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

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

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

#rockbox log for 2010-03-09

00:02:42domonokyi think rbutil fails on 10.4 ppc because of the mac tts acting up somehow. and the 10.4 intel rbspeex just uses the wrong sdk. that leaves us with noise with rbspeex on 10.4 ppc and intel
00:06:43pixelmajust to clarify - the TTS itself works fine there. So it can "only" be the integration in RbUtil (or another sdk issue). I'd love to test a different TTS but I'd need to find a binary that works on 10.4 PPC and gave up searching earlier (haven't tried too long though). If someone has a link I'd be thankful, I really don't want to compile anything on those Macs
00:08:41pixelmaan espeak binary or similar that is
00:09:35Stephen__could a themesite admin delete my basicblueart and basicblackart the original author has his versions up aswell and it seems a bit silly having both up. and as he is the original author his should stay.
00:09:38domonokybluebrother mentioned that the tts interface is somehow different in 10.4 so i think its that what causes the rbutil hangs on 10.4 ppc.
00:11:46 Quit CaptainKewllllll (Quit: Page closed)
00:12:46domonokyStephen__: which targe/lcdsize have these themes ?
00:13:03Stephen__e200 thanks domonoky
00:16:24 Quit MaadMan (Ping timeout: 245 seconds)
00:17:01CIA-5New commit by kugel (r25078): Fix a bunch of files so that they build with newer ld. The stack/bss sections need NOLOAD. otherwise the linker would try to include those ...
00:17:05CIA-5New commit by kugel (r25079): Use ldr instead of adr, so that it compiles with a newer ld, although setting lr at that point hardly matters anyway.
00:17:37 Quit stripwax (Quit:
00:18:57 Quit ender` (Quit: Why do people give each other flowers? To celebrate various important occasions, they're killing living creatures? Why restrict it to plants? "Sweetheart, let's make up. Have this deceased squirrel.")
00:23:28 Join fml [0] (
00:23:53fmlpixelma: hello. Could you please test FS #11088 ? I can't build at the moment.
00:27:39 Quit karashata (Quit: The fluffy dragon has left completely!)
00:28:09domonokyStephen__: sorry, admin interface is somehow acting up, will do that tomorrow or so. please ping me if i forget it :-)
00:28:38Stephen__will do there's no rush anyway domonoky thanks again
00:28:41 Nick S_a_i_n_t_ is now known as S_a_i_n_t (S_a_i_n_t@
00:29:12*TheSeven spotted an FTL bug!
00:29:32TheSevennow *that*'s was caused nandfsck to be that slow
00:29:38S_a_i_n_tan FTL bug, or *the* FTL bug?
00:30:05TheSevennothing critical, just causing a lot of overhead during recovery
00:30:40TheSevenbut it's the bug i have been wondering about where it could be coming from since several days, which i just spotted by accident
00:32:27fml...or anybody who is able to build the manual: please try FS #11088
00:32:47CIA-5New commit by kugel (r25080): Fix red.
00:34:49 Quit fml (Quit: CGI:IRC)
00:36:34 Join karashata [0] (
00:37:47 Quit domonoky (Read error: Connection reset by peer)
00:39:47 Quit Zarggg (Ping timeout: 268 seconds)
00:40:58 Join Zarggg [0] (
00:42:33 Join yelped [0] (
00:43:13CIA-5New commit by kugel (r25081): Make mini2440 compile again.
00:43:24yelpedAnyone wants to check this out, please?
00:44:20 Quit yelped (Client Quit)
00:46:28S_a_i_n_twhat the hells the point in that? come in and say "can you look at this" (and several people have replied to his thread anyway), and then leave before anyone can answer?
00:46:47 Quit jgarvey (Quit: Leaving)
00:46:56 Quit flydutch (Quit: /* empty */)
00:54:22TheSeventhe answer would have been that nagging won't really help
00:56:56CIA-5New commit by kugel (r25082): Fix two warnings that appear with gcc4.4.3. The one in recording.c ("compact_view is used uninitialized" seems to be false positive unfortunately).
00:58:34***Saving seen data "./dancer.seen"
00:59:35 Join CaptainKewllllll [0] (~quassel@
01:00:02 Quit CaptainKewllllll (Remote host closed the connection)
01:01:33funmanre: r25082 : does gcc warn about dead code by default ?
01:02:03funmankugel: when trying the same diff than you, gcc wouldn't give a warning, so it might ignore dead return statements
01:03:04kugelfunman: I think it doesn't see that it's dead code
01:04:07kugelexit is a macro which expands to a differently called function. maybe the new gcc doesn't get trapped by the pure "exit" anymore. or it's more stupid if you like
01:05:38kugelI think 4.4.x has a new feature, __attribute__((never_reached)), maybe they gave up some detection in favor of letting devs decide themselve
01:06:20kugeloh, that was 4.5
01:06:47kugeland it was __builtin_unreachable()
01:07:04funmankugel: when disassembling the function (for arm), the return was removed (so gcc saw it was dead code)
01:07:40funmaniiuc, longjmp() was marked as noreturn, perhaps based only on the function name
01:07:51kugelthen I guess the warning sensitivity has changed
01:08:17kugelI think it's good if it warns and still optimises it away
01:09:25funmanah, longjmp (in setjmp_arm.S) doesn't use bx lr but rather the opcode for it, so perhaps gas sees it as no return
01:11:00kugeldo you think gas removes the return?
01:11:19kugeldid you check gcc -S as well?
01:11:33funmanno i didn't
01:11:49funmando we support armv4 non -T CPUs ?
01:12:12 Quit Schmogel (Quit: Miranda IM! Smaller, Faster, Easier.
01:12:32kugelwell, there's no target. I assume we would support them as we don't do thumb anyway :9
01:12:36funmanthen setjmp_arm.S could perhaps be changed (and an extra warning introduced?)
01:12:36saratogai think pre -T CPUs were well before mp3 players
01:12:55kugelfunman: that file is imported, we rather avoid to change imported code
01:13:36funmankugel: only if it harms syncing. one-liner doesn't, and I doubt we would need to sync it in the future considering the small size
01:13:53kugelwhat would you like to change?
01:14:28funmanRET => bx lr
01:14:39funmanit would save 8 bytes of binsize !
01:14:40 Quit planetbeing (Remote host closed the connection)
01:14:45 Nick planetbeing__ is now known as planetbeing (~planetbei@
01:14:53 Join planetbeing_ [0] (~planetbei@
01:17:56 Quit liar (Quit: Verlassend)
01:17:57 Join n1s [0] (~n1s@rockbox/developer/n1s)
01:21:29funmanhm os_exit is noreturn, but not longjmp
01:22:26CIA-5New commit by funman (r25083): Use bx lr unconditionally in setjmp_arm.S ...
01:22:54kugelawesome #ifdef :D
01:23:21funmanself descriptive ^^
01:24:55funman(this change didn't add a warning with gcc 4.4.2, although configure warned me about not using 4.4.3)
01:25:16funman+ using arm-elf-eabi-gcc
01:25:50 Quit DerPapst (Quit: Leaving.)
01:27:05 Part toffe82
01:28:55kugelthe rating of the build system is strange
01:29:29kugelprevious round I had an avg of 130, and a round speed of 170. this round I start with an avg of 31
01:29:44kugeland get round speed of 312 because I only get bootloader :\
01:32:06 Quit karashata (Quit: The fluffy dragon has left completely!)
01:42:13 Join karashata [0] (
01:43:38 Quit funman (Quit: free(random());)
01:51:34 Quit guymann (Quit: update time!)
01:57:20 Join Strife89 [0] (
01:59:37 Quit iq (Remote host closed the connection)
02:02:56 Quit Strife89 (Ping timeout: 248 seconds)
02:04:58 Join Strife89 [0] (
02:10:09 Join CaptainKewl [0] (~jason@
02:13:33 Quit Stephen__ (Quit: Leaving)
02:17:19 Join noob1 [0] (
02:18:44 Quit noob1 (Client Quit)
02:18:52 Join noob1 [0] (
02:19:20n1spixelma: still awake?
02:19:44 Join planetbeing__ [0] (~planetbei@
02:20:16 Quit froggyman (Ping timeout: 265 seconds)
02:21:50 Quit planetbeing_ (Ping timeout: 276 seconds)
02:22:05 Join Noob123456 [0] (
02:23:54 Quit noob1 (Quit: CGI:IRC (Ping timeout))
02:24:58 Join webguest113 [0] (
02:26:04 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
02:27:04 Quit Noob123456 (Quit: CGI:IRC (Ping timeout))
02:29:38kugeln1s: seen my mail? You could save me a lot of logs downloading and grepping :)
02:29:56 Quit webguest113 (Quit: CGI:IRC (Ping timeout))
02:33:35n1skugel: no, i hadn't, writing reply
02:34:14 Join noob1 [0] (
02:37:55 Join yelped [0] (
02:38:24yelpedSorry for leaving before so suddenly, had to go.
02:39:50noob1Question for anyone that knows: WPS function: Next/Prev on Ipod 5th Gen is supposed to go to next track and pervious track according to the manual, but mine just fast tracks and doesn't go to next song. How to I fix that.
02:40:14 Quit MethoS- (Remote host closed the connection)
02:41:10n1snoob1: are you just pressing them a short while or holding them in?
02:41:12 Quit Sajber^ (Read error: Connection reset by peer)
02:41:35noob1I understand holding it long should fast track
02:41:49noob1but pressing it quickly does nothing except fast track 1 second.
02:42:13n1snoob1: then you probably have the "track lock" (or w/e) it's called enabled
02:44:42saratogan1s: why are the test_codec results you posted so odd?
02:44:51saratogadoes test_codec not work correctly on the beast?
02:45:14n1ssaratoga: ? (they are decoding times)
02:45:26saratogaoh not MHz?
02:45:34n1syep, seconds
02:45:50n1smaybe should have mentioned that...
02:45:52saratogado you have the raw logs for svn anywhere?
02:45:56saratogai'll put them on the wiki
02:46:06n1sno, Unhelpful made the pdf
02:46:30n1sand all my logs are on my desktop on the other side of the world unfortunately
02:46:41saratogaok i'll bug him to post the logs or else update teh wiki himself (its easy now that we have a perl script to generate the wiki tables)
02:47:02 Join FlynDice [0] (
02:47:18n1si don't know how usefull it is to post these results now since most codecs have changed since then (and yeah the script is nice)
02:47:30 Quit yelped (Quit: CGI:IRC (Ping timeout))
02:50:29 Join planetbeing_ [0] (~planetbei@
02:50:53saratogan1s: it'll be a nice before and after for the mdct lib
02:51:02 Quit planetbeing_ (Remote host closed the connection)
02:51:17n1sah, yes
02:51:33noob1n1s: thanks
02:51:40n1show did the speed change over the old mdct on cf btw?
02:51:47noob1Found the option, "Prevent Skip playing"
02:51:49 Quit komputes (Ping timeout: 245 seconds)
02:51:54 Join planetbeing_ [0] (~planetbei@
02:52:17saratogan1s: it got marginally faster
02:52:30saratogabut thats mostly because we have no ASM at all for CF, not even for simple stuff like butterflies
02:52:30 Quit noob1 (Quit: CGI:IRC (EOF))
02:52:56n1sah, nice, there are a few places that stand out as good asm candidates :)
02:53:16saratogayeah everywhere theres arm asm :)
02:53:30saratogabut FFT8 is a pretty obvious place, FFT16 woudln't hurt either
02:53:45n1smainly gcc basically doesn't use movem so it does a lot of move.l stores which can stall etc
02:53:46saratogathe transform macros using EMAC would help a lot
02:53:54 Quit planetbeing__ (Ping timeout: 245 seconds)
02:54:09n1si think i'll look at it when i get home
02:54:20saratogacould probably shave 5-10MHz off Vorbis/WMA/AAC/ATRAC/Cook/AC3
02:54:34n1s*that* would be amazing
02:56:04 Quit planetbeing_ (Client Quit)
02:56:50saratogathats basically what we've done for arm over the last 2 years of playing with the MDCT :)
02:58:38***Saving seen data "./dancer.seen"
02:59:36*chrisb enjoying rockbox on sansa e200 now
02:59:59kugel"Unhelpful-O, but on my e200 flac was faster with -O2 than svn with -O2"
03:01:07 Quit karashata (Quit: The fluffy dragon has left completely!)
