem for copying the line.
00:50:44kyle6513batch files even have input if you really need it
00:50:55saratogathe whole point of the catagory system is to discourage users who are simply too stupid to read what they are installing
00:51:10ShapeShifter499on r23121M the database hasn't been working
00:51:12saratogaif we assume users are intelligent and can become informed then we should just link the device status port and let them decide
00:51:52saratoga"device status page" rather then "port"
00:52:17 Quit pamaury ("exit(*(int *)0 / 0);")
00:53:09linuxstbsaratoga: I disagree. I would say the point is to have a clear central place where the status of ports is clearly shown. If an install procedure is too hard for end-users, then the port shouldn't go into "Usable".
00:53:25 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
00:54:10ShapeShifter499how do I get a svn of a earlier build?
00:54:17saratogawe do have a clear central place for status, its the DeviceStatus wiki page
00:54:55saratogaits color coded and everything
00:55:12ShapeShifter499I want 23088
00:55:13linuxstbsaratoga: Yes, but that's just an expanded version (or at least should be) of what's on the home page.
00:55:49linuxstbsaratoga: Although I've lost track of what we're arguing about - are you now saying you're happy with how the front page is now?
00:55:59saratogabut its really not since on the front page we're not actually saying anything about what a port can do, only about how useful we think it is to some average user
00:56:10saratogai like the idea of moving things around on the front page
00:56:23saratogai just don't see much sense in putting information about targets that aren't useful to developers
00:56:37saratogaoriginally i didn't even put the unusable targets on the front page, but instead a link to the wiki
00:56:38kyle6513it says in giant bold letters PORT STATUS
00:56:41saratogabut someone else changed that
00:56:45kyle6513dont understand how it would be hard to find that
00:57:45saratogai dont really disagree with mentioning that they exist on the front page, as its a good way to interest people in development, but I do think we should say as little as possible beyond "its in progress and you're welcome to help"
00:57:50TheSevenlinuxstb: which interrupt was the lock switch one?
00:58:06saratoga"go look at the device status wiki page and see for yourself"
00:58:15 Join kugel [0] (n=kugel@rockbox/developer/kugel)
00:58:35linuxstbTheSeven: I can't remember...
00:59:35saratogahuh the linux development guide doesn't say to checkout the source code, but goes straight to running
00:59:50linuxstbShapeShifter499: See the UsingSVN wiki page
01:00:45linuxstbsaratoga: Which one? This one does -
01:01:04saratogalinuxstb: ah that one is better, the one on the development guide does not
01:01:10saratogalinked from the front page
01:01:27ShapeShifter499found it
01:03:09kyle6513linuxstb, has there been any progress on the ftl bug?
01:03:09saratogaalthough that simple guide is fucking awful
01:03:42kyle6513saratoga, i followed it and it worked fine for me and i have never used SVN in linux before
01:03:45linuxstbkyle6513: Read the Rockbox commit messages - see the front page of the site, and the "last four weeks" link underneath.
01:04:14kyle6513linuxstb, thanks
01:04:45saratogayes but its a wall of text and pointless smile gifs
01:04:49 Quit tmzt (Read error: 104 (Connection reset by peer))
01:06:23kyle6513saratoga, that is true
01:08:54saratogaits only simple in that it talks down to you, not in that its actually simple to read
01:09:36 Quit GeekShadow ("The cake is a lie !")
01:14:53 Nick |H|Strife89 is now known as |W|Strife89 (
01:17:03saratogai wonder why running seems to use only a single processor for compiling
01:21:12 Quit DerPapst ("Leaving.")
01:22:06 Nick fxb is now known as fxb__ (n=felixbru@
01:23:25ShapeShifter499I 2g2
01:24:19 Part ShapeShifter499
01:27:09TheSevenlinuxstb: IIRC you said that the hold switch is connected to an IRQ. However I can't find it?
01:27:35linuxstbTheSeven: I sort of remember saying that... Let me grep my own logs...
01:27:50TheSeveni already greped mine and didn't find it
01:28:06TheSeveni can only find you saying which GPIO it is
01:28:20linuxstb"interrupt 4 seems to fire when the hold switch is turned on"
01:28:48linuxstbJuly 18th, 14:10 in #linux4nano-dev
01:32:13 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
01:33:06 Join Lynx [0] (
01:33:33 Nick Lynx is now known as Guest19873 (
01:33:45TheSevenah, you spelled it out...
01:34:09TheSevenwith int 4 you mean 0x10?
01:34:18linuxstbI would imagine so
01:37:07TheSeventhat bit doesn't seem to ever turn up in SRCPND
01:37:33linuxstbThis was from loader.htm - so maybe the OF enabled it somehow.
01:39:49 Join Capt_Blackwood1 [0] (
01:40:38 Quit MethoS- (Remote closed the connection)
01:41:15 Join Eragon55 [0] (
01:42:37 Join intrados1 [0] (
01:43:22 Quit intrados (Read error: 110 (Connection timed out))
01:43:58TheSevendiagmode doesn't reinit the clickwheel either - it has the same bug as rockbox
01:44:07TheSevenso no use poking around in there
01:45:48Capt_Blackwood1how's the work on the iPod 6th Gen Classic going?
01:46:14kyle6513didnt he already ask that in #linux4nano?
01:46:25TheSevenhe did.
01:46:33*linuxstb considers it answered
01:46:58kugelsaratoga: you need to hack -j in, or use MAKE_FLAGS (or whatever than env var was called)
01:47:14Capt_Blackwood1i was askin about the "Rockbox" i know about "iPodLinux"
01:47:39linuxstbCapt_Blackwood1: The answer is the same.
01:47:51TheSeven...and linux4nano isn't ipodlinux either
01:47:58 Part Capt_Blackwood1
01:47:59 Quit Lynx_ (Read error: 110 (Connection timed out))
01:47:59 Nick Guest19873 is now known as Lynx_ (
01:54:38 Quit Thundercloud (Remote closed the connection)
01:57:05rashersaratoga: yeah, set MAKEFLAGS="-jX" before running
01:57:23 Join webguest56 [0] (
01:57:41rasherBetter yet, set a reasonable default value in your login script
01:57:54saratogahow would i do that?
01:58:04TheSevenlinuxstb: I have a working, but very ugly, fix for that wheel issue now
01:59:06rashersaratoga: add export MAKEFLAGS="-j 4" to ~/.bashrc (or your shell's login script)
01:59:25linuxstbTheSeven: Do you have a patch?
01:59:28 Join JackWinter3 [0] (
02:01:02TheSeveni really need a better trigger for that
02:01:39 Quit Eragon55 ("ChatZilla 0.9.85 [Firefox 3.0.14/2009082707]")
02:02:40 Quit Rob2223 (Read error: 110 (Connection timed out))
02:03:25 Quit JdGordon ("Leaving.")
02:04:17linuxstbTheSeven: Isn't there already code for that in that file?
02:04:39linuxstb(search for calls to opto_i2c_init)
02:04:54TheSevenno, that's something different
02:05:02TheSevenand it isn't called for nano2g
02:05:48TheSevenoh, you mean the one at the end of the file
02:06:38linuxstbMaybe those could be functions (static inline?) defined in the target-specific parts of the file - e.g. disable_wheel and enable_wheel
02:07:26TheSevenfor now I'll just add an else to that ifdef
02:08:19 Nick Naked is now known as Hadaka (
02:11:35*TheSeven just tried powering down the clickwheel SPI controller while hold is on, but then it breaks again
02:12:05CIA-85New commit by 03theseven (r23122): Fix the iPod Nano 2G clickwheel when releasing the hold switch
02:12:08 Quit JackWinter2 (Read error: 110 (Connection timed out))
02:14:14*TheSeven wonders why one first needs to call make before doing a make fullzip
02:15:14 Join AndyIL [0] (n=pasha_in@
02:17:48 Join Plouj- [0] (
02:18:03 Quit intrados1 (Read error: 110 (Connection timed out))
02:18:04Plouj-my rockbox 3.4 h300 prints *PANIC* storage: -71
02:18:32TheSevenlinuxstb: this was my last fix for today. I need to go to bed now
02:19:41linuxstbTheSeven: So you can make a zip, even when the build failed.
02:20:50Plouj-good thing it has a reset button
02:22:50linuxstbPlouj-: Were you running an earlier version of Rockbox before 3.4?
02:23:40***Saving seen data "./dancer.seen"
02:23:43 Quit amiconn (Nick collision from services.)
02:23:46 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
02:23:54 Quit pixelma (Nick collision from services.)
02:23:55 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
02:24:06 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
02:24:15 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
02:26:30TheSevenlinuxstb: so what's left? booting the apple firmware from the rockbox bootloader? (did you have a try at this again?) usb? what else?
02:26:43 Quit AndyI (Read error: 110 (Connection timed out))
02:27:08linuxstbTheSeven: So is charging working?
02:27:33linuxstbTheSeven: Useful if you could update the wiki page with progress - I haven't been following all your changes today...
02:27:38TheSevenit's not properly controlled yet, but yes, we can switch between off/100mA/500mA
02:28:07*linuxstb also mentions an accidental revert...
02:28:25TheSevenoh yes, this also needs to be sorted out
02:28:40TheSeven(and merged with the charging code)
02:28:50linuxstbNo, the bootloader isn't booting the Apple firmware yet, but I haven't tried any more this evening. I can give you my current patch if you want to work on that.
02:29:14linuxstbOr I could probably just commit it...
02:32:07 Join tmzt [0] (
02:33:16TheSevenI'll go to bed now anyways, and I'll be busy tomorrow
02:35:24linuxstbOK, goodnight.
02:39:53Plouj-linuxstb: yeah, I updated to 3.4 recently
02:40:31linuxstbPlouj-: And you get that panic every time you try to start Rockbox?
02:41:21kyle6513plouj- do you see the menu of rockbox for a few seconds?
02:45:38 Join Rob2222 [0] (
02:51:59Plouj-linuxstb: no, I got the panic only this one time (after many successfull boots) and can't reproduce it
02:52:21Plouj-linuxstb: I had the device charging over USB while this happened
02:53:04linuxstbPlouj-: OK, so it's working fine now?
02:53:35Plouj-sucks, eh?
02:57:09CIA-85New commit by 03dave (r23123): Nano2G: Initial (non-working) attempt to load an encrypted copy of the original firmware from the "osbk" image in the firmware partition. Code based ...
02:57:15 Quit amiconn (Nick collision from services.)
02:57:17 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
02:57:37 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
02:57:42 Quit pixelma (Read error: 110 (Connection timed out))
02:57:50 Join pixelma [0] (i=quassel@rockbox/staff/pixelma)
03:03:02 Quit Lynx_ (" HydraIRC -> <- The professional IRC Client :D")
03:03:18saratogalinuxstb: can you run test_codec yet on the nano2g?
03:05:09linuxstbYes, it should work.
03:05:51saratogawell if you get bored, we now have all the test_codec files on the download server
03:08:43linuxstbThat collection is looking a bit sparse - e.g. no ALAC, AC3, any of the RM codecs... Maybe mt can encoded some Real test files - he has realproducer I think.
03:09:45saratogalinuxstb: yes that would be nice
03:10:13saratogai'm running it now on the e200v1
03:10:22saratogaeventually i need to fix the AMS players so they can run test_codec
03:10:38saratogai bet the Nano2G will be among the fastest per clock targets
03:10:58saratogaupdated the clip logging thread if anyone wants to try
03:11:20linuxstbDo you have an he-aac encoder?
03:11:53linuxstbI guess "nero" just means lc-aac?
03:14:32 Quit BHSPitLappy (Remote closed the connection)
03:16:02kugelsaratoga: they can*t?
03:16:31saratogakugel: well they can do any file smaller then the audio buffer
03:18:34linuxstbSo how can you fix the AMS players to solve that? Do you mean fix test_codec to support files larger than the buffer?
03:18:42saratogalinuxstb: yes
03:18:46saratogashould be fairly simple
03:19:04saratogajust pause the timer, rebuffer, restart the timer
03:20:20saratogai have an AAC-HE version of the sample track right here, but its 180kbps so i don't know what use it is to benchmark
03:21:42 Quit evilnick (Read error: 60 (Operation timed out))
03:21:55 Join Horschti [0] (n=Horscht2@xbmc/user/horscht)
03:22:34 Join Valued_Customer [0] (
03:22:52 Nick Valued_Customer is now known as Eragon55 (
03:23:22 Join funman [0] (n=fun@rockbox/developer/funman)
03:28:50saratogahopefully AAC/Ogg/WMA will all be improved with a newer mdct real soon
03:29:16saratogaATRAC/AC3/Cook as well but i don't have test samples for them to benchmark
03:32:26 Quit dmb ("Leaving")
03:33:34 Join dmb [0] (n=Dmb@unaffiliated/dmb)
03:37:41linuxstbsaratoga: I'm trying that AAC-HE file now - it seems only just realtime...
03:37:58saratogaat what clock?
03:38:05linuxstb200MHz I think
03:38:24saratogahow fast is MP3?
03:38:45linuxstbI'll try one now. Any preference for bitrate?
03:38:51 Quit Horscht (Read error: 110 (Connection timed out))
03:38:55saratoga192k is what i always use
03:39:17linuxstbOK. The AAC is still running...