03:01:20kugeln1s: how do you enable -O2 in the code?
03:02:45n1skugel: edit the makefile for the codec
03:03:41n1syou can get an example from tremor for example
03:04:01saratogayou have to regex the make strings usually, but theres examples in various codecs like tremor
03:05:09kugelwill try. thanks
03:05:18 Join BiptoN [0] (
03:05:24*kugel now has a script to download the logs properly
03:06:56kugelif flac is faster with -O2 and the new compiler, there's no real reason to stick to 4.0.3, is there?
03:10:18 Quit BiptoN (Quit: CGI:IRC (Ping timeout))
03:10:33n1sif the codecs are faster overall and everything still works i don't think so
03:11:24 Join iq [0] (~iq@unaffiliated/iq)
03:14:37 Join planetbeing_ [0] (~planetbei@
03:15:19kugelwow, 704% rt, instad of 593%
03:17:20kugelflac_8 improved by 100 percentage points as well
03:20:40saratogastripwax: I took a shot at converting svn tremor over to our mdct, it compiles but the output is wrong, not sure why yet
03:20:46saratogaif you want to take a look
03:25:26saratogai have to get back to work so I can't debug it anymore right now
03:27:51kugelflac svn is also improved by -O2 but not as much
03:31:14saratogakugel: which target?
03:31:40kugelI can only test flac on PP, so..
03:36:17saratogakugel: maybe its worth commiting the change then?
03:37:14kugelI haven't verified the output, or coldfire
03:38:05n1siirc 02 made no or little difference on coldfire with gcc 3.4
03:38:33n1sprobably it was slightly slower or i would have committed the change already ;)
03:38:44n1sbut testing is of course nice
03:39:08kugeltremor does -O3 on cf and -O2 on the rest, so it can be arch dependant in the .make
03:39:16 Quit Adnyxo (Quit: Leaving)
03:39:29saratogayeah i wouldn't touch CF its been picked clean already
03:39:31 Join Connor [0] (
03:39:33ConnorHello, I have a question about rockbox on my 5.5G 80 gig iPod Video. Whenever I pick a song from my database, the iPod pops up a "searching for files" box, and it scrolls from 2700 to 0, and then plays the song. I have the database loaded to ram and dircache enabled. Any way to fix th
03:41:21n1ssaratoga, kugel a nice way to find the optimal O level is to do what Unhelpful did when he made those benchmarks: remove all the individual codec O flags and set it in the codecs.make and then do 3 test runs over the test set, at O1, O2 and O3 and just compare logs
03:41:36Connorn1s you ever hear of this bug?
03:41:43ConnorOr, what I believe to be a bug?
03:41:56n1snope, never used database
03:42:04Connorn1s what mp3 player are you using?
03:42:08saratogamaybe try searching the forums
03:42:12ConnorIt didn't help much.
03:42:14saratogai doubt many people up at this hour have used the database
03:42:20n1si have a h300, c200v1 and gigabeast
03:42:39Connorsaratoga what do you recommend doing?
03:42:47ConnorAside from not using the database.
03:43:07saratogaeither searching or asking during the day on IRC
03:43:07kugelwell if only a test_codec run would last an hour or so :(
03:43:09S_a_i_n_tJdGordon: Around?
03:43:18n1sConnor: so what problem does this cause?
03:43:33n1skugel: that's why the beast is nice :)
03:43:39Connorn1s a 10-20 second delay in a song starting to play.
03:43:49n1sah, it takes that long
03:44:11kugeln1s: you can hardly compare its performance to the PPs one can you ?
03:44:14ConnorI'll try it not saved to ram.
03:44:52kugelConnor: how big is the list you're selecting
03:45:01Connorkugel 2700 songs.
03:45:02n1skugel: no, you probably can't but a full test_codec run takes much less than an hour
03:45:11kugelit enqueues all songs in the current list, so it may take a while to fetch all and put them into a playlist
03:45:27 Quit JdGordon (Ping timeout: 260 seconds)
03:45:56ConnorHmm. How do I keep the songs from being put into a playlist?
03:46:02ConnorOr is this required?
03:46:02saratogaoh so you're creating a 2700 song playlist
03:46:09ConnorI guess?
03:46:11n1sah, yes it's basically a limitation in the design of the database/playlist stuff
03:46:31ConnorI just want to play my music without a 20 second delay of picking a song from the list of songs to the time the music actually starts.
03:46:31kugelConnor: browse something which gets you a smaller list
03:46:46n1sConnor: make you database searches return fewer results or i suppose long press the son and insert it
03:46:58ConnorSo there's nothing I can do?
03:47:16n1sConnor: <n1s> Connor: make you database searches return fewer results or i suppose long press the son and insert it
03:47:29kugeldon't choose 2700 song lists. I'm sure the database has enough filters which lead to smaller lists
03:47:55ConnorWell I usually have music on shuffle.
03:48:12kugelis that 2700 all of your songs?
03:48:42kugelI'm not sure why you need the database then
03:48:53saratogacouldn't you just make a single playlist with all your songs and shuffle that?
03:49:05saratogainstead of endlessly regenerating a new one from the database
03:49:13kugelyou could simply go to resume playback after boot without even entering the database. or, once you play all those songs save the playlist on disk
03:49:30ConnorHmm, how would I play my music without the database then?
03:49:58saratogaplay the playlist
03:50:19ConnorI am confused.
03:50:34 Join BiptoN [0] (
03:50:46saratogai think the problem is that you're trying to create a gigantic playlist everytime you use the database, which your slow cpu can't handle well
03:51:01saratogaso i'm saying just save the playlist to disk and play it instead of trying to regenerate it everytime
03:51:05Connorsaratoga I think that is the problem.
03:51:10ConnorAnd how would I go about saving the playlist?
03:51:30saratogai don't know the buttons on your player, but its in the manual
03:51:44saratogaor of course you could always do it on your PC from iTunes or whatever
03:52:17kugelConnor: in the main menu->playlists->save current playlist
03:52:24ConnorThank you.
03:52:32TheSevenhm, what is the charging time estimate based on?
03:52:39ConnorAnd all I have to do is select a song, have it build the playlist, and then save it?
03:52:41TheSevenis it assuming a linear charging curve?
03:53:05kugelConnor: that should work yes
03:53:11ConnorThank you
03:54:32TheSevenand is battery_level_dangerous just a warning level, or causing a shutdown?
03:54:45n1sTheSeven: i think it's done the same way as discharging but with a different curve
03:55:05kugelTheSeven: I think it shuts down at that level
03:55:16kugelbut I'm not sure
03:55:19TheSeventhe existence of battery_level_shutoff proposes the earlier, but my ipod did shut off at "dangerous" during the bench
03:55:43TheSevenn1s: the difference is that while discharging has a more-or-less constant current, the charging on li-ion is highly non-linear
03:56:16saratogaon most targets we have no access to current, so I assume its all based on voltage
03:56:29TheSeventhere is a voltage to percentage mapping for charging, but no mapping to charging time
03:56:32kugelI think the battery estimation is based on the current voltage and the CURRENT_* defines from the config file
03:56:47TheSevenare there current_* things for charging?
03:57:11kugelI think there is one
03:57:31TheSevenreading the current won't help either, what we really need is a mapping from percentage to remaining charging time
03:57:42kugelmaybe not
03:57:53TheSeven(and no, a constant doesn't suit that purpose at all)
03:58:34*TheSeven wonders where that charging time estimation code is
03:59:02n1sTheSeven: it is done the same way for both charging and discharging but with different curves mapping percent to voltages
03:59:51TheSevenn1s: the voltage to percent part is clean, i'm just saying that we would also need a percentage to charge completion time mapping
04:00:11n1sneed for what?
04:00:22TheSevento predict the charging time
04:00:37n1sah, yes
04:00:49TheSeven(and the "voltage while charging" mapping is somewhat superfluous if we have a battery that's calibrated well)
04:01:43n1si think that the time estimation is now just done by reading the voltage at an interval, and then extrapolating from percent charged over a window
04:02:05TheSevenwhich doesn't make any sense either for li-ion
04:02:44 Quit Strife89 (Quit: Bed.)
04:03:43TheSeven powermgmt_est_runningtime_min = (100 - level)*battery_capacity*60 / 100 / (CURRENT_MAX_CHG - runcurrent());
04:03:58n1sit's most likely a thing from the archos days
04:03:58TheSeventhat doesn't make any sense at all
04:05:02TheSevenit's assuming a 100% efficiency of the battery, a constant input current until it is 100% charged, and that the power consumption of the player will be deducted from that, which ist just *all wrong* for nano2g
04:05:50TheSevenon nn2g (if we allow it to draw 500mA) it will do a CC chage at 120mA until the voltage hits 4.32V and then do a CV charge at that level from that point on
04:06:22n1syeah, i was wrong, i though power_history list was used by apparently that is only used by the debug menu and the archos charging code
04:06:45TheSevenif we limit it to 100mA, the archos or whatever scheme will fit better, as during the CC phase, the input current will be limited to 100mA, so the battery current will be 100mA minus the player's current consumption
04:06:55n1sTheSeven: yeah, most li-ion chargers should be the same
04:07:21TheSeventhe funny thing is that there doesn't seem to be a charge completion event
04:07:35TheSevenit will keep the battery floating at 4.32V until you remove USB power
04:07:57TheSevenbut then, if you re-plug it, it won't start charging unless it's below some limit (in the range of 4.2V somewhere)
04:10:01n1staht's a bit weird
04:10:43TheSevenand i haven't yet foudn a GPIO that indicates whether it's actually charging, which is a bit nasty
04:17:30chrisbare links supported? MUSIC/Beethoven/Symphony5.ogg MUSIC/VonKarajan/Symphony5.ogg can these be links?
04:17:54S_a_i_n_tThe OF must know its charged it displays a fullscreen battery cell to indicate charging is finished.
04:17:57chrisbhard or soft links?
04:18:05 Quit planetbeing_ (Quit: planetbeing_)
04:18:19S_a_i_n_t*Nano2g that is.
04:18:34chrisbsaratoga: how do people handle this case?
04:18:48saratogai don't know what that case is
04:19:24chrisbone song, with two or more "parent" artists
04:20:11chrisbjam the info into id3 ?
04:20:18CIA-5New commit by nls (r25084): Simplify makefile
04:23:07CIA-5New commit by nls (r25085): Statically init array to 0 instead of runtime memset
04:25:20BiptoNdoes anyone know if the idea to port to Arhos 5 IMT has been discussed?
04:26:48Unhelpfulkugel: last i checked i still can't use mpegplayer with my eabi build, unsure if it's an eabi or 4.4 issue
04:27:04saratogaBiptoN: I don't think so
04:28:29kugeln1s: heh, I wasn't sure whether I don't understand the .make files enough or if that was really messed up :)
04:28:50kugelUnhelpful: hrm, I haven't checked actually
04:29:44n1syeah that makefile had 2 rules for making the same objects, i guess make selected the last one (which was correct)
04:33:27 Join CGL [0] (~CGL@
04:37:22 Quit BiptoN (Quit: CGI:IRC (EOF))
04:40:54 Quit TheSeven (Disconnected by services)
04:41:09 Join The_Seven [0] (~theseven@rockbox/developer/TheSeven)
04:41:19 Nick The_Seven is now known as TheSeven (~theseven@rockbox/developer/TheSeven)
04:42:36 Join planetbeing_ [0] (~planetbei@
04:45:11kugelUnhelpful: shit, indeed
04:45:22kugeldata abort. the disassembly looks fine though
04:45:56 Quit planetbeing (Ping timeout: 240 seconds)
04:45:59 Nick planetbeing_ is now known as planetbeing (~planetbei@
04:46:27kugelhave you further looked at it?
04:48:23 Quit kugel (Remote host closed the connection)
04:53:11 Join Barahir [0] (
04:56:41 Quit Barahir_ (Ping timeout: 256 seconds)
04:58:37 Quit panni_ (Read error: Connection reset by peer)
04:58:39***Saving seen data "./dancer.seen"
04:59:55 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
05:02:52saratogaUnhelpful: you should put your test codec results for the beast on the wiki
05:02:58saratogawe have no scores at all for it on the page
05:03:27saratogaif you have old pre-MDCT logs, even better, we could do before and after
05:04:16 Quit mt (Ping timeout: 240 seconds)
05:05:57mc2739kugel: (for the logs) the bootloader you sent me worked on the e200
05:08:58 Join mt [0] (~mtee@rockbox/developer/mt)
05:16:18Unhelpfulkugel: crashed? :D
05:20:09Unhelpfulshould probably check doom also, it and mpegplayer both had warnings related to short enums
05:23:00 Join demi` [0] (WarMinion@
05:25:17 Part demi`
05:27:13Unhelpfulalso might be a struct thing, if we have yuv->rgb asm that expects a struct rgb something like { u8 r; u8 g; u8 b;} then it's probably only 75% of its expected size
05:32:31 Quit Horscht (Quit: Verlassend)
05:32:54 Quit n1s (Quit: Lmnar)
05:45:08 Join planetbeing_ [0] (
05:51:36 Join n1s [0] (~n1s@rockbox/developer/n1s)
05:58:43 Join Rob2223 [0] (
06:02:36 Quit Rob2222 (Ping timeout: 265 seconds)
06:12:22CIA-5New commit by theseven (r25086): Nano2G battery calibration
06:23:50 Quit anewuser ()
06:30:41 Join FOAD_ [0] (
06:34:00 Quit FOAD (Ping timeout: 264 seconds)
06:34:00 Nick FOAD_ is now known as FOAD (
06:41:44 Join planetbeing__ [0] (
06:43:29 Quit planetbeing_ (Ping timeout: 248 seconds)
06:58:41***Saving seen data "./dancer.seen"
07:08:21 Join Codefest_ [0] (~dce36163@gateway/web/freenode/x-nnmiloyogbqgiito)
07:09:17 Quit Codefest_ (Quit: Page closed)
07:14:59 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
07:37:10 Join Codefest [0] (~dce36163@gateway/web/freenode/x-dknjfupnxbguftxh)
07:37:14CodefestWe invite you to be a part of the Largest Online Coding festival in the Indian Subcontinent.[ ] Sit right in front of your systems and take part in challenging and mind boggling competitions on application development, algorithm design etc. PS:Attractive Prizes.Adobe & BT certificates.
07:37:42CodefestWe invite you to be a part of the Largest Online Coding festival in the Indian Subcontinent.[ ] Sit right in front of your systems and take part in challenging and mind boggling competitions on application development, algorithm design etc. PS:Attractive Prizes.Adobe & BT certificates.
07:37:47 Part Codefest
07:39:08 Quit CaptainKewl (Remote host closed the connection)
07:41:03 Join planetbeing_ [0] (
07:44:54 Quit planetbeing__ (Ping timeout: 276 seconds)
07:49:04 Join stoffel [0] (
07:57:37 Quit JdGordon (Ping timeout: 248 seconds)
07:57:44 Quit CGL (Quit: Saliendo)
07:57:53 Quit n1s (Ping timeout: 258 seconds)
08:01:10*S_a_i_n_t shakes his fist at stupid "spam & leave" guy.....Grrrrrrrrrrr!!!
08:09:32 Join Xerion_ [0] (
08:11:35 Quit Xerion (Ping timeout: 240 seconds)
08:11:35 Nick Xerion_ is now known as Xerion (
08:12:28 Join Zagor [0] (
08:12:28 Quit Zagor (Changing host)
08:12:28 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
08:13:48 Quit chrisb (Ping timeout: 245 seconds)
08:19:17 Join LinusN [0] (~linus@rockbox/developer/LinusN)
08:21:32 Join flydutch [0] (
08:22:53 Join ender` [0] (
08:28:36 Quit mc2739 (Ping timeout: 256 seconds)
08:29:16 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
08:50:34 Join einhirn [0] (
08:53:43 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
08:56:07 Quit Zagor (Quit: Clint excited)
08:58:05 Quit JdGordon (Ping timeout: 245 seconds)
08:58:42***Saving seen data "./dancer.seen"
09:07:24 Quit planetbeing_ (Quit: planetbeing_)
09:08:51 Join petur [0] (~petur@
09:08:57 Quit petur (Changing host)
09:08:57 Join petur [0] (~petur@rockbox/developer/petur)
09:13:12 Part Speedy2 ("Konversation terminated!")
09:14:14 Join stripwax [0] (
09:14:55 Quit stripwax (Client Quit)
09:18:39 Join gill0r_ [0] (
09:21:41 Quit gill0r (Ping timeout: 252 seconds)
09:22:38 Join FOAD_ [0] (
09:25:36 Join liar [0] (
09:26:05 Quit FOAD (Ping timeout: 252 seconds)
09:26:05 Nick FOAD_ is now known as FOAD (
09:37:58 Join karashata [0] (
09:40:18 Join pamaury [0] (
09:42:06 Quit liar (Read error: Connection reset by peer)
09:57:37 Join liar [0] (
10:05:29 Join Zagor [0] (
10:05:30 Quit Zagor (Changing host)
10:05:30 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
10:09:08 Quit lyngaas (Ping timeout: 240 seconds)
10:09:45 Quit planetbeing (Remote host closed the connection)
10:11:39 Quit shai (Quit: Leaving)
10:12:00 Join shai [0] (
10:14:44 Join DerPapst [0] (
10:19:46 Join efyx_ [0] (
10:20:12 Join planetbeing [0] (
10:48:05 Quit liar (Read error: Connection reset by peer)
10:51:27 Quit BHSPitMonkey (Remote host closed the connection)
10:58:45***Saving seen data "./dancer.seen"
11:04:06 Join DataGhost [0] (
11:04:07 Quit DataGhost (Changing host)
11:04:07 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
11:07:25*petur wonders if somebody already applied for participation in GSoC
11:10:52linuxstbHave we? i.e. has someone filled in the paperwork?
11:11:13peturwhich is what I'm asking ;)
11:11:52linuxstbAh, I thought you meant students...
11:12:08*linuxstb pings scorche|sh and Bagder
11:35:43 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:43:13 Join lyngaas [0] (
11:45:26 Join liar [0] (
11:47:53n1sdoes any coldfire wizard here know how the syntax for trapf.{wl} is?
11:59:10 Quit shai (Quit: Leaving)
12:03:12 Join Sajber^ [0] (~Sajber^
12:03:46 Join dfkt [0] (dfkt@unaffiliated/dfkt)
12:04:24 Quit Kitar|st (Ping timeout: 260 seconds)
12:05:31n1sok, using it with a literal value works but that is not very nice,
12:06:03 Join Kitar|st [0] (
12:18:05 Join Marctraider [0] (
12:18:24MarctraiderHello and goodday everybody.
12:21:37MarctraiderJust tried installing iloader on my iPod Nano Black 2G yesterday but I kinda screwed up afterwards. Not totally but I cannot seem to get back to my old state. But, it did worked. So you can add me to the list of 7 tested devices ;)
12:24:19S_a_i_n_tDo you know what you did wrong?
12:24:35 Join shai [0] (
12:24:35MarctraiderI did make back-ups of the files appleos.bin and norflash.bak but i cannot seem to recover them. Tried to write them back with ipodpatcher for nano 2G with all different parameters and it did write to the firmware partition successfully but nothing changes. iloader keeps appearing at the ipod. Also uninstalling the bootloader succeeded and when repeating it it said bootloader already uninstalled, but iloader keeps appearing..
12:25:05Marctraideralso dd'ing random data in linux to /dev/sdd and sdd1/sdd2 every possible way. Nothing changes :(
12:25:37Marctraiderperhaps someone knows how this is happening or has the same problem
12:26:04S_a_i_n_tyou need to flash norflash bask to nor, but you'll need to ask TheSeven about how exactly to do that.
12:26:16MarctraiderOhh ok
12:26:59Marctraideryes well.. the ipod starts with a recoveryscreen 'norloader' and quicky says to switch the hold switch and i do and then ibugger pops up
12:27:20Marctraiderbut this norloader screen only appears for 10ms or so haha
12:27:42MarctraiderBut I think i'll just have to wait for the TheSeven then.
12:27:49S_a_i_n_tat the present state (with norloader installed) even if the apple OS is in the firmware partition, and the bootloader is installed to OSOS, the device isn't really aware of this...
12:28:06S_a_i_n_tas the norloader takes no notice of them
12:29:04S_a_i_n_tbut it does sound very wrong, it will probably flash up quickly like that if the iLoader directory is not on the disk
12:29:12S_a_i_n_tor if it cannot find the right files.
12:31:44S_a_i_n_tTheSeven will be asleep at the moment, but I'm sure he won't mind if you PM him and let him know the state you've got yourself into. He wrote the installer, iLoader, iBugger...and a good deal of code for the he really is the authority on this.
12:31:49 Quit stoffel (Remote host closed the connection)
12:33:37 Join hebz0rl [0] (
12:40:14 Quit DataGhost (Ping timeout: 252 seconds)
12:47:23liarMarctraider: if you use norloader iloader will be in the NOR flash. if you write random stuff to the NAND(/dev/sd?) that will not stop iloader from starting
12:50:09MarctraiderS_a_i_n_t: thanks for the help already I will pm him later today, i've read everything you wrote
12:50:26Marctraiderliar: ok
12:50:55Marctraiderbut norloader popups just 10ms when resetting ipod and instantly goes to iloader
12:51:16Marctraiderin that 10ms it tells to switch the hold button. When i do it starts ibugger.
12:51:21liaryes, thats normal
12:53:32liari dont understand the problem, do you want to uninstall norloader?
12:53:36MarctraiderI DID manage to change one thing... namely that I got another screen with 3 options; telling there is a flash error or something. Then it gives me three options: Disk mode, ibugger, and something else. but this is also still part of the iloader.. I find it strange if ipodpatcher tells me it uninstalled iloader it still loads. But i'm not that into this stuff so i can be very wrong ofcourse.
12:54:21*TheSeven needs to catch up
12:54:26Marctraiderliar: The prob is i want to recover my ipod and i have 2 back-up files norflash.bak and appleos.bin.
12:54:44S_a_i_n_there we go :D
12:55:01*S_a_i_n_t thought TheSeven was still asleep... ;)
12:55:04MarctraiderTheSeven: Can i contact you later today for my problem? (I dont blame you ofcourse it was totally my own risk :)
12:55:26MarctraiderBut I need to take the children out of school
12:56:01MarctraiderI'll be back as soon as i Can, please read what i've posted here TheSeven, see you later today perhaps.
12:56:25TheSevenS_a_i_n_t: I was until a minute ago ;-)
12:56:36S_a_i_n_thehe :D
12:57:02TheSevenhehe, I didn't build an uninstaller yet, but I can quickly make one
12:57:12Marctraideroh! ok...
12:57:25Marctraiderbe back later ;)
12:57:43TheSeven(you could always go back using ibugger and genflasher, but i'll prepare an automated way)
12:58:43 Join Adnyxo [0] (
12:58:48***Saving seen data "./dancer.seen"
13:02:43 Join roolku [0] (
13:04:48roolkuZagor: did you get my email regarding the build system problems?
13:05:34 Join perfectdrug_ [0] (
13:06:24 Quit karashata (Quit: The fluffy dragon has left completely!)
13:09:32 Quit perfectdrug (Ping timeout: 258 seconds)
13:10:16 Join krazykit` [0] (
13:12:57 Quit moos (Ping timeout: 264 seconds)
13:13:34amiconnn1s: What are you trying to do with trapf?
13:13:38 Quit krazykit (Ping timeout: 240 seconds)
13:14:25amiconnThis instruction is meant for avoiding an extra branch in single instruction alternatives.