03:39:25saratogai need to ask stripwax to profile AAC-He again
03:39:41saratogabut i'm sure its the same problem early ATRAC3 had, slow QMFs
03:40:21linuxstb105.73% realtime - 181.59MHz
03:40:33saratogai'm running it now on my e200
03:40:40saratogabut mp3 will be telling
03:41:02saratogaif everything is working right and you have 0 latency IRAM like PP, I would expect about 38MHz
03:42:08 Nick |W|Strife89 is now known as Strife89 (
03:42:49linuxstbI don't think it is zero latency. But TheSeven should know
03:43:09saratogalinuxstb: 134.72 MHz on PP
03:43:21saratogamakes me think the MP3 results won't be pretty . . .
03:43:32linuxstbYes, mp3 seems slow as well... Maybe we're not actually running at 200MHz...
03:43:52linuxstb50.60Mhz - 379.38% realtime
03:44:15saratogayeah something isn't right
03:44:59linuxstbSeems like it may be running at around 100MHz. IIRC, the CPU should be comparable to the Gigabeat F
03:44:59saratogathats much slower then the Gigabeat, so its not just IRAM
03:45:18saratogathat would mean 25MHz decode, which is too fast
03:46:07 Quit Eragon55 ("ChatZilla 0.9.84 [Firefox 3.5.3/20090824101458]")
03:46:25saratogayou've got the same slow multiplier as PP, but a slightly faster load/store unit, so if everything else is equal you should be very slightly faster
03:46:38saratoga(compared to 1 core MP3 on PP)
03:47:06 Join Eragon55 [0] (
03:47:22kugelsaratoga: though it has only small caches
03:47:37kugelFX make up their lack of iram with a bunch of cache
03:47:51saratogasame total cache as PP
03:47:58saratogaand anyway for mp3 you should be entirely in IRAM
03:50:39 Quit efyx (Remote closed the connection)
04:00:00 Quit webguest56 ("CGI:IRC (Ping timeout)")
04:01:42 Quit kyle6513 (Read error: 60 (Operation timed out))
04:03:32 Join Lss [0] (
04:05:48 Quit TheSeven (Nick collision from services.)
04:06:06 Join The_Seven [0] (n=theseven@rockbox/developer/TheSeven)
04:06:18 Nick The_Seven is now known as TheSeven (n=theseven@rockbox/developer/TheSeven)
04:06:40 Quit kugel (Remote closed the connection)
04:08:15 Join Kajarii [0] (
04:17:00 Quit funman ("leaving")
04:19:14 Quit dmb (Read error: 104 (Connection reset by peer))
04:23:02 Quit Eragon55 ("ChatZilla 0.9.84 [Firefox 3.5.3/20090824101458]")
04:23:41***Saving seen data "./dancer.seen"
04:32:18 Join funman [0] (n=fun@rockbox/developer/funman)
04:36:43 Join intrados1 [0] (
04:36:44 Quit nosa- (Read error: 54 (Connection reset by peer))
04:36:45 Join nosa-j [0] (
04:38:00 Quit Rondom (Nick collision from services.)
04:38:10 Join Rondom [0] (
04:40:04 Quit nosa-j (Read error: 104 (Connection reset by peer))
04:40:07 Join nosa- [0] (
04:41:12 Quit saratoga ("Page closed")
05:06:27 Join ShapeShifter499 [0] (
05:25:41 Part toffe82
05:31:35 Join lifeless_ [0] (n=lifeless@
05:33:56 Quit panni_ (Read error: 104 (Connection reset by peer))
05:45:40 Quit lifeless_ ("Ex-Chat")
05:46:02 Quit rvvs89 (Read error: 54 (Connection reset by peer))
05:48:26 Join lifeless_ [0] (n=lifeless@
05:51:19 Quit Horschti ("Verlassend")
05:51:23 Quit funman ("free(random());")
06:11:45 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
06:14:26 Join fdinel [0] (
06:17:28 Quit fdinel (Client Quit)
06:19:05 Quit gevaerts (Nick collision from services.)
06:19:17 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
06:23:43***Saving seen data "./dancer.seen"
06:24:35 Quit elinenbe (Read error: 113 (No route to host))
06:28:58CIA-85New commit by 03tomers (r23124): LCD scrolling - fix a typo
06:29:48CIA-85New commit by 03tomers (r23125): LCD scrolling - reduce one 'if' nesting level
06:30:30 Join tomers [0] (n=chatzill@
06:51:35 Quit tomers (Read error: 148 (No route to host))
06:57:06 Quit liar|netbook ("Verlassend")
07:04:41 Quit HBK (Read error: 131 (Connection reset by peer))
07:05:48 Join HBK [0] (
07:09:42 Quit ShapeShifter499 (Read error: 60 (Operation timed out))
07:12:07 Join stoffel [0] (
07:15:12 Quit Rob2222 (Read error: 104 (Connection reset by peer))
07:15:53 Join Rob2222 [0] (
07:28:21 Join funman [0] (n=fun@rockbox/developer/funman)
07:30:20 Part Kajarii
07:35:47 Join Riku [0] (
07:48:38 Quit intrados1 (Read error: 110 (Connection timed out))
07:54:00 Quit Lss (Read error: 110 (Connection timed out))
07:58:29 Join locutox [0] (
07:58:47locutoxyou glorious mofos, you've finally got rockbox working on 2g nano:D
08:05:36 Quit Strife89 ("Alvin, it's 2:00 in the morning. Go to bed!")
08:23:44***Saving seen data "./dancer.seen"
08:25:04 Quit stoffel (Remote closed the connection)
08:27:45 Join ender` [0] (
08:32:04 Join Rob2223 [0] (
08:33:40funmanis it possible to browse rockbox menus / use plugins while recording?
08:34:49funmanit looks like i'm not able to do this on my fuze but i'm not sure if it's a limitation of rockbox or of the fuze keymap
08:38:07 Join ruj1 [0] (
08:38:42funmanfor instance, i would like to open the debug menu while recording is on, to check clock frequencies and ascodec registers
08:39:50 Join Zagor [242] (n=bjorn@rockbox/developer/Zagor)
08:42:13funmanZagor: can you do some chmod o+r and ln -s mkamsboot-1.1 mkamsboot please ?
08:44:09Zagorsure. but what is the symlink for?
08:44:36funmanwe can refer to it in the wiki / manual and not have to correct for each new version
08:44:44Zagorgood point
08:45:03funmanor perhaps mkamsboot-latest
08:45:38funmani think sansapatcher/ is a symlink as well ?
08:46:12pixelmafunman: for some reason I think browsing the menus while recording is not possible (I seem to remember that the action to leave the recording screen is the same as the stop recording action which I stumbled upon for the c200 keymap), not a 100% sure though
08:46:21Zagoror actually, no. sansapatcher is a dir and sansapatcher-0.8 is a symlink pointing to it... it's a bit of a mess, this directory :-(
08:47:08 Join petur [50] (n=petur@rockbox/developer/petur)
08:50:08 Quit Rob2222 (Read error: 110 (Connection timed out))
08:55:13funmansometimes I reload the page and see the mkamsboot-1.1RC directory, how are the 2 servers synced ?
08:55:49funmanthe mirror is hosted by videolan, i should ask them
08:57:01topikis there a reason why test_codec is not built by default?
08:57:16funmantopik: it's useless to most users
08:57:30topikmany things are
08:57:53topikit wouldn't hurt to have it, would it?
08:57:59funmanright, but i think the point is people needing to use test_codec know how to build it
08:58:13topikthat's true
09:03:01topikas written on the forum it is more useful to have coders than benchmarkers, but for unstable builds it could be useful/convenient to have test_codec included. then again, compiling your own is not _that_ hard
09:05:05funmanZagor: what is the new rsync url ? doesn't work anymore
09:07:33Zagorfunman: umm, I'm not sure. I haven't changed anything rsync related lately.
09:08:59funmanZagor: hum it works in fact, i'll check why the videolan server isn't up to date
09:13:04 Join flydutch [0] (
09:16:18funmanworks now, perhaps related to permissions of mkamsboot-1.1
09:26:39 Join LinusN [0] (n=linus@gateway/web/cgi-irc/
09:28:17 Join Thundercloud [0] (i=thunderc@
09:29:35 Join JackWinter4 [0] (
09:30:09 Quit lifeless_ (Remote closed the connection)
09:30:40 Join lifeless_ [0] (n=lifeless@
09:36:10 Quit ruj1 ()
09:37:38 Join MethoS- [0] (n=clemens@
09:41:51 Quit JackWinter3 (Read error: 110 (Connection timed out))
09:46:13 Quit Thundercloud (Remote closed the connection)
09:52:09 Join robin0800 [0] (
09:56:29 Join nls_web [0] (
09:57:10 Join DerPapst [0] (
10:02:33funmanI wonder how FM radio can be heard on sansa e200/c200 v1 since all line inputs are muted by audiohw_set_monitor(false) in audio_input_mux()
10:18:30TheSevensaratoga, linuxstb: I don't know about IRAM latencies either, I just know that IRAM also runs through the cache, probably for a reason. And SDRAM is running at something ridiculous like 32MHz.
10:21:56 Quit Riku (Read error: 104 (Connection reset by peer))
10:23:35linuxstbTheSeven: BTW, I think the NAND/FTL still isn't perfect. It seems to shutdown cleanly, but more often than not the NOR bootloader seems to be doing something - i.e. booting is delayed by a few seconds.
10:23:38TheSevenlinuxstb: I just looked at r23123... Do you actually need interrupts in the bootloader?
10:23:48***Saving seen data "./dancer.seen"
10:24:11linuxstbTheSeven: Also, I tried reinstating my USB changes, and get a "panic: could not unmount nand" error every time I inserted USB.
10:25:06TheSevenIf not (which I would hope) you could just leave the bootrom mapped to zero like iloader does, and just change IRAMORIG to 0x22000000
10:26:01*TheSeven can't think of a reason why unmounting panicing on USB connect, but not shutdown, should be coming from the FTL
10:26:05linuxstbTheSeven: I guess it depends on whether any of the drivers used by bootload use interrupts - do they? e.g. if they sleep(), then we need the tick timer (unless we fake the sleep for bootloader builds -other targets do that).
10:26:28TheSevencan't remember any of the interesting drivers doing that
10:26:32TheSevenAFAIK they all just yield
10:26:59 Join antil33t1 [0] (n=Mudkips@
10:27:02 Quit antil33t1 (Remote closed the connection)
10:27:02linuxstbTheSeven: Is the system_reboot() in system-s5l8700.c still correct - I think that's where the panic is coming from.
10:27:09 Join antil33t1 [0] (n=Mudkips@
10:27:10 Quit antil33t ()
10:27:11TheSevennot having the tick timer will just mean infinite timeouts (and thus probably a lockup while scanning nand banks on 2GB ipods)
10:27:15 Join antil33t [0] (n=Mudkips@
10:27:17 Quit antil33t (Nick collision from services.)
10:27:23 Join antil33t [0] (n=Mudkips@
10:27:29TheSevenso faking the tick timer from the usec one like on PP could be an idea
10:27:42linuxstbWe have a usec timer?
10:28:23 Quit antil33t1 (Client Quit)
10:28:32 Join antil33t1 [0] (n=Mudkips@
10:28:38 Quit antil33t (Nick collision from services.)
10:28:43 Join antil33t [0] (n=Mudkips@
10:28:45 Part antil33t1
10:29:07 Join antil33t1 [0] (n=Mudkips@
10:29:13 Quit antil33t (Nick collision from services.)
10:29:41TheSevenlinuxstb: The datasheet doesn't have one, but apple seems to be accessing one
10:29:50TheSevenit's somewhere after the timer D regs
10:29:51 Join antil33t [0] (n=Mudkips@
10:30:03TheSevenstill needs investigation
10:30:15TheSevenand i found the reason for the weird FTL/rebooting behavior
10:30:40TheSevensince we moved that to powermgmt.c errors while shutting down are just ignored and not paniced on any more
10:31:50TheSeveni think we should remove that panic in system_reboot, and add one in nand_flush to catch all cases
10:32:45 Join daurn [0] (i=daurnima@freenode/staff/daurnimator)
10:39:18 Quit nls_web ("CGI:IRC")
10:39:30 Join maruk [0] (
10:41:27 Join daurn| [0] (n=daurnima@freenode/staff/daurnimator)
10:46:21linuxstbTheSeven: Do you mean your fix to the shutdown panics was not to panic any more? Or am I misunderstanding?
10:48:02 Quit daurn (Read error: 60 (Operation timed out))
10:48:43 Join daurn [0] (i=daurnima@freenode/staff/daurnimator)
10:48:59TheSeventhere are 2 types of shutdown panics: one, where the FTL code itself panics, because we're entering untested and potentially destructive code (i.e. remapping - the ones we've been seeing recently), and the other kind is where something just goes wrong, the ftl returns a bad rc, and the caller paniced on that. the latter were only visible on reboots, but not on shutdowns any more, since 23057
10:50:05 Join Contos [0] (
10:50:28Contoshey i had a question, whats on the sansa rockbox?
10:50:40scorche"whats on"?
10:50:41Contoslike games and whatnaught?
10:51:09linuxstbSee the "manual" link on the left of every page of the Rockbox website.
10:51:28Contosokay... rephrase, features
10:51:35linuxstbSee the "manual" link on the left of every page of the Rockbox website.
10:52:03Contoskay thx
10:52:29 Quit Contos (Client Quit)
10:53:13*TheSeven can't reproduce FTL fails, even with re-enabled panics
10:54:53CIA-85New commit by 03theseven (r23126): Bring the iPod Nano 2G shutdown FTL panics back.
10:57:47TheSevenlinuxstb: do you get any shutdown panics with that revision? I don't get any
10:57:57linuxstbLet me try...
10:59:48 Quit daurnimator (Read error: 110 (Connection timed out))
11:00:03TheSevenand my norboot ("apple-logo") time is at the usual 2 seconds, so it seems to be just working
11:00:21TheSevenbut there could be some weird conditions on your FTL that make it choke
11:00:48linuxstbI just turned my ipod on, and the NOR spent about 10 seconds doing something...
11:01:37linuxstbHmm, and now I get a "can't load rockbox.ipod - bad checksum..."
11:01:42linuxstb(from the RB bootloader)
11:01:50linuxstbSeems like my Nano isn't happy still
11:02:55TheSevenis the bootloader using an FTL_READONLY build?
11:03:32TheSevenok, so that one can't be at fault
11:04:34 Join Omlet [0] (
11:04:42 Join Omlet05 [0] (n=omlet05@
11:04:55*linuxstb tries again, and this time does an "eject" as well as unmount.
11:05:17 Quit Omlet05 (Client Quit)
11:05:50linuxstbOK, it booted now. The "eject" didn't return instantly, so maybe I should always be doing that... (although I've never needed in the past)
11:06:55linuxstbTheSeven: OK, it seems happy at the moment. Reboot to USB also works without panic. (I have my usb changes re-applied)
11:07:16linuxstbSo maybe the disk mode only does its own flush on "eject".
11:08:36gevaertslooks like it, but why wouldn't it do that on disconnect?
11:09:11 Join rvvs89 [0] (n=ivo@pdpc/supporter/base/rvvs89)
11:09:13linuxstb"eject" == "disconnect" I think.
11:09:43linuxstbi.e. if I just unmount, then the ipod just says "do not disconnect", but after eject it says "OK to disconnect"
11:10:05linuxstbgevaerts: Ah you mean a physical disconnect?
11:10:17 Quit daurn| (Read error: 110 (Connection timed out))
11:10:19linuxstbYes, you would think it would do that...
11:10:22gevaertsunless it's a host-sode issue
11:11:42linuxstbWhat could that be? I always unmount before unplugging - could that not be flushing things to the disk?
11:12:23gevaertsit should, unless I'm seriously misunderstanding things. I just don't like eliminating possibilities too soon
11:12:52TheSevenlinuxstb: so you just unmounted and reset it?
11:13:02linuxstbI unmounted then unplugged.
11:13:32gevaertslinuxstb: maybe run "vmstat 1" and see if there's bo activity during eject?
11:13:40TheSevenand before you were resetting it from disk mode? or what were you doing?
11:13:41linuxstbAh no, I didn't... When I force it into disk mode (holding select+play), it doesn't automatically leave disk mode, so I would have needed to reset.
11:13:49gevaertsor iotop, or...
11:14:13TheSevenlinuxstb: heh, that's the advantage of not manually booting it to disk mode :-)
11:14:16linuxstbSo yes, perhaps the "reset from disk mode without eject" was upsettig it.
11:14:21TheSevenit automatically shuts down cleanly and reboots
11:14:33linuxstbTheSeven: Yes, that's what happens when Rockbox reboots.
11:14:35TheSevenlinuxstb: that's what i was expecting
11:14:54linuxstb(I only had to enter manually because someone reverted that feature....)
11:15:54TheSevenok, at least you spotted that "no panic on unclean shutdown" bug :-)
11:16:13linuxstbI'm here to help.
11:19:18 Join einhirn [0] (
11:25:12linuxstbTheSeven: Have you implemented backlight brightness on the Nano2G?
11:25:58*linuxstb sees he did - nice...
11:27:19linuxstbgevaerts: Adding HAVE_USB_POWER wasn't enough to stop my Nano2G rebooting on usb when holding menu
11:29:42 Quit daurn (Read error: 110 (Connection timed out))
11:30:01 Join daurn| [0] (n=daurnima@freenode/staff/daurnimator)
11:34:06maruklinuxstb: mine does not reboot when plugging an usb cable (even if I don't hold the menu button).
11:34:22linuxstbmaruk: Yes, that was reverted from svn. I'm testing it again now...
11:34:48maruklinuxstb: ok.
11:41:37TheSevenlinuxstb: you didn't notice the backlight wasn't as bright as usual by default, did you?
11:42:13linuxstbTheSeven: No, although I did just notice that it was brighter in my bootloader...
11:42:18TheSeveni set the default brightness to half the current apple uses!
11:42:54TheSevenin fact we could double what apple uses (from the PMU side) but i doubt the LEDs would survive that for long - so i set the max value to what apple did
11:43:00 Join nls_web [0] (
11:44:26TheSeven(the brightness setting is using a logarithmic scale)
11:50:02linuxstbThe default seems fine to me, and I think I normally like things brighter than average.
11:50:36TheSevenyes, that's what I thought, too
11:52:14marukTheSeven: for me, the default is also fine. I reduce the default by 1 or 2 (can remenber).
11:52:33TheSevenit looks a little like a printout at that brightness. not really luminous, but perfectly clear
11:53:00TheSeventhe default value seems to be bright enough to reduce contrast again
11:53:14TheSeventhe apple value*
11:54:14TheSevenand of course it's a huge power saving... backlight off: 22mA, backlight at the default value: 34mA, backlight at the apple value: 46mA
12:02:24 Join KBH [0] (
12:04:25linuxstbTheSeven: Have you started looking at USB?
12:04:36TheSevenwell yes, *looking* at it
12:16:04TheSevenI might have another go at that tonight
12:16:24TheSevenat least if there is nothing more important to do, but I think we've caught all the other major things now
12:16:41linuxstbYou could look at the bootloader if you want - I'm probably not going to get time today.
12:17:05TheSevenFYI, the USB controller seems to match the one in the S3C6400X datasheet
12:17:28linuxstbAs well as booting the OF, it would be nice to have button detection - currently the dual-boot only happens if the hold switch is on, but it would be better if you could hold a button instead.
12:18:00linuxstbBut I remember that on earlier ipods, I struggled to detect very early button presses.
12:19:15*linuxstb wonders how controversial a "check for .rockbox/bootloader.bmp file and if it exists show that as a boot menu" feature would be.... ;)
12:19:27 Quit HBK (Read error: 110 (Connection timed out))
12:19:31TheSevenyes, the wheel won't send anything if a button is just held
12:19:40TheSevenonly if you press it while monitoring it
12:19:53gevaertslinuxstb: you mean as some sort of illustrated help file?
12:20:51 Quit funman ("free(random());")
12:21:19linuxstbgevaerts: Yes. I quite like iloader - it shows a clickwheel image with various logos in the button locations for the different options - Rockbox, AppleOS, disk mode, iBugger (TheSeven's debugger). You then press the relevant button to start that option.
12:22:03linuxstbBut boot menus have long been a "no-do", so I guess I'm wondering if people have mellowed...
12:22:11TheSevenlinuxstb: not to forget the pear for "whatever you would like to have here" :-)
12:23:02gevaertsI think it would be a good idea, although I wonder if having just a marker file and the bitmap file compiled in wouldn't be easier/less code. Do we want a bmp reader in the bootloader?
12:23:33linuxstbMaybe s/bmp/raw/
12:23:49TheSevenlinuxstb: I would go for the way i did it in iloader
12:23:51linuxstbi.e. just load it directly to the framebuffer. But yes, compiling it in could work, but you know what users will want...
12:23:52***Saving seen data "./dancer.seen"
12:23:57TheSeveni.e. native-lcd BMPs
12:24:27TheSevenyes, users like the easy themability of iloader :-)
12:24:56gevaertsmaybe I'm overestimating the cost of having the bmp reader
12:25:07TheSevendepends on its features
12:25:28linuxstbI don't think speed is an issue - if the user is going to use a boot menu, they can wait a fraction of a second for it to be loaded.
12:25:51TheSeveni think gevaerts was rather thinking of bin/ramsize
12:25:58TheSevenbut i think a trivial BMP reader in the bootloader should be possible
12:26:07TheSeventhe one iloader uses is 15 lines of code
12:26:08linuxstbEmbedding the bmp would be more binsize...
12:26:29TheSeven(but of course it only supports native bitmaps)
12:26:30gevaertsI was mostly thinking about complexity actually
12:26:59gevaertsif we do loadable files, I think it should be normal bmp
12:27:14TheSevengevaerts: what I'm doing is basically get height and width out of the BMP header, set up the LCD, and blit the rest of the BMP directly to the LCD
12:27:38linuxstbTheSeven: That's not very portable... If we do this for the Nano2G, it would make sense to do it on the older ipods.
12:28:00linuxstb(some older ipods have a byte-swapped lcd buffer, I guess that could just be a #ifdef...)
12:28:52TheSevenyes, not very portable, but it was the easiest to implement - i didn't want to waste time on that, but rather get that thing finished
12:29:03linuxstbBut anyway, the main question is whether "we" (as Rockbox) want it...
12:29:15TheSevenmost of the iloader code is pretty crappy, but at least it was done in about a day
12:30:41gevaertslinuxstb: I think it would be fine, except of course that people will always want more
12:31:11linuxstbWhen do we give people what they want? ;)
12:31:19gevaertswhen it suits us :)
12:31:38gevaertsMaybe actually implementing a graphical configurable bootloader will reduce support issues due to people using untested bootloaders?
12:33:17linuxstbYes, although iloader is likely to be well-tested with Rockbox (as TheSeven is using it I assume), it would be nice to not fork users into two groups based on their bootloader.
12:35:00TheSeventhe RB bootloader has the same issues as GRUB legacy, I think ;-)
12:35:07TheSevenTime for a 2.0 :-D
12:35:21linuxstbWell, we don't have a 1.0 yet (for the Nano2G)...
12:35:33 Quit Omlet ("( :: NoNameScript 4.22 :: )")
12:35:43TheSevengrub also never had a 1.0
12:37:29TheSevenlinuxstb: btw, i found another issue in your code: we're probably running into caching trouble again...
12:37:55TheSeveni would suggest using 0x0A000000 or 0x48000000 as the load buffer
12:38:18 Join intrados1 [0] (
12:38:23TheSevenand then of course disable the protection unit before jumping to 0x08000000
12:39:53TheSevenaddress bus wrapping to circumvent caches is the best trick ever I ever spotted :-D
12:41:51linuxstbTheSeven: I thought the NAND driver did that? Ah, but the decryption doesn't...
12:42:17linuxstbSo would just passing 0x48000000 to aes_decrypt() be enough?
12:42:28TheSevenno, i would use that address all the time
12:42:46TheSevenelse things might still be stuck in the write buffer when aes_decrypt tries to access them
12:43:03TheSevenhey, can anybody tell my why this yellow hasn't been there all the time?
12:43:46TheSevenah, someone had removed these buggers already. why didn't svn update catch that?
12:43:47linuxstbTheSeven: I fixed that...
12:44:06TheSevenbut for some reason it neither made its way into my tree nor conflicted...
12:44:49CIA-85New commit by 03theseven (r23127): Fix the yellow again
12:45:09linuxstbTheSeven: Yes, you re-committed them with r23126...
12:46:28*linuxstb likes to do an "svn diff" before committing, just to double-check what he's actually going to commit
12:48:27*TheSeven usually does that, but didn't expect that file to be out of sync before committing that one-line patch
12:52:46 Join daurn [0] (i=daurnima@freenode/staff/daurnimator)
12:55:10TheSevenis there currently a known bug with text scrolling?
12:55:28TheSeveni'm right now seeing a very weird thing
12:57:36TheSevenif text starts scrolling, it will first scroll from the left to the right end correctly.
12:57:52 Quit daurn| (Read error: 110 (Connection timed out))
12:58:28TheSeventhen it will scroll back from the right to tle left end again, but twice as much as expected, i.e. beyond the left end
12:59:05TheSevenand it shows very interesting vertically-wrapped characters beneath the left end of the text
12:59:27TheSeventhen it will jump to the right end again, and again scroll beyond the left end
12:59:37TheSevenany idea what's causing this?
12:59:56TheSevensomething related to RTL or that centered patch yesterday?
13:00:51linuxstbTheSeven: I've just tried disabling IRQs in the bootloader, and it feezes some time after turning the backlight on...
13:01:23TheSevenyou have a 2gb/4gb ipod?
13:02:13TheSevenyes, then it's probably locking up while scanning the NAND banks
13:02:27TheSeven(the timeout for detecting the chips needs the tick timer)
13:03:25linuxstbAlthough it doesn't seem to update the LCD either, so I'm not sure it's getting that far.
13:03:56*TheSeven suggests faking a tick timer
13:04:29linuxstbI'm calling backlight_init() - maybe that's doing too much...
13:05:05TheSevenno, it should only do some i2c stuff
13:05:23TheSeven(i2c timeouts won't work either, but shouldn't happen in the first place)
13:05:38TheSevensee sub_22001208 and it's xrefs for how to access the usec timer or whatever it is
13:06:20 Join kugel [0] (n=kugel@rockbox/developer/kugel)
13:07:11kugelTheSeven: don't forget to handle usb connection in the nand thread
13:07:31TheSevenhuh? how's that related?