13:14:33 Join moos [0] (moos@rockbox/staff/moos)
13:14:46n1samiconn: according to the programmers manual it can be used to jump forward 2 or 4 bytes so yeah i try to hump past one instruction to avoid doing a bra
13:15:02n1seh s/hump/jump/
13:15:06 Quit Marctraider (Ping timeout: 276 seconds)
13:15:08amiconnAs such, you cannot use it as a mnemonic in the assembler (except the operand-less form, which is essentially a single cycle nop)
13:15:46n1syeah, that's what i gathered :(
13:15:52amiconnIt's ugly, but I know of no other way than to use .word
13:15:57 Join Marctraider [0] (
13:16:34n1sdoesn't work well with inline asm unless you hardcode the registers, and a single cycle save is probably not worth it
13:17:14n1si wonder why the assembler can't handle it though
13:17:54amiconnThe assembler can't handle it because you effectively turn the next one or two instructions (or one insn plus one extension word) into data
13:19:28amiconnI used this in two places: The remote lcd driver for the irivers, and the bit-banging i2c for the pcf50606 on H300, X5 and M5
13:20:06amiconnLibspeex also uses this in the filters on coldfire
13:21:00amiconnOh, and the dsp code
13:22:54n1sso that's why grep gave no hits
13:24:21amiconnIf you grep for 'trapf' you'll find those places, because they say in a comment what this .word means
13:24:40amiconn(apart from the one in dsp_cf, since that says 'tpf')
13:26:25n1sah, i forgot the -r swithch, duh
13:27:20 Quit moos (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
13:27:33 Join moos [0] (moos@rockbox/staff/moos)
13:29:13n1saha, you only have to .word encode the trapf instr not the following ones, that makes it nicer
13:35:32n1samiconn: thanks, i had misunderstood it a bit and thought you'd have to use .word for the instructions following the trapf
13:44:41S_a_i_n_tTheSeven: I think I may have found a reason as to why the Nano1g is so much slower than the Nano2g...can you take a look at this please?
13:45:33S_a_i_n_tIt seems a little odd to me, but I don;t know much about the CPU's of either target, or the similarities and differences.
13:45:59TheSeventhat looks sane, as it specifies the *default* freq
13:46:19TheSeven(which is the highest possible on nano2g, and the lowest possible on nano1g for some reason)
13:47:08TheSevenbut soon after boot, the boosting code will take over and adjust that dynamically (between 48 and 192MHz on the 2g, and between 24 and whatever on the 1g)
13:47:52linuxstbS_a_i_n_t: The Nano1g has an 80MHz CPU - I would expect it to be slower...
13:48:48TheSevenlinuxstb: we're talking about a ~factor 50 slowdown during disk accesses or whatever, making the pictureflow init take a minute on the 1g and a second on the 2g
13:49:01TheSevenS_a_i_n_t: erm, regarding your symmetry thing, it's totally unusable for me
13:49:10TheSevensomething seems to have went wrong with it ;-)
13:49:26S_a_i_n_theh...why's that?
13:49:53linuxstbIs that problem evident just in pictureflow?
13:50:20TheSevenit seems to assume a wrong screen size, scrollbars, icons and the backdrop are missing, and the wps is some brown unreadable mess that looks roughly like the default wps, just in weird colors
13:50:27amiconnTheSeven: The default freq is usually the one that can be used with the pll disabled
13:50:38S_a_i_n_tas long as you have the right fonts, it should run sweet. ;/
13:50:45amiconnIt is not used during normal operation, apart from the radio screen when not recording
13:50:45TheSevenamiconn: that would be 32kHz on the 2g then :-P
13:50:50S_a_i_n_tthat's really fucking screwed up...
13:51:28amiconnThe frequency used during operation is the *normal* frequency, unless the cpu can't cope, in which case we switch to the *max* frequency (boost)
13:51:35TheSevenskin buffer: 57.8K used
13:52:01TheSevenwhere is cpu_freq used at all then?
13:52:25 Join Lss [0] (
13:52:56TheSeven#define CPU_FREQ 191692800 to be exact
13:55:53TheSevenS_a_i_n_t: oh nice. rebooted it, and now i just get a black screen
13:56:17TheSevenand that's with a build without the ftl changes
13:56:43TheSeveni read the files back from the ipod and compared, they're fine
13:58:02S_a_i_n_tSomething *must* be fucked up on your end...or *perhaps* its currepted somehow on the RB server...but I doubt it.
13:58:23S_a_i_n_tI have the same theme running on 2 1g's and 2 2g's here at home.
13:58:40TheSevenif i just load the wps, i'll get the default wps instead
13:58:52amiconnTheSeven: It is mostly used in a target specific way
13:58:58S_a_i_n_tand it has been check out by perfectdrug on a sim he built...and reported it worked fine ;/
13:59:39amiconnOn targets without boosting it's also used to define the operating frequency in general
14:00:19TheSevenamiconn: I set it to the operating frequency during boot (which is also the PLL speed, boosting is just done through divisors)
14:00:37amiconnCPU_FREQ is a fixed value
14:01:20amiconncpu_frequency is reflecting the current value, and gets initialised to CPU_FREQ
14:01:26TheSevenwrong evaluation order ;-) (I set it to the (operating frequency during boot ))
14:03:05amiconnSo you can use CPU_FREQ on Nano G2 as you want. Having a CPUFREQ_DEFAULT that's fairly low would probably make sense though
14:03:31amiconn(although I don't think there are many occasions when this frequency is chosen - Nano G2 doesn't have a radio, does it?)
14:03:55TheSevenonly through accessories
14:04:05TheSeven(which aren't implemented yet)
14:05:11*amiconn thinks he found the bug that caused the non-working battery bench
14:05:53amiconnThere is a pair of functions to hook into a "disk is spinning" status. register_storage_idle_func() and unregister_storage_idle_func()
14:06:23amiconnThe latter can optionally call the registered function (a last time, directly), and battery_bench uses that
14:06:43amiconnBut if the target in question doesn't implement storage callbacks, that doesn't work
14:14:19 Join robin0800 [0] (
14:14:20 Quit roolku ()
14:15:25S_a_i_n_tCan I do an appeal to iPod Nano 1st / 2nd gen owners to test this theme *please* : it is working on four of my target devices (two nano1g's and two nano2g's) but is apparently failing in spectacular fasion on TheSeven's Nano2g :-/
14:17:00 Join anewuser [0] (anewuser@unaffiliated/anewuser)
14:17:02 Join kugel [0] (~kugel@rockbox/developer/kugel)
14:21:45 Join Schmogel [0] (
14:24:20 Quit anewuser (Ping timeout: 260 seconds)
14:25:59 Join anewuser [0] (anewuser@unaffiliated/anewuser)
14:26:23 Quit Marctraider (Ping timeout: 252 seconds)
14:26:34 Join Marctraider [0] (
14:26:55 Join panni_ [0] (
14:30:09amiconnTheSeven: In order to make idle notification work you need to implement the call_storage_idle_notifys(false) call in the nand thread (compare to how other storage drivers do this), and enable it for nano G2 in ata_idle_notify.h
14:30:40amiconnIt's nevertheless a bug in battery_bench that it doesn't work without idle notification
14:31:33 Join froggymana [0] (
14:39:39 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
14:53:43 Join FOAD_ [0] (
14:54:01 Join CGL [0] (~CGL@
14:57:33 Quit FOAD (Ping timeout: 245 seconds)
14:57:33 Nick FOAD_ is now known as FOAD (
14:58:51***Saving seen data "./dancer.seen"
15:00:52MarctraiderTheSeven: thanks i'll check it out
15:00:56Marctraiderand report back
15:01:17Marctraideri thought with ipodpatcher you could uninstall the bootloader and everything was fine, but i thought wrong i guess ;)
15:01:43TheSeventhat only applies to the old osos bootloaders, not to the nor ones
15:02:27TheSevenyou actually need to *reinstall* the old apple bootloader, overwriting iloader, and the place where it is stored can't be accessed directly from a pc
15:06:31Marctraidermhh ok
15:06:43Marctraiderim trying the uninstaller now as we speak
15:07:22MarctraiderI assume one has a problem if one doesnt have norflash.bak backupped?
15:08:24Marctraiderok. Well thats not the case dont worry, but i just thought about it
15:08:42S_a_i_n_tyou freaked me out there...
15:08:43TheSevenwell, *in theory* one could build it from several pieces scattered around in apple firmware images and the iloader flash, but that won't be fun and requires some more advanced tools and knowledge about how this stuff works
15:09:01S_a_i_n_tits not *so* bad, but you could never have a "stock" ipod again.
15:09:09S_a_i_n_t*if that were the case that is.
15:09:10MarctraiderAnd i have one more question. The actual reason why i wanted to try this is because i wanted to edit the apple firmware to get rid of that stupid EU volume limit.
15:09:20Marctraideri see
15:09:21 Quit moos (Quit: ChatZilla 0.9.86 [Firefox 3.6/20100115144158])
15:09:47MarctraiderI have edited an appleos.bin to remove the EU volume cap, can i use this one to replace my original?
15:09:50S_a_i_n_tit is easy to "uncap" the ipod
15:10:00TheSevenyou can also do that without iloader, but it will require decrypting, patching and re-encrypting the firmware
15:10:02S_a_i_n_tTheSeven could point you to a link I believe
15:10:24TheSeven(which is well possible, just a bit of work)
15:10:28Marctraiderwell ill try to uninstall first, uncapping comes later i guess
15:10:57TheSevenuninstall = copy 2 files and reboot it, then restore using itunes
15:10:58*S_a_i_n_t lives in has iPods that can deafen him :D
15:11:21MarctraiderTheSeven: aight!
15:11:25*TheSeven didn't notice *much* difference in the output levels whether it's capped or not
15:11:54TheSevenand afterwards you can patch the firmware using ibugger, ipodcrypt and ipodpatcher
15:11:55Marctraiderdo i need to rename it to iLoader.conf or .cfg?
15:12:08TheSevendid i misspell it somewhere?
15:12:18Marctraideri think so
15:12:29Marctraidercfg vs conf
15:12:33TheSevenok, fixed
15:12:39Marctraiderno serious thing though ;)
15:12:45Marctraideri was just confused for a sec
15:12:56TheSeventhat's why it needed to be fixed ;-)
15:13:05S_a_i_n_tAnd with it uncapped, or with Rockbox which the "capping" is unaffected bym an iPod Nano 1st/2nd gen is plenty capable of destrying a cheap pair of IEM's
15:13:30S_a_i_n_teven a halfway decent pair...if you take it up past 0dB
15:14:00TheSevenmy nano won't go above -1dB with 16 Ohms headphones
15:14:30S_a_i_n_t*wont* go, or sounds shit if you go past?
15:14:41Marctraiderdoes the iloader screendissapear in some time or do i now just have to restore with itunes?
15:14:56TheSevenMarctraider: which screen exactly?
15:14:57Marctraiderok the screen goes of but nothing happens
15:15:06*S_a_i_n_t got himself some sennheiser IE8's
15:15:11Marctraiderplease press play and extract an iloader theme etc etc. Iloader free your nano!
15:15:21TheSevenhm, it didn't find the iloader.cfg file then
15:15:27TheSevenpress play and fix it
15:15:52S_a_i_n_tI can take both my 1 & 2 g's up to -6dB...but it sounds pretty dodgy past -3
15:16:07TheSevenS_a_i_n_t: it sound shit *because* the output amplifier can't go above -1dB, or actually 20% less at 16 ohms
15:16:31S_a_i_n_tthe IE8's can take then punishment on -6db...but your ears can't (not for long anyway)
15:16:38 Quit DerPapst (Read error: Connection reset by peer)
15:16:58TheSevenso the actual output level the headphones get will be around -3dB maximum. the amp isn't capable of supplying more
15:17:28kugelarg, this stupid data abort
15:17:44kugelhow can I debug those better than getting the failure address? the reading of -6 is not actually true?
15:17:57 Quit robin0800 (Read error: Connection reset by peer)
15:18:01TheSevenwhich reading?
15:18:08S_a_i_n_tthe volume
15:18:11kugelI would like a reg dump and/or the current ins
15:18:17TheSevenyou mean +6?