13:07:53kugeleach thread must respond on SYS_USB_[DIS]CONNECT
13:08:36kugel(that was regarding what you said earler today)
13:13:00amiconnThe rockbox bootloader doesn't have a boot menu on purpose. It should boot as quick as possible, and be self contained
13:13:16amiconnA boot menu is a big nodo, imo
13:13:44TheSevenamiconn: would you also object splitting/ifdefing it into a fast and a fancy one?
13:13:59TheSevenkugel: I can't see how a thread itself would get such an event
13:14:13TheSevenaren't it rather the queues that get it?
13:14:56kugelnot sure :p
13:15:20TheSevencan't see why threads should be interested in such events in general
13:15:49TheSevenand i can't see how a thread should get an event at all, besides using a queue, which the nand thread doesn't have
13:16:03amiconnIt's also a question of reliability. On some targets a non-working bootloader would mean it's bricked
13:16:48amiconn(not even considering the space constraints on a few targets)
13:17:25kugelTheSeven: could be that only queues get it, maybe I'm confused since every other thread has a queue
13:17:27TheSevenamiconn: that's why ifdefing it would make sense
13:17:54TheSevenon nano2g, it can be up to 16MB in size, and a bad one is very easy to replace
13:18:09 Quit kugel (Nick collision from services.)
13:18:15amiconnMore ifdefs == mor epotential bugs
13:18:17 Join kugel [0] (
13:19:01*TheSeven would pretty much like to get rid of that thread altogether, but can't make the spindown timeout work without it
13:19:07*kugel wouldn't like a boot menu as well
13:19:33kugelTheSeven: you can add a timeout
13:20:14TheSevenhow? i.e. using which api?
13:20:33kugeltimeout_register(power_off_func, data, HZ/5); see kernel.h (HAVE_TIMEOUT_API)
13:22:06kugelthe function will run in a tick task, which may (or may not) cause some problems
13:22:25kugelin a tick interrupt*
13:22:48TheSevenhmm, yielding from an interrupt is probably not the best thing to do
13:23:01kugelyou wouldn't need to yield
13:23:25kugelthe function runs only once, when the timeout wears off
13:23:45TheSevenbut trying to lock a mutex will either lock up entirely or need to yield if it's done in an IRQ
13:24:14TheSeven(in case the mutex is already locked)
13:24:26amiconnyield() must not be called in interrupt context
13:24:32kugelthat's why mutex aren't interrupt safe :)
13:24:55TheSevenand that's why mutexes aren't timeout-safe either
13:25:22amiconnIf you need to call a function at a certain timeout, and that function needs to access mutexes or similar, you pretty much need a thread
13:25:22TheSevenso I'll need to keep that thread around
13:25:43amiconnAfaik all storage driver use a thread for oe thing or another, so I don't see a problem with that
13:26:09LinusN"apps/plugins/crypt_firmware.c: Copyright 2009 TheSeven" - don't we want real names there?
13:26:29kugelI think we do
13:26:43TheSeveni think linuxstb just copy-pasted that
13:27:11TheSeveni'm fine with removing it altogether, or rather replacing it with a rockbox header
13:27:19linuxstbLinusN: Well, it was a verbatim quote from the other source file...
13:27:45linuxstbLinusN: The (C) at the top has TheSeven's real name.
13:28:13linuxstb(but yes, there's nothing explicitly linking the two, although that's implied)
13:30:51linuxstbamiconn: I'm only suggesting an optional boot menu on targets where it's safe, such as ipods.
13:32:31CIA-85New commit by 03theseven (r23128): Ditch additional copyright notice quoted from iBugger.
13:34:59nls_webi don't see the problem with loading an optional bmp in the bootloader but of course i wouldn't use it
13:35:06TheSevenamiconn: it's more or less the question if rockbox will do it and keep it based on the same, known-to-work, drivers and source code, or if a third party will do it with possibly more compatibility issues.
13:35:24*linuxstb wouldn't use it either...
13:36:05TheSevennow that iLoader is around, I don't think people will stop using it, even if i would explicitly drop support for it
13:36:48TheSeven(of course, I'll try to keep it as compatible as possible, but there are already some FTL fixes that haven't been backported to it yet)
13:36:56 Quit antil33t ()
13:37:48kugela boot menu will just get us annoying people asking for it on other targets
13:38:28linuxstbkugel: So? We're not obliged to do what users want...
13:38:46kugelit's annoying
13:38:49linuxstbBut if that's the case, it shows it's a wanted feature.
13:39:01linuxstbkugel: Then just ignore such people...
13:39:15*TheSeven wonders if #if (CONFIG_CODEC == SWCODEC) || !defined(SIMULATOR) for the metronome plugin does make any sense
13:39:53TheSevenshouldn't it rather be ... || defined(SIMULATOR) instead?
13:41:12linuxstbNo, I think that means that it's always compiled for swcodec (real and sim), and only real hwcodec.
13:41:47linuxstb(hwcodec are the Archos targets that have a hardware mp3 decoder which isn't simulated)
13:41:49TheSevenwhy should it work on real hwcodec and not on hwcodec sim (which is probably internally swcodec?)
13:42:30linuxstbThat's the problem - hwcodec sim is not internally swcodec.
13:43:16nls_webyeah, that pp codition is in a few places and always takes me a few seconds untill i get it :)
13:44:20nls_web#if (CONFIG_CODEC == SWCODEC) || (CONFIG_CODEC != SWCODEC && !defined(SIMULATOR)) would be slightly clearer imo
13:44:54linuxstbnls_web: Or a short comment...
13:45:01nls_webor that :)
13:47:36TheSevenare the oscilloscope and vu meters designed to work for line in and/or playback?
13:47:51linuxstbJust for playback I think
13:48:07linuxstb(maybe the function isn't implemented in our pcm driver if they don't work)
13:48:09TheSevenso there's still a missing api... they just show zero
13:48:25TheSevenoscilloscope for recording would be really nice :-)
13:48:25linuxstbYes, search for "peak" in pcm-s5l8700.c
13:48:35 Part LinusN
13:48:45*TheSeven wonders why that's target specific
13:49:12linuxstbI think it's to get access to the data as close as possible to when it enters the DAC
13:49:33TheSevenwell, shouldn't every PCM api just be passing it around?
13:51:01linuxstbYes, perhaps it could be..
13:52:15 Join fdinel [0] (
13:53:57TheSevenah, it's latencies that are tried to avoid there
13:57:14TheSevenok, the scope works now
13:57:47*TheSeven wonders why someone bothered writing a stub for that one that's longer than an actually working one
13:58:38CIA-85New commit by 03theseven (r23129): S5L870x: Implement pcm_play_dma_get_peak_buffer
14:01:21 Join elinenbe [0] (
14:03:36 Quit locutox (Read error: 101 (Network is unreachable))
14:03:39linuxstbTheSeven: BTW, I think we should keep interrupts in the bootloader - in the future I think we would want a USB driver there, which I'm assuming will need to be interrupt-driven?
14:04:19linuxstbWhen IRAM is remapped, I'm assuming it's also still available at 0x22000000 ?
14:04:52linuxstb(plus a button driver if a boot menu is accepted...)
14:06:35TheSevenUSB can be done without interrupts (iBugger Loader is doing it that way), but I'm not sure if that would work with rockbox's USB stack
14:06:57 Join TheSeven|Mobile [0] (n=TheSeven@
14:07:02TheSevenyes, it's still available
14:07:32TheSevenbut the int handlers will be called from 0, which may upset them, depending on how they work
14:07:40TheSevenan app build had issues with this at least
14:08:38 Quit DerPapst ("Leaving.")
14:09:16linuxstbIt should be fine I think - I'll simply specify IRAM as being at 0x22000000 in, but remap IRAM.
14:09:52linuxstbThe vectors themselves are simply "ldr pc, [pc, #20]" - and those memory locations contain the 0x22xxxxxx address.
14:23:55***Saving seen data "./dancer.seen"
14:25:51 Quit TheSeven (Read error: 113 (No route to host))
14:47:04 Join ruj1 [0] (
14:49:51 Quit Zarggg (Read error: 110 (Connection timed out))
14:50:15 Join kyle6513 [0] (n=kyle6513@
14:50:54 Quit robin0800 (Read error: 110 (Connection timed out))
15:00:05 Join TheSeven [0] (n=theseven@rockbox/developer/TheSeven)
15:00:25 Quit kyle6513 ("Leaving")
15:01:23TheSeveni just listened to my nano during a bus ride, and noticed 3 issues:
15:01:25TheSeven- the scrolling problem (does this also happen on other targets?)
15:01:44TheSeven- the "external power" backlight timer is always used, even if it's running on batteries
15:02:31TheSeven- pause when headphones removed doesn't work
15:02:46linuxstbThe third requires you to enable it in the settings - did you do that?
15:03:10 Join teru [0] (
15:03:20TheSeveni did, but it might have been reset by me installing a new build tonight
15:03:23TheSevenlet's have a look
15:04:13linuxstbIf you just unzip over the top of an existing build, then it shouldn't change your settings.
15:04:27linuxstb(they are in config.cfg, which isn't in
15:05:38TheSevenyes, but there was some FS corruption caused by that NAND bug, so it might have it hit
15:05:41TheSevenok, works again
15:05:52TheSevens/it hit/hit it/
15:06:44TheSevencan anybody have a quick check (on any target) if that scrolling issue turns up?
15:08:55*TheSeven wonders where to find the sim current builds
15:10:28marukTheSeven: what scrolling issue ?
15:10:33 Quit antil33t1 (Read error: 104 (Connection reset by peer))
15:10:40 Join antil33t [0] (n=Mudkips@
15:11:37TheSeven[12:56]<TheSeven>if text starts scrolling, it will first scroll from the left to the right end correctly.
15:11:39TheSeven[12:56]<TheSeven>then it will scroll back from the right to tle left end again, but twice as much as expected, i.e. beyond the left end
15:11:40TheSeven[12:57]<TheSeven>and it shows very interesting vertically-wrapped characters beneath the left end of the text
15:11:42TheSeven[12:57]<TheSeven>then it will jump to the right end again, and again scroll beyond the left end
15:11:58 Part Plouj- ("oh well")
15:13:14marukTheSeven: nano2g specific issue ?
15:13:28TheSeventhat's what I'm trying to find out
15:13:54TheSeveni've only seen it since yesterday, and there was some patch committed that I suspect to cause that on all targets
15:14:09marukTheSeven: do you know if r22977 is affected ?
15:14:21TheSeventhat's probably before it
15:14:23linuxstbTheSeven: You could try a UI simulator for a different target (I'm busy with real work atm...)
15:14:26 Join faemir [0] (n=faemir@
15:14:45TheSevenI suspect r23105
15:14:55TheSevenlinuxstb: if you tell me where i find the current builds of them?
15:15:21mc2739TheSeven: I am seeing this scrolling problem on e200v2 with r23123
15:15:50TheSevenok, so that's not a target specific issue at least. fine.
15:15:51linuxstbTheSeven: See (for both win32 downloads and instructions for building the sim yourself)
15:16:13marukTheSeven: no issue on my mini with r22977.
15:16:47teruTheSeven: I think the issue is related to FS #10670.
15:16:50TheSevenlinuxstb: these are way too old to contain the issue and I don't have access to my build machine right now
15:17:00TheSevenaren't there build system builds of them to be downloaded somewhere?
15:17:21linuxstbTheSeven: No
15:17:22kugelTheSeven: yes?
15:17:36linuxstbrasher: Is there something wrong with your uisim builds?
15:18:18TheSevenkugel: could it be possible that your patch caused that?
15:18:27kugelwhich patch?
15:19:01kugelI don't think so, but everything is possible
15:20:25mc2739linuxstb: I think they are broken from the server move - I believe it was around that date
15:20:36TheSevenkugel: nvm, it's in fact FS #10670
15:20:51kugelmy patch just added center drawing, which is probably as broken as RTL in regards to scrolling. but normal drawing shouldn't be affected any more than it is by rtl
15:21:33TheSevenkugel: the issue only turns up when scrolling
15:21:56mc2739Zagor: ping
15:22:10*TheSeven will try as a possible fix when he gets home
15:22:14mc2739Zagor: when was the server move?
15:23:28Zagormc2739: sep 16
15:23:30linuxstbmc2739: 16th September - it's on the news page
15:23:40 Quit lifeless_ (Remote closed the connection)
15:23:53mc2739sorry, should have looked there first
15:24:49mc2739Zagor: rasher's sim builds page has not had updated sims since the 17th. Could that be related to the move?
15:25:09gevaertsrasher updates his sim builds when he feels like it
15:25:40mc2739gevaerts: ok, I thought it was automated
15:27:42kugelmc2739: the page says so too
15:27:48kugel"updated daily"
15:28:35gevaertshm, I might misremember of course
15:29:52 Join evilnick [0] (i=0c140464@rockbox/staff/evilnick)
15:30:34mc2739I am not seeing "updated daily" on that page
15:34:31 Quit TheSeven (Nick collision from services.)