15:18:18S_a_i_n_tmine goes up to -6dB
15:18:23 Join DerPapst [0] (
15:18:27S_a_i_n_tahh, yes
15:18:41S_a_i_n_tsorry, forgot to shift
15:18:50TheSevenmaximum RMS output power of the amp is ~32mW, which is -3dB
15:19:11TheSeven(which you'll get when setting it to -1dB, due to some inaccuracies)
15:19:25TheSeveneverything above that will just clip at 16 ohms
15:19:35S_a_i_n_tHmmm...howcome it displays up to +6dB then?
15:19:48S_a_i_n_toh, right. I get it now.
15:20:17TheSevenit will apply up to +6dB to audio data that doesn't go above -9dB ;-)
15:20:45n1skugel: you can hack up the exception handler to print a regdump etc
15:21:07TheSevenbtw, when it's feeding into a line load instead of headphones, it will reach +1dB (and the setting will be accurate)
15:21:10kugelI think funman once had a patch for that stuff
15:21:47TheSevenn1s: why didn't that go into svn?
15:22:30*TheSeven likes the "crashkernel" approach, that would also allow to get a memory dump and look at the hardware
15:23:04kugeln1s: it seems to be on the tracker (grepping logs is awesome :) )
15:23:19n1sTheSeven: i think it was because it used some cpu feature that some cpus were missing or something like that
15:23:46TheSevenwe should be able to get around that, at least for ARM, which would be quite helpful for a start
15:23:59n1syeah, it was arm only afaik
15:24:12kugelthis one should be committed
15:24:19n1sbut you should ask funman or whoever was against it :)
15:24:30n1skugel, whic fs#
15:24:31kugelthe problem seems to be proper #ifdef'ing for ARMs without cp15
15:24:42kugeln1s: FS #10296
15:27:19n1syeah, just a whitelist #ifdef including it for cores we know have the necessary features should be fine
15:28:00 Join jgarvey [0] (
15:28:23 Join funman [0] (~fun@rockbox/developer/funman)
15:29:56 Join preglow [0] (
15:30:30preglowjoin #qt
15:30:37preglowtake 2
15:30:52funmanFS #10296 displays fault status and address causing the data abort (not the adress _where_ it happened), but we need to examine each model and see if it supports those registers (they are mandatory in ARMv6+, we need to check v4/v5)
15:31:27kugelis there a armv5 which doesnt support it?
15:31:56preglowi doubt v4 supports those
15:32:11kugelwe could extend our ARM_ARCH define to include some bitmask for certain features
15:32:32preglowwhy not just check for the arm revision? most features are consistent across revisions
15:32:37kugelpreglow: 922t does, at least
15:33:18funmanTorne knows the details, the registers are present if a MMU or a MPU is present
15:33:31preglowyeah, that's what i expected, most v4s don't have that, if any
15:33:49funmansansa AMS at least do
15:34:11kugelso, why would the cpu throw a data abort on address 60B4AE04? does it know where an address is good or bad?
15:34:45preglowi'm wrong, btw
15:34:51preglowloads of armv4s have mmus
15:34:53kugelthe address if foobar, btw
15:34:58amiconnkugel: That's not valid ram or rom, is it?
15:35:09kugelpreglow: we already knew you were wrong :p
15:35:21kugelamiconn: no
15:35:23preglowyeah, but not the extent to which i was :>
15:35:31funmankugel: you should look at lr and print a bit of stack to see where it comes from (perhaps a borked function pointer?)
15:36:10funmanthough you will have a different stack pointer so you need to switch to supervisor mode before reading the stack
15:36:21Tornefunman: or if they feel like it
15:36:31Tornefunman: it has occasionally been present on chips with neither :)
15:36:37Torneit's not there on arm7tdmi though
15:36:52 Join evilnick_B [0] (~0c140464@rockbox/staff/evilnick)
15:36:54kugelfunman: how can I get lr?
15:36:56Tornefunman: also, I was going to work on that patch for you..
15:37:08Tornefunman: to add support for ARMv6 and to make it ifdef'ed right. I just hadn't gotten around to it :)
15:37:25kugellr - 8 seems to the be the fault address
15:37:29funmanTorne: well i'll let you do ;) i try to look for a good place in headers to discriminate CPUs
15:37:57Tornefunman: well your code is fine for armv5, but it won't work for prefetch/undef on armv6 as they have a separate IFSR :)
15:38:32kugelfunman: the fault is in an .S function, btw
15:39:18amiconnkugel: Is this the mpegplayer crash with eabi?
15:40:03kugelit doesn't seem to be caused by short enums, I added a dummy with the value of INT_MAX to all
15:40:15amiconnSubtle difference in parameter passing between old abi and eabi?
15:40:25amiconnWhich file+function?
15:41:20funmanTorne: We "only" need to check TCC77x TCC780x DM320 AT91SAM9260 PNX0101 S3C2440 DSC25 and S5L870x CPUs
15:41:25kugelI only know that the passing of 64bit types has changed. but the function in question doesn't have those
15:41:42kugelamiconn: idct_arm.S:mpeg2_idct_add
15:42:06n1sfunman: pnx0101 is arm7tdmi iirc and the only port using it is dead
15:42:51kugelfault 8 is "precise external abort, nontranslation" apparently
15:44:47*kugel still thinks we should remove dead ports from svn to make it easier for people to work on active code
15:46:47n1skugel: it's not in they way much
15:46:55funmanTCC780x DM320 and S3C2440 do have an mmu (they use mmu-arm.S), I don't know about TCC77x (iaudio7/m200/c100/logikdax, all abandonned?), AT91SAM9260 (lyreproto1 = abandoned?), DSC25 (archos av300 = no code = abandoned) and S5L870x (meizus, nano2g)
15:47:43kugeln1s: what's the point in keeping them? It's not like we can't recover their code from the svn history
15:47:48funmans5l870x have a mmu code so they're ok (why specific code btw?)
15:48:55n1skugel: true but since tey don't really get in the way i don't think removing them is very important but i wouldn't be against it
15:49:05Tornefunman: well, if they have a "standard" ARM cache or MMU or MPU they must have CP15
15:49:25funmanI see that tcc77x manipulate cp15 cache registers
15:49:56Tornedo we not have a CPU type define of some kind?
15:50:07kugelfunman: non-standard caches I believe
15:50:08Torne#define ARM946EJ-S or similar
15:50:16funmanTorne: no, only V4 V5 and V6
15:50:22TorneWell, then I think we should? :)
15:50:28Tornesince that's sufficient to determine cp15 presence afaik
15:50:42Torneand it also determines dozens of other things
15:50:42funmanwhat about ARM_CACHE_PRESENT ?
15:50:50Tornewhich we may not have a need for right now, but may be usefu lone day :)
15:51:01Torneis that an existing define?
15:51:06TheSevenfunman: s5l8700/s5l8701 doesn't have a mmu
15:51:10Tornenot all ARMs with caches have cp15, see portalplayer
15:51:13n1sthere is a CPU_ARM7TDMI
15:51:15Tornethey have weird external cache
15:51:30TorneYeah.. do we not have more of those?
15:51:33 Quit froggymana (Quit: CGI:IRC)
15:51:40 Quit anewuser ()
15:51:50TheSevenbtw, why do we need to access cp15 at all?
15:52:06n1sTorne: i don't think so
15:52:17funmanTheSeven: what's mmu-s5l8700.S for ?
15:52:40TheSevenonly cache coherency stuff
15:53:00kugelTheSeven: fsr and far
15:53:01TheSevenit actually has a MPU
15:53:06funmanTorne: hm or MMU_PRESENT ? this would make #ifdef simpler if we only check this information in system-arm.c
15:53:33Tornefunman: Well, I would suggest that we define the specific CPU type in the target configs
15:53:44Torneand then use that to infer a bunch of other defines like MMU/cache/etc
15:53:54Torneand then use *those* to enable specific code
15:54:04TorneThis is probably a reasonable amount of effort from where we are now though
15:54:08Tornebut i think useful
15:54:49Tornefunman: lemme find an example
15:56:12amiconnkugel: Hmm, that's really odd. Do you know whether it works on the beast using eabi?
15:56:56Torne <- is this publically visible?
15:56:59Tornei think it should be :)
15:57:03Torneit's been open sourced
15:57:05 Join b0hoon [0] (~quassel@
15:57:19Tornethe top of the file has all Symbian's implications for CPU type -> features
15:57:23Tornewe'd want something like that
15:58:03*amiconn builds an eabi toolchain
15:58:21funmanTorne: sounds good, perhaps in a new cpu_arm.h file
15:58:41Tornefunman: that way you can write a single api for cache/mmu/etc that works for whatever
15:58:49Tornewithout target-specific defines
15:58:52Tornejust ARM feature defines
15:58:55kugel"#if !defined(__CPU_ARM1136_ERRATUM_411920_FIXED)" :)
15:58:59Tornekugel: Yeah
15:59:13TorneIf you get silicon that has a fix for the erratum you can define that to turn our workaround off
15:59:18Torneregaining whatever was lost by the workaround :)
15:59:25Tornesee now this code is all open i can talk about it ;)
15:59:39Torneyou'll notice there are a *lot* of erratums that are relevant :)
15:59:42TorneARM are awesome like that
16:00:16kugelsymbian (core) is OSS since a while
16:00:51funmaneclipse public license 1.0 isn't compatible with GPL but we can use the idea of this file
16:00:59kugelamiconn: the parameter passing looks different in the disassemblies
16:01:00TorneI didn't mean to copy it
16:01:04Tornejust the concept
16:01:22funmanTorne: it would have been faster :P but screw those millions of licenses
16:01:46n1sreinventing the wheel is good fun after all
16:02:07funman <- this diagram shows a MPU for the arm946e-s used in tcc77x
16:03:02funmannobody is listed having an iaudio7 on RockboxTesting :/
16:04:24kugelit's also a dead port
16:04:41funmantoo bad vitja left :/
16:05:03funmanbtw I contacted pbxy over the forums, he was the first to run code on Clipv2 and he might come back in the effort
16:05:46amiconnkugel: Hmm, binutils did build, but gcc doesn't build:
16:05:57amiconnconfigure: error: Building GCC requires GMP 4.1+ and MPFR 2.3.2+
16:06:08amiconnDoes that sound familiar?
16:06:09kugelyea, you need those since 4.4 to buiiöd gcc
16:06:50kugelyou can install them from your distros package manager
16:08:33amiconnIf I would know the package names...
16:08:48 Quit Adnyxo (Ping timeout: 264 seconds)
16:08:53kugellibgmp and libmpfr I think
16:08:54n1slibgmp3 iirc
16:09:12n1smaybe -dev even
16:10:22 Join Farthen_ [0] (
16:11:09n1syeah libgmp3-dev and libmpfr-dev on debian
16:11:29amiconnAh. The latter also depends on the former
16:12:29funmanTheSeven: do all the s5l870x have the same CPU ? tools/configure say 'arm940tbe'
16:13:28funmanah, be is bigendian for meizum*, ysp3; nano2g is littleendian
16:13:44TheSevenafaik the 8700/8701 have an arm940tdmi
16:13:55kugeldoes it need the -dev packages?
16:14:16TheSeventhe 8702 has an arm926ej-s
16:14:29funmando we have targets usign the 8702 ?
16:14:33TheSevennot yet
16:14:36kugelI thought both libs are auxillary utils for the build process. I can't remember anymore if I tried without -dev
16:14:40TheSeventhat would be nano3g and classic
16:16:23*amiconn curses
16:16:32kugelamiconn: in the meanwhile: if you're interested in how it's called
16:16:58amiconnIf the process is interrupted for whatever reason, and you forget to clean up the intermediate mess, it *will* fail next time
16:17:23funmanTorne: is there a difference between 't' and 'tdmi' suffixes?
16:18:23kugeljust a shorthand I think
16:18:40linuxstbDoesn't each letter in tdmi mean something?