15:34:52 Join The_Seven [0] (n=theseven@rockbox/developer/TheSeven)
15:35:04 Nick The_Seven is now known as TheSeven (n=theseven@rockbox/developer/TheSeven)
15:41:42 Join stoffel [0] (
15:47:19 Join jgarvey [0] (
15:48:25kugelit says "built daily", sorry
15:49:11mc2739kugel: thanks, I was looking on">
15:56:01CIA-85New commit by 03teru (r23130): New plugin theme_remove which offers a way to remove specified theme. ...
15:57:42 Quit MethoS- (Remote closed the connection)
16:00:21 Join lifeless_ [0] (n=lifeless@
16:22:49 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
16:23:57***Saving seen data "./dancer.seen"
16:30:14 Quit TheSeven (Read error: 110 (Connection timed out))
16:33:49 Quit lifeless_ (Remote closed the connection)
16:34:06 Join lifeless_ [0] (n=lifeless@
16:56:15 Quit nls_web ("CGI:IRC (EOF)")
16:56:24 Join mc2739_ [0] (n=mc2739@rockbox/developer/mc2739)
16:56:38 Quit mc2739 (Nick collision from services.)
16:56:40 Nick mc2739_ is now known as mc2739 (n=mc2739@rockbox/developer/mc2739)
17:01:28 Join toffe82 [0] (n=chatzill@
17:01:30 Join n1s [0] (n=n1s@rockbox/developer/n1s)
17:01:38 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
17:02:54 Quit Zagor ("Don't panic")
17:03:31 Join fdinel [0] (
17:09:39 Nick fxb__ is now known as fxb (n=felixbru@
17:10:55kugeltiny albumart in my custom statusbar! :D
17:11:30kugelnext to the normal big one that is!
17:12:40kugel at the upper left
17:14:01gevaertsthat's reasonably tiny :)
17:14:17 Join robin0800 [0] (
17:15:17n1sif you didn't know it's albumart it'd be quite hard to tell :)
17:16:49kugelI need to decouple albumart loading from buffering, and rework the mechanism in playback.c completely
17:18:18kugelthat was the last item on my custom statusbar block list
17:18:21n1show do you mean "decouple albumart loading from buffering" isn't it a good thing they are done at the same time?
17:20:13kugeln1s: the albumart is still buffered
17:20:36kugelbut now buffering doesn't know anymore that it's doing albumart, but rather "just a generic image"
17:21:20 Quit mt (Read error: 101 (Network is unreachable))
17:21:38kugelthe problem is that buffering calls get_albumart_size directly, that doesn't quite work with multiple sizes (and is unclean anyway)
17:24:57n1sdoes it scale on load currently? and are you changing that to load the original image and scale later?
17:25:16kugelit scales on load yes, and I don't change it
17:25:37n1sso you will buffer the image several times?
17:26:13kugelin fact, it scales the same image twice now, and buffers the scaled version, which may or may not result in a noticeable delay
17:26:30kugelbut I don't see a way around, scaling on display is worse
17:27:50n1sdoes it not get loaded into the wps image buffer? if so it could get scaled on that copy, would of course waste a lot of buffer for people with unreasonable aa sizes
17:27:52kugelalso, with my current way of doing this, it will even buffer images of the same size twice, I'm not sure how to fix that best
17:28:32kugelno, it's in the audio buffer and remains there
17:28:41 Quit teru ("Quit")
17:30:02rasherIs anyone strongly opposed to me removing the marquee theme and putting an AUTHORS and COPYING file in the wps dir?
17:31:37linuxstbrasher: Are they two independent things?
17:31:57rashercopying is just a copy of the cc-by-sa 3.0
17:32:03kugeln1s: there's no "wps image buffer" :p
17:32:16kugeldid you miss the grand rework?
17:32:35rasherlinuxstb: marquee is the only one I haven't gotten permission to (re)license
17:32:45linuxstbrasher: Wasn't there talk of removing all themes from SVN once the theme site was up and running?
17:33:14n1skugel: eh ok, s/wps image buffer/buffer where wps images are/
17:33:15rasherlinuxstb: yeah.. but so far that's not happening
17:33:37kugelbut still, albumart is in the audio buffer, and that makes sense imo
17:33:43linuxstbrasher: Yes, I guess it means someone (or some script) packaging them all and uploading to the site...
17:34:00rasherlinuxstb: pretty much, yes
17:34:18linuxstbrasher: But if there are licensing issues which would make things simpler to remove one old WPS, I say go for it.
17:34:39gevaertsrasher: speaking of licensing, the Mountain Fog theme (176x220) seems to use GPL art
17:34:40rasherWe'll also still have cabbiev2 in SVN even after that happens though, so we'll need the license and authors anyway, in my opinion
17:35:31 Join TheSeven [0] (n=theseven@rockbox/developer/TheSeven)
17:35:42n1skugel: sure it makes sense i'm just interested in how it works and how you will change it :)
17:36:16 Quit amiconn (Remote closed the connection)
17:36:16 Quit pixelma (Read error: 54 (Connection reset by peer))
17:36:44 Join dfkt_ [0] (i=dfkt@unaffiliated/dfkt)
17:37:10kugeln1s: buffering isn't changed much (just multiple versions of the albumart per track), the bigger change is in playback.c
17:37:26 Join amiconn [0] (i=quassel@rockbox/developer/amiconn)
17:37:27 Join pixelma [0] (i=quassel@rockbox/staff/pixelma)
17:37:44 Join mcuelenaere [0] (
17:38:15kugelthere's now a struct holding various function pointers. each skin (or anyone who wants albumart buffered), adds his struct to the linked list, on buffering playback iterates through that list handling each skin's albumart
17:38:26mcuelenaereamiconn, kugel, linuxstb, TheSeven: the Onda's already have a boot menu in the boot loader
17:38:40kugelmcuelenaere: argg!
17:38:44mcuelenaerebut that's mostly because the VX777 doesn't have any physical buttons so it needs to use the screen for dual boot
17:38:53mcuelenaereto select the OF*
17:39:06mcuelenaerekugel: I don't see what the problem is?
17:39:12kugelwe can have an exception for that I think
17:39:41mcuelenaerecurrently it has 4 options: boot Rockbox, boot OF, USB mode & reset Rockbox configuration
17:39:50kugelI just don't like the idea, in a year someone comes up with the idea to port grub/lilo because the rockbox bootloader menu is too crappy
17:40:11linuxstbkugel: Then say yes/no to that, not something else...
17:40:20CIA-85New commit by 03rasher (r23131): Clear up theme licensing: ...
17:40:23mcuelenaereso what? let them do that, that doesn't mean it should get included in our tree..
17:40:33*linuxstb doesn't buy into the "don't do X because then people will want Y" argument
17:40:44n1skugel: playback dealing with the aa feels very wrong to me; although i'm not familiar with how the playback/buffering works
17:41:49kugeln1s: there's no way around, only playback knows which track is currently played and which are buffered, and what cleanup is needed, etc
17:42:07mcuelenaereTheSeven: FYI, there are several USB implementations of the S3C64xx OTG/device controller in Linux
17:42:48n1skugel: that sounds to me like something that buffering shoudl deal with
17:43:10kugelhell no
17:43:55rasherlinuxstb: my sims were built on soap's box, and he's vanished.
17:44:08kugelbuffering is a module alone, just doing what it is told: buffering stuff, and removing stuff from the buffer, and the memory management. it shouldn't know what it buffers, or when to buffer
17:44:42linuxstbrasher: Yes, I saw you say that.... I just wanted to check if you knew.
17:45:15kugelalbumart is tied to the currently playing, and upcoming, songs. that's a job for playback
17:45:25 Quit dfkt (Read error: 145 (Connection timed out))
17:45:45 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
17:46:12n1skugel: thei i suppose i think there should be a layer between the current buffering and playback that knows what is buffered and deals with the details of that and to which playback gives events
17:46:43kugelplayback does that
17:47:05n1syes and to me it seems like playback is soing too much
17:47:14kugelit's the glue between buffering, codecs and audio thread. another layer doesn't make much sense
17:47:33 Quit dfkt_ (Nick collision from services.)
17:49:27n1ssince it is preforming such a complex task, further division of labour seems logical to me
17:51:19 Join kkurbjunW [0] (n=karlk@
17:51:22kugelit could need a bit of cleanup yes (like moving some parts to playlist, or splitting playback.c into more source files), but I think the general layout is fine
17:51:32kugeli.e. I don't see a use for another layout
17:51:45kkurbjunWgevaerts: Is there a wiki page on using USB serial?
17:52:11gevaertskkurbjunW: PortalPlayerUsb
17:53:12kkurbjunWhaha, I wish I found that before, I got it working on my own, but only after hours of messing with it
17:54:48 Quit petur ("work->home")
17:55:09kkurbjunWcan wiki topic title's be changed? Since this is no longer specific to the portal player it seems a different title might be helpful
17:55:39toffe82interesting for the sigmatel chips : : SigmaTel 35XX SDK Device Updater
17:57:14kkurbjunWgevaerts: USB high speed works fine on the m66591 - is the fullspeed/highspeed note still pertinent for the portal players?
17:57:46toffe82I think it can be used with all the sigmatel chips..
17:57:55kkurbjunWI mean, specifically for USB serial that is
17:58:24 Join tomers [0] (
17:58:34gevaertskkurbjunW: ah, no. The usb_serial was fixed to work around the issues
17:59:18 Join uflops [0] (
17:59:35gevaertsThe problem was that packet sizes larger than 96 bytes don't work properly on PP without boosting. Serial now uses 32 or 64 byte packets (can't remember which)
17:59:37 Quit polobricolo ()
18:02:02kkurbjunWgevaerts: I see that we have a UsbSoftwareStack page - it looks like it is there for historical reasons, but would it make sense to have the USB serial on it's own "driver" page? I found some other information online that seems to indicate that some people have made it work in windows too by writing an INF
18:02:48kkurbjunWgevaerts: the PP connect at high speed correct?
18:03:19gevaertsif the host does, yes
18:03:55 Join liar|netbook [0] (n=liar@
18:04:21kugelkkurbjunW: why don't you ask earler then? :)
18:04:29 Quit TheSeven|Mobile (Read error: 104 (Connection reset by peer))
18:04:30kkurbjunWdo you know if the serial driver takes the speed into account? packet sizes are limited to 64 bytes in full speed right?
18:04:39kkurbjunWkugel: I assumed no one was awake
18:05:17kkurbjunWit was in the evening yesterday that I was messing around with it
18:05:30kkurbjunWI guess it doesn't have to
18:06:21kkurbjunWbut I was wondering if the problems were showing on systems that only connected at fullspeed, or if it was having trouble at high speed.
18:06:22gevaertskkurbjunW: it does, yes
18:10:21kkurbjunWgevaerts: the other thing I was running into is that the start of some of the packets coming from the serial driver were not word aligned which was causing data aborts on the M:Robe 500. I can fix that in the driver for the m66591, but I was wondering if you know offhand whether the other USB drivers can handle packets that do not have a starting point that is word aligned? I was thinking about forcing the serial driver to align the start of all of it's p
18:12:01kkurbjunWthe m66591 driver usually works with shorts, and switches to byte mode when it gets to the end of a packet and it is not half-word aligned. In doing so it assumes that the start of the packets are word aligned, which works on the storage driver, but fails when I enable logging for buffering
18:13:47kkurbjunWsorry, this: "end of a packet and it is not half-word aligned" should read: "end of a packet and it's length is odd"
18:14:36 Quit mikroflops (Read error: 110 (Connection timed out))
18:16:11 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
18:16:48 Join bertrik_ [0] (
18:18:21 Quit bluebrother (Nick collision from services.)
18:18:23 Join bluebroth3r [0] (n=dom@rockbox/developer/bluebrother)
18:20:08 Quit tomers (Read error: 113 (No route to host))
18:23:56 Join polobricolo [0] (
18:24:00***Saving seen data "./dancer.seen"
18:27:25 Join tomers [0] (
18:28:13tomersI am about to commit the patch in "FS #10670 - The first letter of a scrolling line starts to appear at the end of the line" which finally seems to solve the scrolling issue. I've tested it on both RTL and LTR languages. Any objections?
18:28:21linuxstbrasher: So themes in SVN are now dual-licensed?
18:29:38rasherlinuxstb: They'd have to be, considering they've been implicitly gplv2 or later up until now
18:31:47CIA-85New commit by 03tomers (r23132): Fix FS #10670 - The first letter of a scrolling line starts to appear at the end ...
18:32:25linuxstbrasher: Yes, I understand that. But what about future modifications?
18:33:17rasherlinuxstb: I guess we have free hands there. Maybe we should just switch completely
18:34:46 Quit maruk ("Leaving.")
18:36:12linuxstbrasher: Yes, I think so - just to keep things simple.
18:37:47linuxstbrasher: BTW, how did you determine the authors list? I recall making some changes to some themes...
18:39:20rasherlinuxstb: I'm not entirely sure.. this was a while ago, but I think I looked mainly in the theme files themselves, and commit messages
18:39:36kugeltomers: I'm not sure if that fix is correct
18:39:45linuxstbYou're probably right - looking back at the changelog, my changes were all relatively minor...
18:39:56kugelthe "ofs" is a bad variable name.
18:40:02tomerskugel: It worked both for English and Hebrew scrolling
18:40:36tomersBoth in RTL mode (Hebrew interface) and LTR mode (English interface)
18:40:56kugelhave you tried if it breaks holding left/right in the main menu? this moves the entire menu to the left/right (that's what ofs is for)
18:42:46kugelif ofs is > 0, it actualy draws the text (for ltr) with the first few pixels cut away ( "File Browser" gets "le Browser". all other items too)
18:42:59kugelI'll just try
18:43:28CIA-85New commit by 03rasher (r23133): Move to only CC-BY-SA 3.0 for future changes.