16:18:46kugelI don't think there's any t which doesn't support dmi as well
16:18:50funmant is for thumb, not sure what is dmi
16:19:14linuxstbAh yes, "Thumb Instruction, Debugger, Multiplier, ICE"
16:19:15kugeld is debug, m is fast multiplier, i..I forgot :)
16:20:42linuxstbin-circuit emulator
16:21:22*linuxstb doesn't know what that is though ;)
16:21:34*n1s googles
16:21:49*n1s finds
16:21:51 Join Adnyxo [0] (
16:22:01funmanTorne: I've compiled the list of ARM CPUs on FS #10296
16:22:33n1sit appears to mean jtag basically
16:22:46TorneJTAG is a protocol, ICE is a thing to access over it :)
16:25:46Tornefunman: you probably want to print the entire FSR, not just the bits from it, because some of the other bits mean something on various processors
16:25:58amiconnkugel: That call looks fishy. The new gcc version is a bit short to tell what it's actually trying to do
16:26:00Torneand some of htem mean something but aren't documented publicly :)
16:26:24kugelit also appeared fishy to me
16:26:32amiconnYou can clearly see that either r0 or r1 must be wrong, because they're identical when calling mpeg2_idct_add()
16:26:41funmanit appears only the arm7tdmi (PP, pnx0101, dsc25 SoCs) don't have those registers
16:26:56kugelcompiling with -O0 would show how it expects the parameter passing, right?
16:27:22amiconnfunman, Torne: There already is a macro for arm7tdmi specifically
16:27:33Torneamiconn: this isn't the only feature that should be conditional on chip type, though
16:27:48Torneamiconn: all the MMU/cache stuff should really be derived from ARM CPU type, not from target
16:28:05Torneat least where the target uses ARM's standard macrocells
16:28:22Tornewhich is also true of basically everything but PP
16:29:34*funman agrees and will let Torne do the work
16:29:46Torneyeah thanks for that
16:30:09Tornewell, i will, at some point, if nobody else does. but i can't make many promises re when tha tmight be
16:30:14Tornefinal fantasy xiii came out today
16:30:36ranmachanfunman: Went to Tokyohackerspace today, but left the JTAG cables on the board for now :)
16:30:38funmanfor now i'll just check ARM_ARCH < 6 and !CPU_ARM7TDMI
16:30:51Tornefunman: you can do it for 6 as well, it'll just be incomplete
16:30:59funmanranmachan: didn't test already, it'd be good if someone with an e200v2 tests it first
16:31:06ranmachanSadly a bit cramped in there, would be nice to fit a connector...
16:31:09Tornejust i suggest you make it print the whole FSR val instead of just those two fields
16:31:26funmanalthough I could move the check after the known-to-be-working button check, i'm a bit lazy those days :)
16:31:28ranmachanfunman: Well, with the ifdefs it should only be used on C200V2.
16:31:29 Join MethoS- [0] (~clemens@
16:31:57funmani noticed that, perhaps we should just commit it now and check incrementally on each other model
16:32:06funmanTorne: sure
16:32:23funmani have an armv6 manual handy, i can add the armv6 code unless you prefer to do it
16:32:27TorneWell, do the armv6 stuff too
16:32:36Torneit's basically just checking IFSR instead of DFSR
16:32:47TorneDFSR on ARMv6 is at the same place as the unified FSR on armv5
16:32:59Torneso you just need to check a different cp15 reg if it's a prefetch abort.
16:33:15Torne(and obviously for prefetch abort, the FAR is irrelevant, it's lr that tells you where the abort was)
16:33:17funman'on armv5' => but it's the same on armv4 & armv5 ?
16:33:27Tornearmv4/5 have one fault status register for all faults
16:33:37Tornearmv6 makes that register for data aborts only, and adds a new one for prefetch aborts
16:33:53kugelhm, that -O0 doesn't appear to have worked
16:34:55 Join toffe82 [0] (~chatzilla@
16:35:23 Quit leavittx (Read error: Connection reset by peer)
16:37:55funmanTorne: domain & status bits will always be the same, i.e. there could be additional info in bits 31:8, but 7:0 still have the same meaning ?
16:38:46Tornestatus is [10,3:0]
16:38:48Tornenot just 3:0
16:38:53Torneso you are printing the wrong bits anyway
16:39:20TorneOh, that's onlt armv6
16:39:24kugelamiconn: eabi doesn't seem to care about r1 at all
16:39:28Tornealso armv6 has bit 11 for "write"
16:39:33Torneso you can differentiate read/write easier :)
16:40:10kugelamiconn: this is with -O0:
16:40:13Tornefunman: er, so "sort of". the domain field will always be 7:4 but the status encodings were greatly expanded in armv6
16:40:20Torneas well as the extra bit for "write"
16:40:36Tornefunman: read ARM ARM rev I or later :)
16:40:53funmanwhere ARM = Architecture Reference Manual ?
16:41:13funmanand I is capital i
16:41:16funmangot that
16:42:07kugelamiconn: of course not even the -O0 version works
16:43:55kugel1 parameter seems to be passed via stack
16:45:35MarctraiderHello guys, TheSeven has helped me with my ipod black nano 2G and its recovered to original Ipod firmware, WITH EU volume limit removed! :) i'm just glad i didnt have to throw my ipod away... going for a smoke now see'yal later!
16:52:44 Quit advcomp2019 (Ping timeout: 260 seconds)
16:53:40 Join Strife89 [0] (~michael@
16:53:45 Join advcomp2019 [0] (~advcomp20@unaffiliated/advcomp2019)
16:54:35n1spixelma: want to test something re stkov in properties?
16:55:05pixelmalater, sure
16:55:20kugelamiconn: it seems like this from old to new gcc: r0 = r0; r1 > r2; r2 > r3; r3 > stack
16:56:31Tornekugel: have you got something that's become 8 bytes because of the eabi requirements?
16:56:42Torneit sounds like the first arg is 8 bytes big..
16:57:00kugelno, void mpeg2_idct_add(const int last, int16_t * block, uint8_t * dest, const int stride);
16:57:21kugelyes, it appears this rule applies. but it shouldnt
16:57:23Torne..const int?
16:57:46n1spixelma: ok, then try declaring the mp3entry struct on line 69 outside the function so it's not on the stack; it's the only obviously large stack eating thing in the plugin except for the fact that gcc might go crazy with inlining since all functions are static
16:58:10kugelTorne: is there a problem with that?
16:58:16 Join anewuser [0] (anewuser@unaffiliated/anewuser)
16:58:18Tornenot that i know f
16:58:32Tornei've just never found declaring parameters as const to be useful :)
16:58:42Tornewell, values
16:58:46kugelit rather sounds like the 2nd arg is 8 bytes now
16:58:47pixelman1s: umm, well, how?
16:58:52***Saving seen data "./dancer.seen"
16:59:02kugelah well, that doesn't make sense either :)
16:59:52kugelconst params make sense. that tells the caller that the regiser being left untouched
16:59:53n1spixelma: just move the line outside the function
17:00:24 Join LambdaCalculus37 [0] (~rmenes@rockbox/staff/LambdaCalculus37)
17:00:31kugelr0-r3 are caller saved, so if the caller needs them after the call it may omitt the saving
17:01:06 Part LinusN
17:01:31n1spixelma: above the function even (or it won't build)
17:02:12kugelTorne: the stack parameter goes away when I swap the 2nd and 4th parameter
17:02:42 Quit n1s (Quit: Lmnar)
17:05:13kugelwooh, mpegplayer runs now
17:05:20kugelbut, what the hell is this!?
17:06:06kugelmaybe I should write to the gcc ml
17:06:11kugelthat appears to be a bug
17:07:10Tornekugel: yes, sounds like it might be
17:07:16Tornekugel: however your argument for const params is wrong.
17:07:26Tornekugel: You still have to save r0-r3
17:07:45Tornecalling them const only guarantees that the called function doesn't assign to those variable *names*, it says nothing about the registers
17:07:51TorneThe function is still free to overwrite r0-r3
17:08:05TorneThe parameters don't necessarily stay in those registers.
17:08:20kugelgood point
17:08:23Tornethe const *only* protects you from accidentally assigning to the params when you meant to write something else
17:08:29Tornethere is no difference in the generated asm, no optimisation benefit.
17:08:55 Join CaptainKewllllll [0] (~2669ecc2@gateway/web/freenode/x-gbpvdlabrmxzmter)
17:08:59funmanTorne: can you check ?
17:10:06funmanperhaps I could enhance comments "no MPU/MMU" can refer to arm7tdmi or to !arm7tdmi in the context
17:10:28Tornefunman: looks good, except don't print FAR for prefetch aborts
17:10:33Torneit's not updated
17:10:50Torneso it would just be the address of the last data abort, which is not useful ;)
17:11:03funmanhm right
17:11:25Tornealso you might consider for undef aborts, fetching the instructoin and printing that
17:11:39Torne(you know it's fetchable or you wouldn't get an undef)
17:11:55Torneit's not vital, though, i guess, you can find it in the elf binary
17:12:50Tornemeh probably not worth it
17:12:56Torneit's gonna be vanishingly infrequent
17:13:21TorneDo all the targets we're using here have room for 7 lines of text on screen?
17:13:29Tornei assume yes, but..
17:13:47funmanassuming 8 pixels font yes
17:13:53kugelmost should handle 7 in the 8px sysfont
17:14:10Torneok, cool.
17:14:31funmanlength should be ok, i had tested back then on Clip (128x64)
17:14:46 Part b0hoon ("GTG Bye.")
17:15:02 Quit liar (Quit: Verlassend)
17:17:49 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
17:19:53CIA-5New commit by funman (r25087): Display Fault status address register on data/prefetch aborts ...
17:20:43Torneanyone know what ARM1136 revision the i.MX31 is? :)
17:21:45funmandebug menu could print the ID code on arm
17:22:14 Quit MethoS- (Remote host closed the connection)
17:27:45Torneagain, only if it has cp15, but yes ;)
17:28:59UnhelpfulTorne: it does, cycle counter is somewhere in cp15 ;)
17:29:22Unhelpfuls/cycle counter/terribly generic performance counter/ ;)
17:29:40CIA-5New commit by funman (r25088): mkamsboot: Use audio master irq_enrd0 usb_status bit instead of gpio on C200V2 ...
17:30:10TorneUnhelpful: what?
17:31:43Unhelpfulcp15 is the system control coprocessor, yes? i've used parts of its functionality on beast.
17:31:45ThomasAHfunman: any news on the clip+? sound (even if it's just radio) would be nice :)
17:32:35funmanThomasAH: no, when we have sound we'll send the 'gentlemen' email
17:32:45*kugel sent the mail
17:32:57ThomasAHfunman: what is a 'gentlemen' email?
17:33:06TorneUnhelpful: i meant funman's point about the id code
17:33:10TorneUnhelpful: not perfctrs.
17:33:36ThomasAHah, "Gentlemen, we have SOUND!" ... found a few
17:33:43funmanThomasAH: see
17:33:58UnhelpfulTorne: ah, so the beast's CP15 may not have the ID code?
17:34:08FlynDiceThomasAH: I've been looking at the clocks setup for awhile now trying to sort things out but it's being rather finicky....
17:34:43TorneUnhelpful: no, all ARMs with cp15 have the id register, but arm7tdmi does not have cp15 and thus no id register
17:35:10amiconnkugel: Is that only for this function, or all functions?
17:35:21amiconnIn fact that looks like a gcc bug to me
17:35:28funmanTorne: isn't cp15 mandatory & standardized on armv6 ?
17:35:34UnhelpfulTorne: ...ah... i think my error was in thinking funman was replying to your <Torne> anyone know what ARM1136 revision the i.MX31 is? :)
17:35:46kugelit's possible that other functions are hit as well
17:35:56funmanUnhelpful: but i was ! :)
17:35:59ThomasAHFlynDice: clock works fine for me after setting it on both sides in the correct order :)
17:36:05amiconnWe could work around it by reordering the parameters of the idct_* functions. Of course that needs adjusting all asm implementations, not just arm
17:36:13ThomasAHFlynDice: or clocks as in ram speed and such?
17:36:34kugelI think the problem is exposed by the fact that mpeg2_idct_add() is written in asm. if it was written in C it would probably follow the same strange calling convention gcc dictates here
17:36:49FlynDiceThomasAH: er not the time but fclk, pcclk etc...
17:36:51Unhelpfulkugel: what "strange" calling convention?
17:37:00Tornefunman: yes.. we should do that, just not on arm7tdmi :)
17:37:25amiconnThis *might* because the 2nd and 3rd parameters are pointers - gcc may consider them as (potential) 64 bit parameters, so r1 is skipped
17:37:34kugelUnhelpful: the gcc generated call appears to violate the standard calling convention
17:37:49ThomasAHFlynDice: this is for running without disabled caches I assume?
17:38:03Unhelpfulamiconn: ...are pointers 64b on ARM? and since when?
17:38:26amiconnNo - that's why I said it may be a gcc bug
17:38:37kugelamiconn: that wouldn't make sense. how can it put something into r3 if r2 was a 64bit type (i.e. would take space in r3 as well)?
17:38:50amiconnAnd also why I asked whether it's just this function, or all
17:38:57Unhelpfulkugel: what if you cast the pointers to uint?
17:39:06Unhelpful(in the call)
17:39:50FlynDiceThomasAH: Not sure, but the clocks setup right now just "happens" to work so perhaps that will help. Right now the as3525 fclk is at 6 MHz... clk_main(24MHz)/4
17:39:58 Quit Strife89 (Ping timeout: 256 seconds)
17:40:08FlynDicemake that as3525v2
17:40:19amiconnkugel: And if it's more than just that one, be prepared to see spectacular crashes on various arm targets. Everywhere - not just in plugins
17:40:47kugelUnhelpful: same result
17:41:28kugelamiconn: I assume there's no crash as long as the called function is also compiled by gcc
17:41:51ThomasAHFlynDice: I'm not familiar with such details. fclk is like the FSB speed on PC hardware?
17:42:03FlynDiceThomasAH: That's wrong, FCLK is 120 MHZ PCLK is at 6 MHz
17:42:08amiconnYes, but there are numerous asm functions in the core as well
17:42:15amiconn...e.g. ata transfer
17:42:45kugelmy hdd based yh925 behaved fine in the core
17:43:04kugelit appears to be dependant on the order of the parameters as well. swapping the 2nd and 4th one helps
17:43:19kugel(in the prototype only)
17:43:40FlynDiceThomasAH: and the bus is fastbus mode which means the processor is only clocked at 6 MHz right now
17:45:17ThomasAHFlynDice: thanks for the explanation, I probably have to read more about it
17:45:28Tornekugel: it definately sounds like an 8-byte-alignment issue
17:45:34mtFlynDice: I have dualboot now on the clip+, but when booting rockbox, it's stuck at the rockbox logo. (With I & D cache disabled)
17:45:40TorneAAPCS requires that 8-byte values be passed in r0/r1 or r2/r3
17:45:56kugelbut there is no 8byte value being passed
17:46:01TorneWell yes
17:46:06 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
17:46:09Tornebut that's why it goes away when you reorder them, i mean
17:46:16Tornesee algorithm in AAPCS section 3.3
17:46:26Tornenot that gcc necessarily quite does that.
17:46:40*ThomasAH has to leave ... bye!
17:47:01funmanmt: btw, bootloader doesn't use MMU nor caches, I/D caches needs only to be disabled in rockbox.sansa
17:47:05kugelmy AAPCS doesn't have section 3.3 :(
17:47:36Torneer 5.5
17:48:27FlynDicemt: try putting an mci_delay() after the second MCI_CLKDIV down at line 460 of sd-as3525v2
17:48:48*amiconn wonders whether this a general gcc-eabi bug, a gcc 4.4.3 bug, or a quirk in arm eabi
17:49:04amiconnWhat's the earliest gcc supporting eabi?
17:49:28Torneamiconn: the ABI is fine, but other than that I can't say :)
17:49:39FlynDicemt: I didn't need this until a few days ago but I think funman's clip+ needed it all along
17:50:09funmanFlynDice: storage should be initialized after LCD, so i'm not sure if that's the problem
17:50:21amiconnWell, the reason why I am wondering is because two of those parameters are pointers, and two are ints
17:50:43amiconnIn svn they are int/ptr/ptr/int, and that causes the strange gcc behaviour
17:50:58Tornewell, the order should not matter
17:51:02amiconnI guess that both int/int/ptr/ptr and ptr/ptr/int/int would work fine
17:51:03Tornebecause the width of all those types is 4
17:51:22Tornei can't see any reason in AAPCS why it should spill no matter the order
17:51:27Torneand have never seen this happen in other code
17:51:43amiconnSo it's probably a gcc bug
17:52:13kugelcan I make gcc not call ld?
17:52:25Tornesure, -c
17:52:27FlynDicefunman: mine started hanging there last week and I added the delay and it seemed to make it work at least
17:52:48amiconnkugel: What are you trying to do?
17:52:54funmani've not hacked on my Clip+ for 1 week or so
17:52:56 Join DataGhost [0] (
17:52:56 Quit DataGhost (Changing host)
17:52:56 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
17:54:09kugelthey want a self-contained C file to reproduce this problem
17:54:30amiconnAh, yes
17:54:58funmangcc -E perhaps ? so all #include are replaced by the header content
17:55:00amiconnBtw, because you asked why the SH1 patch isn't pushed upstream: The gcc team is strange if you report a bug
17:55:19kugelI only have bare metal cross compilers, so I need a crt0.o to verify my test case shows the problem
17:55:43funmankugel: can't you just check the asm ?
17:56:12 Quit fyrestorm (Read error: Connection reset by peer)
17:56:14kugelah yes
17:56:41*amiconn now has an eabi beast build
17:57:01mtIs everything else other than USB working fine on the e200v2 ? ( I've seen the wiki, but I don't have a v2 to check myself )
17:57:29funmanwhy is locked : not posted in the correct forum?
17:58:11gevaertsIt would be moved then...
17:58:16amiconnkugel: gcc produces the same mess for ARMv6
17:59:24kugelhm, the test case doesn't show the problem
17:59:54funmangevaerts: i thought forum mods locked topics before moving them, but it seems no one is viewing this topic (was when i checked it)
17:59:58 Join chrisb [0] (
18:00:34gevaertsfunman: no, you'd see a redirect message then, not the original
18:03:12mtkugel : Any idea about my e200v2 question ?
18:04:35 Join guymann [0] (
18:05:03kugelmt: that sounds about right
18:05:58*amiconn builds gcc 4.3.4 for arm eabi
18:06:21gevaertsfunman: I'll unlock and move it
18:07:08 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:13:33 Quit LambdaCalculus37 (Quit: This computer has gone to sleep)
18:15:09 Quit pamaury (Quit: Quitte)
18:21:00 Join MethoS- [0] (~clemens@
18:24:19mtFlynDice: mci_delay() didn't work
18:29:45 Join liar [0] (
18:30:32 Join Buschel [0] (
18:31:42 Join LambdaCalculus37 [0] (~rmenes@rockbox/staff/LambdaCalculus37)
18:33:10 Quit petur (Quit: *plop*)
18:34:00 Join CFP [0] (
18:34:39 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
18:37:22domonokygsoc project application time has begun !! who fills out our application ? ie who wants to be our gsoc admin ?
18:38:49*chrisb enjoying his sansa e200 today
18:39:06chrisbit took a while to understand the background database update
18:40:23 Join fml [0] (
18:40:29 Join bertrik [0] (
18:41:22fmldomonoky: hello. If you're able to build manuals, could you please try FS #11088 and tell me whether it compiles? I have no build environment at the moment.
18:42:49domonokyfml: sorry i also cant build the manual at moment.
18:46:06fmldomonoky: ok, no problem. Anybody else with latex and build environment installed?
18:46:30amiconnHmm, arm-elf-eabi-gcc 4.3.4 produces the same nonsense
18:46:41Buschelfml: I can try after another build finished.. will take some minutes
18:48:13 Quit Marctraider (Ping timeout: 240 seconds)
18:49:05 Join Marctraider [0] (
18:50:30 Quit anewuser ()
18:50:53fmlBuschel: ok, thanks!
18:51:22kugelamiconn: :'(
18:51:34*amiconn is going to try 4.0.4
18:52:53kugelthat one can't do eabi afaik
18:54:08 Quit Tomis (Ping timeout: 260 seconds)
18:54:45amiconnWell, it just configured for eabi
18:55:26 Join Tomis [0] (~Tomis@
18:55:40kugel"The commercial ARM RealView C/C++ compiler was the first to support EABI, and usable EABI support came into GCC from version 4.1.0."
18:55:41 Join Highlander [0] (
18:56:16funmandid you try code sourcery gcc ?
18:56:22Buschelfml: the patch compiled and I have a .pdf file as result. Shall I check for something?
18:57:11funmanhardware.bug tests advice to use CodeSourcery 2009q1 over FSF GCC 4.3.4 (for free compilers)
18:57:58kugelI've tried 2009q3
18:58:56***Saving seen data "./dancer.seen"
18:59:50fmlBuschel: not necessarily, I just wanted to know whether it compiles. But if you have a minute you can look at some table with button maps and tell me what vertical spaces (with or without patch) looks nicer. Or you could just post the PDF to a binpaste so that I could compare as well.
19:03:23 Quit chrisb (Ping timeout: 245 seconds)
19:04:15 Join carpathia [0] (
19:04:54 Part carpathia
19:05:44FlynDicemt: I don't know what else to suggest besides trying to turn it on several times. While I've been playing with the clock setup I've found it will hang right there on startup sometimes but if I turn on/off several times it will finally boot up.
19:06:26funmankugel: which build are you seeing the mpegplayer for ?
19:06:56*kugel can't parse that
19:07:11funmankugel: which build are you seeing the mpegplayer problem for ?
19:07:48kugeleabi builds with the 4.4.x compiler (apparently 4.3.4 too)
19:07:58kugelit's not clear whether eabi or the new compiler causes it
19:08:06kugelit's definitely a compiler bug though
19:08:20funmani meant, any target? i'm building fuze to see if i can make a reduced test case
19:08:36kugelyes fuze
19:10:28 Quit DerPapst (Quit: Leaving.)
19:11:28amiconnfunman: beast also shows the problem
19:12:00amiconnI alreday tried gcc 4.3.4, with the same result
19:12:21 Join Strife89 [0] (~michael@
19:23:32Buschelfml: did you see my pm?
19:25:29 Quit Strife89 (Quit: Be right back.)
19:26:59funmani have a test case: the problem seems to be struct addressing
19:27:48funmantogether with attribute aligned
19:28:13fmlBuschel: sorry, no. But now I see it.
19:28:23kugelfunman: oh, interesting
19:31:40kugelit looks you'reright
19:32:51funmanaligned(4) = no problem, aligned(8) = problem
19:32:53kugel "fixes2 the problem
19:32:53amiconnSo it could be worked around by using an intermediate variable?
19:33:21fmlBuschel: actually, I don't quite like both. The reason is that in many places, the tables with the buttons directly follow section captions which isn't good style. I tried to eliminae the not needed table envs since they are there for nothing. We even fight their nature (being floating) via redefining placement to H (=here!, i.e. not floating at all)
19:33:23amiconnEh, volatile?
19:33:29kugelfunman: align(16) = problem as well?
19:33:33funmankugel: yes
19:33:40kugelamiconn: I wanted to make sure gcc doesn't optimize it away
19:33:55funmankugel: you want to report the test case to gcc ?
19:33:56kugelI forgot I use -O0 anyway
19:34:02kugelfunman: sure, will do
19:34:05amiconnUsing volatile will slow down things...
19:34:35kugelnot in this case I think
19:34:45kugelthe memory isn't even derefenced
19:35:03amiconnTry without it...
19:35:18fmlI think in many (most? all?) cases the sections should be replaced with table captions. But that's too much to do and will probably also require some rewording. So I'll leave it.