18:45:40 Quit polobricolo (Remote closed the connection)
18:46:23 Join polobricolo [0] (
18:46:31tomersrasher: It works just fine, when scrollbar is in either one of the side
18:46:44*tomers away
18:49:33kugeltomers: seems to work, although setting ofs to 0 seems wrong to me at a quick glance
18:49:44 Quit robin0800 (Remote closed the connection)
18:51:31 Quit intrados1 (Read error: 60 (Operation timed out))
18:51:59gevaertskkurbjunW: at least the ARC controller doesn't care about alignment
18:53:18gevaertsI think it would be reasonable to say that you need alignment if you want decent performance, and use slower byte-oriented methods if the buffer isn't aligned
18:56:41 Join robin0800 [0] (
19:03:18*kugel questions the use of tiny albumart now that he sees it on target :P
19:03:44gevaertskugel: you need a bigger status bar of course :)
19:04:18 Nick uflops is now known as mikroflops (
19:06:37 Quit robin0800 (Remote closed the connection)
19:06:40kugelgevaerts: and a bigger screen ;)
19:11:36 Quit liar|netbook (Read error: 148 (No route to host))
19:14:23 Quit kkurbjunW (Remote closed the connection)
19:15:19 Join codearn [0] (
19:15:30 Join kkurbjunW [0] (n=karlk@
19:19:07amiconnrasher: Imo building win32 sims should be turned into a configure option. Actually adding them to the build system will require a workaround for the dreaded sectioned compilation warning though :\
19:20:43rasherIt will also be complicated by having to have a different PATH set
19:21:01rasheror some other way to let it know where the proper sdl-config lives
19:21:01 Quit antil33t (Read error: 104 (Connection reset by peer))
19:21:15 Join antil33t [0] (n=Mudkips@
19:24:01amiconnIt would need to interate through all sdl-config binaries it can find, and check which is which
19:24:20 Quit Rob2223 (Remote closed the connection)
19:24:22amiconnThat's a comparably small problem...
19:25:00 Join Rob2222 [0] (
19:25:43codearnHi everyone. I just put the latest build on my sansa e200v2 and the "limiter preamp" doesn't show up on the sound settings menu. I couldn't find anything on google other than a mention that pitch control has to be off, but I can't find a way to disable it... I'm sure I'm missing something...
19:25:50 Quit Rob2222 (Client Quit)
19:26:06 Join Rob2222 [0] (
19:27:01 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
19:30:11 Join saratoga [0] (i=98039f25@gateway/web/freenode/x-qaalunyzcusrmzpa)
19:30:32saratogaaccording to the changelog: "DSP limiter replaced by a dynamic range compressor"
19:31:01 Join tomers_ [0] (
19:31:31 Quit tomers (Read error: 113 (No route to host))
19:31:42 Nick tomers_ is now known as tomers (
19:31:46tomerskugel: Let me know if there is something more to do with regards to this issue
19:32:45codearnoh i see. thanks a bunch. i had a feeling that might be it.
19:32:47 Join Strife89 [0] (n=michael@
19:32:59rasheramiconn: Do you think crosscompiling should be an advanced option?
19:33:53codearndo you know if there is documentation available on the compressor? the sansa manual doesn't have it yet
19:34:20amiconnDoesn't really matter, as long as there is no solution for that warning. It's one of the few warnings which cannot be disabled individually
19:34:37amiconnAnd in fact the warning doesn't even make sense
19:35:43kugelcould the build system whilelist that warning so that it doesn't case a yellow?
19:36:33 Join merbanan [0] (
19:36:44 Join Grahack [0] (
19:38:08 Quit stoffel (Remote closed the connection)
19:42:05rasherI wouldn't say it's completely useless. The current way to do crosscompiling is pretty gross
19:42:12rasherI'm switching it to an advanced option now
19:43:58amiconnIt warns that sectioned compilation may affect debugging. But that simply doesn't apply to the win32 obect and binary formats
19:45:05 Quit evilnick (Ping timeout: 180 seconds)
19:48:11GodEatermy rsync-fu is broken :(
19:51:12 Join midgey [0] (n=tjross@rockbox/developer/midgey)
19:51:46midgeyamiconn: is the warning something similar to "-ffunction-sections may affect debugging on some targets"?
19:55:01 Join petur [0] (
19:55:01midgeythe same warning shows up on OS X 10.5 when compiling the sim. it appears to be gone on 10.6
19:56:54 Join Zarggg [0] (
20:00:35 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
20:00:35 Join bughunter2 [0] (n=bughunte@unaffiliated/bughunter2)
20:01:27 Join Eragon55 [0] (
20:01:28 Quit codearn ("CGI:IRC (EOF)")
20:04:19 Join Thundercloud [0] (
20:04:27 Join FOAD_ [0] (
20:08:47 Join liar|netbook [0] (n=liar@
20:09:34kugelhmm, I'm having a problem
20:09:43 Join panni_ [0] (
20:09:55kugelI post something to the audio thread with queue_send/_post, but it doesn't come through
20:10:30 Join T44 [0] (
20:11:40CIA-85New commit by 03nls (r23134): Fix typo from r19579 that prevented this code from building, avoid copying lang strings
20:12:20*TheSeven just found a tricky ibugger bug by accident while porting the USB code to rockbox
20:13:05kugeloh, my queue_send comes before audio_init()
20:13:05*n1s wonders if anyone uses that charging screen on their recorders, the text drawing has been broken since late 2008
20:14:18kugelaudio_init comes after wps_init, I'm not sure if I can swap them easily
20:14:29rasheramiconn: is it "warning: -ffunction-sections may affect debugging on some targets" you were talking about?
20:14:36rasheroh ah
20:14:41rasherYou just answered that, I see
20:14:49rasheranyway, I think I have it working now as a configure option
20:15:39kugelamiconn: can I shuffle a bit in init()? I need settings_apply after audio_init()
20:16:36 Quit Topy44 (Read error: 60 (Operation timed out))
20:16:55 Quit FOAD (Read error: 110 (Connection timed out))
20:16:55 Nick FOAD_ is now known as FOAD (
20:20:39 Join Jaykay [0] (
20:21:30amiconnInit order is quite important for certain things. I don't remember whether this is the case for these two things
20:22:19Jaykayhi, I'm getting only seven columns from dev.cgi
20:22:29Jaykayfrom the first table
20:23:20bertrik_kugel, I've been thinking a bit about your idea for doing the DBOP reads
20:23:40bertrik_if we make sure to wait for the FIFO to be empty in the interrupt routine, we might not need the dbop locks at all
20:24:01***Saving seen data "./dancer.seen"
20:24:45amiconnkugel: Idk which one you intended to move, but moving settings_apply past audio_init is a very bad idea and must not be done
20:24:45bertrik_this wasn't possibly previously because we communicated directly with the LCD, but now we're just communicating with the buttons on the DBOP data bus
20:25:40linuxstbJaykay: That's normal now - only columns with yellow or red are shown.
20:25:40TheSevenJaykay: you shouldn't get any :-)
20:25:53amiconnMany modules inited after the current settings_apply() call need the settings already applied (that's why it's done that way)
20:26:54Jaykayok, thanks :)
20:26:58bertrik_kugel, there may be some time wasted in the read button interrupt waiting for the FIFO to drain, but it's probably not *that* bad in practice
20:27:18rasherIs anyone - anyone at all - interested in checking out the patch I have for making simulator crosscompiling a configure option?
20:27:34linuxstbrasher: I'll give it a shot...
20:28:32kugelamiconn: why?
20:29:11amiconnSeveral modules need the values in order to know how much buffer they need to allocate
20:29:16rasherlinuxstb: you should have a crosscompiled SDL in your PATH (preferably last, or you won't be able to compile regular sims...)
20:29:32kugelI mean you didn't say why audio_init() must not be done after settings_apply()
20:30:02amiconnI did not say that
20:30:07FlynDicen1s: Are you working on that unused variable warning for r23134 or is it ok?
20:30:58kugelok, I missread then, can I move audio_init() up? there's mp3_init() for hwcodec which looks suspicous in that regard
20:31:02amiconnI said that moving settings_apply after audio_init would be a bad idea.
20:31:12amiconnThere are dozens of inits happening inbetween
20:31:58amiconnIt *might* be possible to move audio_init() up, but I wouldn't dare to without testing *all* targets
20:31:59kugelnone of them looks like they depend on the audio thread doing nothing in their init, I haven't looked at them though
20:32:13 Quit n1s (Read error: 110 (Connection timed out))
20:32:46 Quit Jaykay ("ChatZilla 0.9.85 [Firefox 3.5.3/20090824101458]")
20:32:52kugelhm, there a few inits which do grab the audio buffer, that won't work
20:33:01linuxstbrasher: I don't think it feels right to have it as an advanced option - I would expect to at least choose "S" in the first menu. "W" is also free in the first menu, even though that's getting very busy...
20:33:39amiconnAnd I also think hwcodec audio_init() needs mp3_init() to be done already
20:33:58linuxstbrasher: Also, if I selected "w" first, then "s" in advanced options, it didn't work. I needed "s" then "w"...
20:33:59kugela) I do settings_apply() again after audio_init(), or b) I don't know how to do it :/
20:34:11rasherlinuxstb: but that would add another question each time you build a regular sim
20:34:16*amiconn wonders why kugel suddenly wants to shuffle inits
20:34:35linuxstbrasher: I'm not suggesting that.
20:34:53rasherlinuxstb: what then? Another top-level entry?
20:36:21kugelamiconn: in my multiple album art work, skins send something to the audio thread (in order to be thread safe). the skins are already parsed in settings_apply(). their post goes into the dark at the boot
20:36:35kugeli.e. it doesn't work unless the skin is reloaded manually
20:37:20linuxstbrasher: I'm not sure.... I guess what's there now is as good as anything.
20:37:40rasherI guess there's no need for crosscompiling to depend on simulator. With any luck the database and checkwps tool could work as well
20:38:14rasherexcept you can't enable those through the advanced menu
20:38:48*amiconn doesn't seem to understand without additional context information
20:39:20linuxstbrasher: But it seemed to correctly detect that I don't have a cross-compiled sdl installed...
20:39:47kugelamiconn: the skins attach a struct of callbacks to a linked list in playback.c. the callbacks are only called in the audio thread, so it would be safe to attach them in the audio thread only too
20:40:04rasherlinuxstb: Great
20:40:21linuxstbrasher: (and I do have a native sdl-config)
20:40:52rasherI've removed the dependency on simulator being set.
20:41:46kugelI could use an event, so that the skins send it again after the audio thread inited
20:41:50 Join n1s [0] (n=n1s@rockbox/developer/n1s)
20:42:16amiconnHow do skins "send" something? Aren't they simply loaded in the main thread?
20:42:19 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
20:42:37kugelyes they are
20:43:00kugelthey call playback_add_aa_cb(), which adds the parameter to a linked list
20:45:12*amiconn still doesn't see the problem
20:46:20kugelthe linked list is accessed for doing the actual albumart work in the audio thread only, while the adding to the linked list is (currently) done in the main thread
20:46:37kugelremoving can also happen
20:47:27kugelthe main thread may remove a struct from the linked list, while the audio thread is accessing it
20:47:35CIA-85New commit by 03nls (r23135): fix yellow
20:47:57amiconnNo it may not. Remember, rockbox does cooperative threading
20:48:03domonokyif the audio thread does not yield when looping over the linked-list, it should be pretty safe to do it from the mainthread. we have cooperative multitsaking :-)
20:48:20domonokys/multisaking/multitasking :-)
20:48:42kugelit does
20:48:51kugelthe scaler yields while scaling
20:49:01linuxstbrasher: I can confirm that a normal sim still builds OK.
20:49:31CIA-85New commit by 03rasher (r23136): Change the Windows crosscompiling logic: ...
20:51:05 Quit midgey ()
20:51:51 Join midgey [0] (
20:52:01amiconnThe scaler is only called when loading album art though, not for loading skins
20:52:39*amiconn is slightly confused
20:53:26kugelif the albumart size in the wps changed (after loading a skin), then audio is restart to pick the new sizes up. I happen to send the stuff after that
20:53:48 Join DataGhost [0] (
20:53:50kugelbut while writing it, I notice should probably change that anyway
20:54:40kugelwhat if someone changes his theme/skin while rebuffering when the scaler is active?
20:55:13CIA-85New commit by 03FlynDice (r23137): AMS Sansa: Remove BUSWIDTH and BLOCKLEN commands and revise strategy for High Speed SD Card. ...
20:55:40amiconnHmm. You could protect skin access with a mutex
20:55:51rashertomers: There seems to be an issue with the bar line-selector when using RTL menus
20:56:02rashertomers: it doesn't cover the entire width of the screen (viewport?)
20:56:03amiconnBut the scaler puts the loaded album art into the audio buffer, not into the skin
20:56:36kugelone of the callbacks in the struct is called just before the scaler gets active to get the albumart dimensions
20:56:41amiconnOnly if that track starts playing, the audio thread copies the bitmap for display (certainly without yielding)
20:57:18kugelthe bitmap isn't copied
20:57:44amiconnThen a pointer is changed somewhere... doesn't really matter
20:59:30kugelit seems possible to me that the linked list is changed while the scaler is doing his work
21:00:18amiconnIf you really want to deal with complex signalling and load skins from the audio thread (btw, you'd then also need to do an implementation for hwcodec, which still doesn't use the same thread), it would still not require shuffling inits
21:00:21kugelin the next look it even may result in a null-pointer access if a struct was removed
21:00:35amiconn[20:55:45] <amiconn> Hmm. You could protect skin access with a mutex
21:00:58kugelI'm gonna think about it more
21:01:07CIA-85New commit by 03rob (r23138): Replace a couple of magic numbers with the appropriate constants.
21:02:25kugelMaybe I'm just gonna use a one-shot event at the end of audio_init(), as it's only an issue at booting
21:02:54amiconnYou could also check if audio is already inited, and if not, just call the function
21:03:07amiconnThere will be no race at boot because the audio thread doesn'
21:03:12amiconnt exist yet
21:03:56amiconnBut don't forget KISS (and that there are two places you need to hook your code if you're going for the complex solution)
21:04:05kugelthen I'd have to expose both, the function that does queue_send(), and the one that does the actual work. but that would work too
21:04:31amiconnEven that isn't necessarily the case
21:04:47kugelah yea, you're right
21:04:52amiconnThe function that normally does queue_send is probably part of audio, right?
21:04:57kugelI think in playback.c there's already a way to detect that
21:05:07amiconnAnd if it is, it should know whether audio is already inited
21:05:22kugelthat seems like the way to go, thanks!
21:05:24amiconnIt could decide itself
21:05:35 Quit avacore^ (Read error: 60 (Operation timed out))
21:05:36 Join pegwole [0] (
21:05:46*amiconn mentions mpeg.c once more, just in case
21:06:00 Join avacore [0] (
21:06:07kugelI'll have a look, maybe :p
21:06:28pegwolei'm pretty sure i'm retarded, but can you put rockbox on the RCA M4303?
21:06:39*TheSeven wonders if this TCC usb driver can actually work
21:06:47kugelthere isn't a hwcodec that does albumart, and since everythign is #ifdef HAVE_ALBUMART I see only a little point (i.e. just for completeness)
21:07:22TheSeven(for more than control transfers)
21:07:29kugelamiconn: does the current hwcodec audio handle albumart in any way?
21:07:43amiconnNo, but it does load skins
21:08:12kugelalright, a stub should do it then
21:08:34amiconnIn theory all bitmap targets could do album art though, just that 1-bit album art isn't really recognisable in most cases
21:09:30kugeland even the stub would be within #ifdef HAVE_ALBUMART
21:10:12 Join intrados1 [0] (
21:10:56 Quit intrados1 (SendQ exceeded)
21:11:48krazykitpegwole, supported targets are listed on the front page
21:12:53 Join intrados1 [0] (
21:16:28pegwolei'm not seeing it...shit
21:16:58pegwolewelp looks like i have to hate freedom using this thing :(
21:25:40TheSevengevaerts: what the hell is usb_drv_set_test_mode?
21:26:13 Quit ruj1 ()
21:27:41 Join shotofadds [0] (n=rob@rockbox/developer/shotofadds)
21:27:44 Join barrywardell [0] (n=barrywar@rockbox/developer/barrywardell)
21:27:59linuxstbTheSeven: AppleOS now loads if I don't enable the cache...
21:28:16TheSevendo you disable the cache before firing it up?
21:28:20shotofaddsTheSeven: the tcc usb driver has worked in the past, I've used it to transfer several hundred megs of raw NAND data once
21:28:38shotofaddsUSB serial works too, but it's rather unreliable all round
21:29:11 Quit midgey ()
21:29:15TheSevenshotofadds: I just keep seeing things like "if (endpoint != 0) panicf_my(..." which should pretty much block all non-control traffic
21:29:19linuxstbTheSeven: No, I haven't tried that. Would the four lines of code in crt0.S with the comment "// disable caches and protection unit" be enough?
21:29:33TheSevenshould work
21:29:41shotofaddsTheSeven: er yes, I remember, I had to comment a few bits out ;-)
21:30:00TheSevenconventions for switching to the next stage of code is irq/fiq disabled, ROM mapped to zero, protection unit off
21:30:19shotofaddsunfortunately I know nothing about USB and never got around to learning enough to try and fix it
21:32:40 Part pegwole
21:34:44linuxstbTheSeven: Yes, that seems to work ;)
21:34:45gevaertsTheSeven: see usb 2.0 7.1.20. It's to put the device in special modes that allow electrical measurements (things like waveform)
21:35:06TheSevenwhat the hell do we need this for?
21:36:19gevaertsonly for strict spec compliance
21:36:33TheSeveni.e. this will never be called anyways?
21:37:35gevaertsI'd at least make sure not to crash if it is
21:39:20TheSevengevaerts: does the core expect endpoint states to be cleared or retained on bus resets?
21:39:51gevaertshm, good question
21:40:27gevaertswhich endpoint states do you meant? Things like stall, or allocations?
21:41:52TheSevenand will the core change the address to zero again, or does the driver do that? (well, I'll just do it, can't hurt)
21:42:12 Join froggyman [0] (
21:42:28gevaertsendpoing allocations are kept, transfers in progress and stalls are not
21:43:32 Quit n1s (Read error: 60 (Operation timed out))
21:44:26gevaertsalthough clearing endpoint allocations won't hurt either
21:44:39 Join stoffel [0] (
21:45:49gevaertsbasically you're allowed to reset everything
21:46:15 Quit Eragon55 ("ChatZilla 0.9.85 [Firefox 3.5.4/20091007001339]")
21:46:46CIA-85New commit by 03dave (r23139): Nano2G bootloader - fix dual-booting the Apple firmware.
21:47:18 Join Sajber^ [0] (
21:51:30TheSevenoh, that USB driver is getting funny
21:51:40TheSevenDMA vs. cache again...
21:53:52TheSevenwhat a waste of space and time...
21:54:03TheSevenI'll need to copy over every single buffer used for USB
21:59:21amiconnWhy copy?
22:01:17TheSevenfor some weird unknown reason flushing the caches doesn't work properly
22:01:38TheSevenso we usually just circumvent them, but we don't have control about the location of the buffers here.
22:02:04TheSevenso I'll need to copy everything from a cached to an uncached buffer and back
22:02:47linuxstbbluebroth3r: What would be needed to add Nano2G support to rbutil? I now have a working patch for ipodpatcher, which I should commit later tonight (after double-checking it doesn't affect earlier ipods)
22:03:27linuxstbTheSeven: Have you given up trying to get cache flushing working?
22:03:34amiconnSounds like investigating cache flush/invalidations would be a good idea
22:05:04TheSevenI'm doing it like it's described in the datasheet, but it doesn't work
22:05:17TheSevenI was told the samsas have the same problem
22:05:25TheSevenor were it the old sansas?
22:05:26 Join notlistening [0] (
22:05:39kugelwe had the problem, but not anymore
22:06:10TheSevenyes, but you also circumvented it by just circumventing the cache at all, right?
22:06:31TheSeveni.e. the flushing itself is still broken?
22:06:46kugelbut "DMA vs cache" sounds awfully known to me
22:08:29notlisteningFlynDice, just be testing the latest cpu boost patch that you posted, it not reads the SD card after startup at 1.05v but after exactly 60 seconds of run time i am getting unreferences instruction error even without the SD card inserted
22:09:06 Quit stoffel (Remote closed the connection)
22:09:53FlynDicenotlistening: Can you try with 1.10 volts and see if that works?
22:09:53notlistening*now reads
22:10:12notlisteningFlynDice, yeah that works nicely
22:10:13kugelI don't know too much about that stuff, but the cpu doesn't have much control about DMA, right (other than starting the transfer)? that somehow sounds to me as DMA with cached buffer can't work reliably. but feel free to ignore me
22:11:20notlisteningFlynDice, o just wanted to try on the lower setting as it is not fair my player doesn't like it
22:11:24amiconnTheSeven: PP is even more fun, as it has two cores with independent caches, yet we managed it without resorting to copying everything
22:11:44linuxstbDon't we just used the uncached memory aliases?
22:11:45 Quit Grahack ("Leaving.")
22:11:52 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
22:12:22TheSevenyes, that works fine, but not for USB, as the sources of all these buffers are scattered all over the place
22:12:39amiconnWe do in several places, yes, but now everywhere
22:13:18JdGordonsomeone fix tuner.c already! its painful to work in
22:13:30FlynDicenotlistening: I think the low voltage is the culprit. It seems when players have problems and we raise the voltage back to 1.10 the problems go away. Some players do ok, others can't habdle it.
22:14:00kugelFlynDice: I'm not surprised, wasn't 1.10V already a questionable setting?
22:14:03amiconnBasic operation is obvious: flush cache before transferring something from a buffer to a peripheral via dma, and invalidate the cache before transferring something from a peripheral to a memory buffer via dma
22:14:07*linuxstb wonders where the platform numbers in rbutil.ini come from
22:14:32FlynDicenotlistening: Not really, the OF uses 1.10
22:14:34amiconnThen there are the little nasty things one needs to observe, e.g. cache line boundaries
22:14:36TheSevenamiconn: exactly what we did, and what didn't work
22:15:14amiconnThese impose additional requirements for buffer alignment
22:15:15 Join Djeepp [0] (n=smiller@
22:15:33kugelFlynDice: questionable as in the datasheet recommends to not use it, IIRC we do use it just becuase the OF does
22:15:46TheSevenamiconn: you should be able to get rid of that by just cleaning/invalidating too much
22:16:10amiconnNot in all cases
22:16:38linuxstbdomonoky, bluebroth3r: ping
22:17:08amiconnE.g. if you're going to transfer something from a peripheral to a buffer, you invalidate the cache, then kick off dma
22:17:32amiconnWhile dma is running, you calculate something in an adjacent buffer, with the cachelines overlapping -> BAM
22:17:34linuxstbdomonoky: Do you know what's needed for ipodnano2g support in rbutil? I'm loooking at rbutil.ini first...
22:17:51linuxstbdomonoky: Why do only some targets have USB IDs?
22:18:05amiconnThe cpu will cache the whole line, possibly before the dma transfer changed the in-memory data
22:18:15DjeeppThe scroll down key on my Gigabeat F40 doesn't work at all unless I turn the touchpad sensitivity to high. I'm guessing I need a new touchpad. Is there a place I can get this part?
22:19:24domonokylinuxstb: the usb-ids are the way they are, because of the current autodetection. (there is a patch in the tracker witch tries to improve it, and use all usb-ids)
22:19:25amiconnAfter the dma finished, the contents of the transfer buffer is invalid as seen by the cpu, because it re-cached too early
22:19:37linuxstbdomonoky: Ah, I'm just looking at the "how to add a new target" in the wiki page.... Given that's it is supported by ipodpatcher, I guess I just need to edit rbutil.ini and recompile with my new ipodpatcher code...
22:19:45amiconnYou can construct such an example for the other dma direction as well
22:19:51domonokylinuxstb: for ipodnano2g you shouldnt enter a usb-id at moment.
22:20:20bluebroth3rlinuxstb: pong
22:20:33domonokylinuxstb: yes, add a entry to rbutil.ini similar to the other ipod entrys, recompile and it should work .
22:20:35linuxstbdomonoky: Yes, I won't. But I've deleted the USB ID from the list later on - I assume those are "unsupported" ?
22:21:07linuxstbAlso, is there any significance to the platform numbers? I see they're similar to the configure menu...
22:21:15domonokyah, yes the list with only usb-ids is for targets we dont support.
22:21:51bluebroth3rno, the platform numbers are only used for sorting (btw, are they used anymore?)
22:21:54domonokylinuxstb: they should be the number from configure, its used for voicefiles (if i didnt fixed that)
22:22:18amiconnThe menu numbers from configure shouldn't be used elsewhere
22:22:19bluebroth3rdomonoky: really?
22:22:51CIA-85New commit by 03jdgordon (r23140): fix a redraw bug when a static token (like %C) is the only token on a sub/line.. hopefully no bad sideeffects...
22:22:53 Nick bluebroth3r is now known as bluebrother (n=dom@rockbox/developer/bluebrother)
22:22:57domonokyat least i introduced them, to use it in voicefiles (its in the header).
22:23:12domonokyoh, its not the menu-number, but the target-id from configure.
22:23:34linuxstbAlso, how should out-of-tree building work? I create a "build-rbutil" and then run "qmake ../rbutil/" ?
22:23:40bluebrotherhmm. Is that used in Rockbox anywhere? It would be much better to get rid of those numbers completely
22:23:43bluebrotherlinuxstb: yes
22:23:56linuxstbbluebrother: I tried that, and got a load of errors at the start about lang files
22:24:03***Saving seen data "./dancer.seen"
22:24:20linuxstbHmm, strange, it working a second time...
22:24:21bluebrotherlinuxstb: that's normal if you are building a release (which is the default)
22:24:29domonokywe shouldnt need the target-id in rbutil.ini anymore, because its also in the rockbox-info.txt file.
22:24:49domonokylinuxstb: you can ignore the lang file errors :-)
22:24:50bluebrotherit's simply rcc complaining that it can't find the *.qm files, which get generated later.
22:25:19bluebrotherto be exact, it's even qmake that is complaining. rcc isn't running that moment.
22:25:37linuxstbDo you have a new rbutil release planned for the near future?
22:25:48bluebrotherbtw, you can also build using -d -p path/to/
22:26:13*bluebrother hasn't planned a release
22:26:26domonokylinuxstb: my rbutil.ini file doesnt contain any target-id anymore.
22:28:15*bluebrother started working on a new install UI, but that will take some time until ready
22:30:05domonokywhen ipod2g gets unstable status, it might be a good time to make a new rbutil release. :-)
22:30:38amiconnlinuxstb: I don't think too much differentiation between targets would be good
22:31:56bluebrotherdomonoky: btw, there's a bug with languages: the tts window saves the language string (like "english") while the system language uses two-letter abbreviations (like "en") as in the ts filenames. Though I'm wondering why those two values are using the same setting at all
22:32:03amiconnAnd as another contradiction to your list - rockbox on the H300 isn't fully functional compared to the OF. A major feature is missing - USB host.
22:32:39TheSevenwow - there are DAPs that support USB host out of the box?
22:32:45 Join Stephen__ [0] (
22:33:18linuxstbamiconn: So you don't want a "top-tier" category of targets? I'm aware of missing USB host on the h300 (and X5?), that's why my description includes "possible minor exceptions".
22:33:28bluebrotherTheSeven: the beast also claims host support. Never tried it, though.
22:33:40domonokybluebrother: i dont really understand. can you point me to the code doing it ?
22:33:49amiconnOn the H300, usb host should be doable as the chip is documented. The one in the X5 is not.
22:33:52toffe82Djeepp: I have a lot of them in stock
22:34:11amiconnIt's "just" a lot of work...
22:35:14Djeepptoffe82: Do you have a website?
22:35:46bluebrotherdomonoky: createvoicewindow.cpp:53: RbSettings::setValue(RbSettings::Language, lang);
22:36:17linuxstbdomonoky: "Unstable" Nano2g should be quite soon... I think the port itself is there, I just need to release a bootloader and new ipodpatccher.
22:36:31bluebrotherprobably something I broke when reworking the settings −− no idea if that was separated from the application language before.
22:36:52bluebrotherbut how should it work? I'd say it should be a setting different from the application language.
22:37:01 Quit Thundercloud (Remote closed the connection)
22:37:11bluebrotherlinuxstb: impressive how fast that was done.
22:37:34domonokybluebrother: true, that looks wrong.
22:37:35domonokyit should use VoiceLanguage
22:38:06domonokyline 79 tries to use this, and Language as a fallback, but it stores it to the wrong setting.
22:38:13linuxstbHmm, my rbutil compiled with speex errors
22:38:35 Join ysae [0] (n=ysae@
22:38:36linuxstbrbspeex.c:(.text+0x59d): undefined reference to `speex_resampler_destroy'
22:38:43JdGordonkugel: how did you end up doing multi-aa?
22:38:47kugelmy attemps to store the handle ids for the albumarts on the skin buffer all failed
22:39:11kugelJdGordon: don't ask :D
22:39:39toffe82Djeepp: no, just rockbox ;) I didn't mentionned them here but I have them...
22:39:40JdGordontoo late
22:40:07kugelwell, some struct and a linked list
22:40:44CIA-85New commit by 03Domonoky (r23141): rbutil: store the voice language in the correct setting.
22:40:51JdGordonshow us a patch?
22:41:00*JdGordon is bored and curious
22:41:09kugelit's a really complicated area, keeping all stuff local to playback nor keeping it all local to the skins doesn't work,
22:41:35kugelmeh, it just doesn't work easily
22:41:37linuxstbbluebrother, domonoky: Any thoughts about my speex problem?
22:42:00JdGordonI thjought you said you got it going ok? or just "good enough"?
22:42:01bluebrotherlinuxstb: I've heard of that before but can't reproduce.
22:42:02Djeepptoffe82: Are they new or recycled from old ones?
22:42:14linuxstbbluebrother: I can! ;)
22:42:31toffe82Djeepp: recycled
22:42:33bluebrotherlinuxstb: then fix it! ;-)
22:42:43*bluebrother checks if he has that VM still around
22:42:55kugelJdGordon: kugel-rb.git?a=commitdiff;h=e1482953a0892db9c59fc1d0e412c14468586e38;hp=288363b2389d5d965700a7018661605367c346db">;h=e1482953a0892db9c59fc1d0e412c14468586e38;hp=288363b2389d5d965700a7018661605367c346db
22:42:58bluebrothermy guess is that it's related to the version of speex. Mine is rather old
22:43:55kugelJdGordon: I thought it worked ok, but then I noticed heavy instability :p
22:43:55linuxstbYes, I recall recent API changes to speex (well, around the time we switched to speex for voice files)
22:43:55toffe82Djeepp: I don't think you will find new ones
22:44:03domonokydont we use the speex copy in rockbox svn for rbutil ?
22:44:03linuxstbbluebrother: I seem to have 1.5.0
22:44:22amiconnbluebrother: Wasn't there a change to use the system's libspeex recently? Maybe that was a bad idea...
22:44:45JdGordonkugel: I really think the best thing right now is to not allow %C in the statusbar skin
22:44:52JdGordonand ignore all the complication
22:45:01bluebrotheramiconn: yes, and that's the cause. That was based on an old task that introduced that. Though I'm starting to wonder if it was a bad idea, indeed.
22:45:17kugelnah, you gotta love the tiny albumart, have you seen my screendump from earlier today?
22:46:26gevaertsJdGordon: I want album art in my status bar!
22:46:27Djeepptoffe82: I guess my question is, will I have the same problem as I have now? The buttons often stick and sometimes don't work at all.
22:46:56JdGordonmulti-aa should be a seperate beast commit anyway
22:47:08TheSevenlinuxstb, amiconn: major problem.
22:47:11kugelit will be, don't worry
22:47:12JdGordonthere is already a "wtf do we do about remote aa's?" comment...
22:47:23linuxstbTheSeven: With what?
22:47:24TheSeventhe 940T can only flush cache *indices*, not cache *addresses*
22:47:33TheSevenso we'll probably need to wipe the whole cache?
22:47:42 Quit JdGordon ("Leaving.")
22:47:55TheSevenbut that will have trouble with yielding
22:48:16amiconnOn most (all?) other targets we just flush or invalidate the whole cache, not individual cache lines
22:48:43amiconnTargets which have a cache of some sort, obviously
22:49:13kugelI think the beast has a function to do it for single lines, no idea if it's used
22:49:48TheSevenamiconn: we can't just *drop* the cache after the operation then
22:50:41amiconnNo, you always need to flush before invalidating
22:51:26bluebrotherlinuxstb: a quick fix is to comment out the lines calling pkg-config in and the librbspeex Makefile
22:51:33amiconnIirc our cache_invalidate() function in fact does both (on targets with a data cache or unified cache)
22:55:24 Join n1s [0] (n=n1s@rockbox/developer/n1s)
22:56:37 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
22:59:08linuxstbbluebrother: Yes, I've reverted 23016 for now...
22:59:14notlisteningIs there something that runs 60 seconds after boot on rockbox?
22:59:56gevaertsI hope so
23:00:44notlisteningdomonoky, I found a few issues with rgutil on openpsapi but but the in task
23:00:51bluebrotherlinuxstb: hopefully I'll be able to reproduce that once my VM setup is finished ...
23:01:05*TheSeven wonders why we aren't having any issues with audio DMA
23:01:28domonokynotlistening: i saw it, but didnt find time to work on this. thanks for testing.
23:02:20notlisteningok, they are little things really, it work really well under linux but no mac to test on :(
23:06:19linuxstbbluebrother: Seems to work nicely ;) Should I commit my rbutil.ini changes, or wait until everything else is in place/
23:08:21bluebrotherlinuxstb: I don't see a problem with committing it now.
23:09:08linuxstbbluebrother: OK. I think I'll wait and commit with my ipodpatcher changes though.
23:11:03kugelI'll do another approach
23:11:48notlisteningkugel, thanks for sorting the scroll wheel for me and everyone ;)
23:12:02kugelwell, I caused the problems :p
23:13:34TheSevenmanaged to do NAND transfers without cache spoiler bits :-)
23:14:21 Quit kkurbjunW (Remote closed the connection)
23:14:26TheSevenhowever, I'll still need to copy everything to prevent alignment problems
23:14:40 Quit JdGordon ("Leaving.")
23:15:21 Join kkurbjunW [0] (n=karlk@
23:15:29TheSevenforcing the alignment upwards on everything won't be possible, huh?
23:16:07 Quit kkurbjunW (Remote closed the connection)
23:17:00saratogacan't you just copy the start of the buffer's unaligned bytes, then DMA the remainder
23:17:09 Join kkurbjunW [0] (n=karlk@
23:17:42TheSevensaratoga: DMA will need things to be continuous
23:18:27saratogayou can't copy via load/store instruction?
23:18:35saratogafor the first couple bytes
23:18:47amiconnNothing prevents you from setting up multiple dma transfers either
23:19:26amiconnI.e. copy & dma the first block, then transfer the main block (and copy the remainder in paralle), then transfer the remainder
23:19:36TheSevenamiconn: We can maybe do this with things like UART, but not with AHB masters like NAND, ECC, Crypto, USB, ...
23:20:32saratogathen copy the start to an aligned buffer, DMA 3 bytes or whatever, and then DMA the remainder of the original buffer
23:20:51saratogaand end I guess
23:20:57TheSevensaratoga: I can't split the transfer
23:21:11saratogawhat is this for?
23:21:11 Join JdGordon| [0] (n=Miranda@rockbox/developer/JdGordon)
23:21:29 Quit petur (Remote closed the connection)
23:22:00 Quit merbanan (Read error: 110 (Connection timed out))
23:22:05TheSeventhese are peripherals that are bus masters, which will organize the transfers themselves. i can't even influence it.
23:22:25TheSeveni just write the address of the data to one reg, a start command to another one, and they do the rest
23:23:16amiconnAnd what stops you to do this more than once, using an isr that's called when each partial transfer ends?
23:23:46*amiconn thinks that you have to specify the length of the transfer as well somewhere
23:24:40TheSevenyes, but this will end up influence the actual length of packets on the e.g. USB bus
23:24:49TheSevenand with ECC, that can't work at all, as I'm not even specifying a length
23:25:30*amiconn wonders whether this newer SoC is really that unflexible
23:25:49TheSevenhey, bus master DMA transfers are a great thing!
23:26:23TheSeventhe problem is just that we don't manage to enforce some alignment constraints
23:27:34TheSevenin fact, it will probably always work nevertheless
23:27:46amiconnDMA can be very useful, yes. But I fail to see why the peripheral transfer sizes need to be linked to the dma block size
23:28:51TheSevenbecause the peripherals are organizing all that crap themselves and fetch the data when they need them automatically instead of having the CPU telling them about each and every bit to transfer
23:29:02 Quit JdGordon| ("Miranda IM! Smaller, Faster, Easier.")
23:29:17 Quit Strife89 ("My number of files is OVER 9000!")
23:29:34TheSevenare there realistic cases where stuff immediately beneath (within 12 bytes) of a DMA buffer will be touched during the transfer?
23:29:43amiconnThey don't do this in a magical way, but they talk to the built-in dma controller. This is basically just a speed control
23:30:18TheSeventhey are DMA controllers itself
23:30:52saratogasomeday i figure out how to use our mailing list without getting my posts garbled
23:30:55saratogabut that day is not today
23:33:36kkurbjunWTheSeven: They are just bus masters correct? A bus master does not necessarily have to be linked to a DMA block size or bus burst size - they could do single transfers if they were intelligent enough. At least with the systems we create.
23:34:13 Join DerPapst [0] (
23:34:37TheSevenyes, they are just AHB masters, and I think they can do that.
23:35:06TheSeventhe problem is just that the CPU caches may trash some things if we touch data immediately beneath unaligned DMA buffers during the transfer
23:38:04 Quit bmbl ("Bye!")
23:39:20tomersrasher: Does the bar line selector should cover the entire screen? Including the icon?
23:39:50kugelwps_reset() is highly annoying
23:40:02rashertomers: It does when using ltr. Also, it seems to just randomly extend past "some amount" of the screen
23:40:39*TheSeven will just forget about alignment and have a look if all kind of weird worms will start creeping around
23:41:03kkurbjunWahh, interesting,I guess even if you did a flush and something in the cache line changes and the line is flushed later it would cause problems - I guess the cache doesn't keep track of what is dirty and write just what has changed
23:41:22TheSevenkkurbjunW: that's what i suspect
23:42:06 Quit Djeepp ("Konversation terminated!")
23:42:12tomersrasher: Can you please clarify? I can't reproduce on e200. Seems ok both RTL and LTR
23:43:53rasherTaken with an e200 sim
23:44:56rasherit should extend all the way to the left
23:46:10tomersrasher: I can't reproduce. What is the theme/font? Is it a clean build (just in case)?
23:46:23rashertomers: Unicatcher theme, clean build
23:47:00 Join Lss [0] (
23:47:23tomersrasher: Font?
23:47:38rasherunifont - the one selected by UniCatcher
23:48:56tomers /me Compiling clean build and testing on e200 target
23:49:58 Join Thundercloud [0] (

Previous day | Next day