19:35:19amiconn(and -O2 or whatever the standard level is)
19:37:14funmankugel: i had to use (volatile int16_t*) -not uint- and cast block to non volatile, now it looks fine
19:38:48 Join Stephen__ [0] (~S@
19:39:49 Join mitk [0] (
19:39:58kugelit needs volatile
19:40:14kugelin the declaration
19:40:26kugelif -O0 is not used
19:40:32 Quit fml (Quit: CGI:IRC)
19:40:49funman : builds & runs with default flags
19:41:35 Join planetbeing_ [0] (~planetbei@
19:41:36 Quit planetbeing (Read error: Connection reset by peer)
19:41:37 Nick planetbeing_ is now known as planetbeing (~planetbei@
19:41:54 Join Strife89 [0] (~michael@
19:42:32 Join planetbeing_ [0] (
19:42:38 Nick Farthen_ is now known as Farthen (
19:43:03mitkHi, can an dev take a look at FS #11082 - Polish character set for 08-Atadore font?
19:43:11funmankugel: even shorter testcase
19:43:23 Join stripwax [0] (
19:43:44funmanworking with aligned(4), not with aligned(8)
19:44:11 Join leavittx [0] (~leavittx@
19:44:26 Join stoffel [0] (
19:44:47funmanperhaps gcc assumes aligned on 64 bits means 64 bits
19:45:27kugelfunman: what happens in the disassembly?
19:45:36kugelblock is on stack, or in r3?
19:45:47funmanjust look if r1 is modified or not
19:45:56kugelI didn't compile it :)
19:46:04funmanr0 is always 0
19:46:28funmans.member goes into r1 for aligned(4) , into r2-r3 for aligned(8) - and thus leaves r1 untouched -
19:46:32amiconnfunman: You're passing a pointer - and why should the pointer be aligned
19:47:15TheSevenS_a_i_n_t: I'm confused. The iLoader release you installed managed to boot /.rockbox/rockbox.ipod? I just found a bug that *must* have prevented this from working
19:47:25funmanamiconn: DCTblock is "int16_t * ATTR_ALIGN(16)" in mpeg2_internal.h
19:47:52amiconnYeah and?
19:48:28amiconnYou're passing a pointer to it. That pointer is 32 bit. Alignment of its destination shouldn't matter at all
19:49:31funmanI don't understand what you're asking me
19:50:39funmanI don't know if alignement of the pointer is relevant for I don't know mpeg2dec code
19:51:01kugelvolatile fixes your test case as well
19:51:47amiconnEh, wait. The struct member is aligned?
19:52:39amiconnThe block this member points to should be 16 byte aligned for performance reasons (cache line alignment, resp. alignment for line burst on coldfire)
19:52:48amiconnThe pointer itself doesn't need to be aligned
19:52:57 Join Strife1989 [0] (~michael@
19:53:49funmanshould it be int16 ATTR * ?
19:53:51 Join michaelcarr [0] (~michael@
19:54:43amiconnImo it should be int16_t * DCTblock;
19:54:59amiconnI'll check the svn log, perhaps there is a reason for this strange alignment
19:55:24 Quit Strife89 (Disconnected by services)
19:55:27 Nick michaelcarr is now known as Strife89 (~michael@
19:55:51 Quit CGL (Read error: Operation timed out)
19:57:57 Quit Strife1989 (Ping timeout: 268 seconds)
20:00:01kugelfunman: it seems likely that gcc confuses something here
20:00:36kugel64bit values must be 8byte aligned afaik, maybe now treats all 8+ byte aligned data as 64bit
20:02:03amiconnThere may be a bug in gcc, but it is also a bug in rockbox that was revealed by eabi
20:02:53kugela bug?
20:03:11kugelhardly a bug. valid code, just unneeded
20:03:24amiconnUnnecessary padding -> waste of ram
20:03:28Strife89Huh. "make: *** No rule to make target `/home/michael/rockbox/apps/codecs/libmusepack/musepack.h', needed by `/home/michael/rockbox/ipod_color/apps/codecs/mpc.o'. Stop.
20:03:45amiconnStrife89: make dep or full rebuild
20:04:20amiconnkugel: Try removing the attribute from the pointer, build and run
20:04:46 Join planetbeing__ [0] (
20:05:04*amiconn spots a possible optimisation for pp5020
20:05:37 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
20:06:03 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
20:07:02kugelamiconn: works
20:07:27 Quit CFP (Quit: Quitte)
20:07:41amiconnIt works on the beast as well
20:08:14amiconnShall I commit?
20:08:25 Quit Schmogel (Ping timeout: 265 seconds)
20:08:40 Quit planetbeing_ (Ping timeout: 276 seconds)
20:09:15funmanamiconn: you found what was that attribute for in the log ?
20:09:19CIA-5New commit by b0hoon (r25089): Packard Bell Vibe 500: reset all allowed devices like in the OF.
20:09:46amiconnfunman: Seems it was just a mistake when taking the buffer itself out of the struct. At least there is no explanation
20:11:22funmanit it serves no purpose for optimization, just commit the change
20:11:48kugelfunman: very good work on this find!
20:12:30amiconnIt *may* have to do with dualcore on PP. I'll try it in a moment
20:13:44*kugel bets the underlying bug is as simple as "if (eabi && var_align >= 8) ..." instead of "if (eabi && var_size >= 8) ..."
20:15:58pixelmabluebroth3r: ok, in my RbUtil testing on PPC I had the wrong TTS set... the Utility doesn't hang with the correct one and "Test TTS" sounds correctly. But again the speex talk files are just noise, going to test my Ondio again
20:16:52amiconnHmm. The decoder struct is dynamically allocated (as part of a larger struct) so that's pretty unlikely
20:17:03pixelmabluebroth3r: and I think RbUtil shouldn't freeze like that with the wrong TTS configuration
20:17:11amiconnThe internal allocation mechanism only aligns to 32bit
20:17:45amiconnkugel: Be prepared for some fun when switching to eabi...
20:18:08amiconn"region `PLUGIN_IRAM' overflowed by 340 bytes" in mp3_enc for ipod color...
20:18:13 Join krabador [0] (
20:19:40kugelcan you actually make sure a struct member is 16byte aligned if the whole struct isn't, and if the struct is dynamically alocated?
20:20:04amiconnThat's my point
20:20:24 Quit planetbeing__ (Quit: planetbeing__)
20:21:12amiconnAnother point is that even on PP libmpeg2 doesn't run dualcore itself. The core split is between video and audio
20:21:22amiconn...and audio is handled by libmad
20:22:25kugelisn't libmad itself dual-core'ized as well?
20:22:57amiconnmpa.codec is, libmad itself just offers the building blocks
20:22:59pixelmabluebroth3r: same effect - MP3 talk clips work :)
20:23:06kugelah, ok
20:23:16 Quit perfectdrug_ (Quit: perfectdrug_)
20:23:19 Join tvelocity [0] (
20:23:33amiconnIn mpegplayer it is used on one core
20:24:20amiconnIirc audio and buffering runs on CPU and video on COP (and the greylib also runs on COP on greyscale PP targets)
20:25:40 Join planetbeing_ [0] (
20:26:43 Join perfectdrug [0] (
20:27:47 Quit flydutch (Quit: /* empty */)
20:28:08 Join lark1313 [0] (
20:28:34saratogalibmad is actually a couple API calls one for parsing and mdct, one for subband filterbank
20:28:44saratogawe just run them on seperate cores in mpa.c
20:28:51saratogawhich is why it was such a trivial change
20:28:59saratogavery nice of them to write the codec in such an odd way
20:29:34 Join Schmogel [0] (
20:30:25 Quit lark1313 (Client Quit)
20:30:51 Quit mitk (Quit: Leaving)
20:30:56 Quit funman (Quit: free(random());)
20:35:51 Quit LambdaCalculus37 (Quit: This computer has gone to sleep)
20:38:33 Quit krabador (Quit: Sto andando via)
20:41:38 Join planetbeing__ [0] (
20:44:14 Quit planetbeing_ (Ping timeout: 276 seconds)
20:45:37CIA-5New commit by Domonoky (r25090): put zip size and zip contents into the db, (updated on checkwps runs) so it doesnt need to always look that up.
20:46:31 Join pamaury [0] (~c2c7a50a@gateway/web/freenode/x-fgmoonyfuhfrzchz)
20:46:59 Quit Seed (Ping timeout: 252 seconds)
20:47:05 Quit Schmogel (Ping timeout: 265 seconds)
20:47:36 Join CGL [0] (~CGL@
20:48:39amiconnmpegplayer runs on the Color, but there are some video glitches
20:50:15CIA-5New commit by Domonoky (r25091): try to fix the themesite.
20:52:19AlexPBuschel: SVN 470.27%, 26.4 MHz, align patch 470.63%, 26.38 MHz Impressive!
20:52:30Buschelhehe :)
20:52:53BuschelAlexP: no effect...
20:53:25AlexPnot as such, no
20:56:05BuschelAlexP: thanks for testing and enough for today...
20:56:49 Join Schmogel [0] (
20:56:59AlexPno worries
20:57:55 Join Seed [0] (
20:58:59***Saving seen data "./dancer.seen"
21:01:55amiconnDoesn't happen on oldabi so must be an eabi problem (or just a glitch)
21:02:29CIA-5New commit by amiconn (r25092): Get rid of unnecessary alignment of struct member which also caused arm-elf-eabi-gcc to mess up parameter passing to mpeg2_idct_add().
21:02:49 Join LambdaCalculus37 [0] (~rmenes@rockbox/staff/LambdaCalculus37)
21:06:51 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
21:06:51 Quit pixelma (Disconnected by services)
21:06:55 Quit amiconn (Disconnected by services)
21:06:57 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
21:07:04 Quit stoffel (Remote host closed the connection)
21:07:08 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
21:07:24 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
21:12:33 Join DerPapst [0] (
21:16:35 Quit LambdaCalculus37 (Quit: Fwump)
21:22:32 Quit JdGordon (Ping timeout: 260 seconds)
21:22:32CIA-5New commit by Domonoky (r25093): Add search function.
21:24:02 Quit mt (Read error: Connection reset by peer)
21:24:11kugelamiconn: the latest answer on gcc-help suggests "char __attribute__((aligned(8))) *member;" to say it points to something 8byte aligned
21:24:18kugeland he says it's indeed a bug
21:24:49kugelbut was never discovered because it's triggered by bizzare code :)
21:25:57 Quit TheSeven (Read error: Connection reset by peer)
21:26:12 Join The_Seven [0] (~theseven@rockbox/developer/TheSeven)
21:26:22 Nick The_Seven is now known as TheSeven (~theseven@rockbox/developer/TheSeven)
21:27:42 Join dfkt_ [0] (dfkt@unaffiliated/dfkt)
21:27:45kugelwhat sort of glitch did you see?
21:28:02 Quit Highlander (Quit: Quitte)
21:28:38 Join Digital_Pioneer [0] (
21:28:45 Join banzairun [0] (
21:29:07banzairunis this thing on?
21:29:11Digital_PioneerHi. I'm trying to install rockbox on a friend's Fuze, following this tutorial: and nothing happened...
21:30:08banzairunhi, I'm a dumbass newb asking a question that likely has an obvious answer that I nonetheless can't find on the web page...
21:30:24 Quit dfkt (Ping timeout: 264 seconds)
21:30:50banzairunI just installed rock box on an ipod video 80gb 5.5g
21:31:15banzairuncan I get my clickwheel itunes store games to work from the rockbox interface?
21:31:55Farthenbanzairun: no, they don't work in rockbox. but you could boot the original apple firmware
21:32:17banzairunhow do I switch between them?
21:32:38Fartheni think you need to hold the left arrow key right after you have reset it
21:32:59banzairunoh so prob like left is one OS and right is the other?
21:34:18Farthenno, it starts rockbox if you do nothing. and if you press the left arrow key it starts the apple firmware
21:34:44FarthenDigital_Pioneer: is it a fuze v1 or a v2?
21:34:47Buschelbanzairun: please see the manual -> turn on Hold switch immediately after turn on.
21:34:51banzairunok how disappointing tho... was hoping for ipod install0us :)
21:35:04FarthenBuschel: ooops, then i mixed them up ;-)
21:36:32Digital_PioneerFarthen: Pretty sure it's a v1, it just calls itself a Fuze.
21:38:03kugellook at the OF version, it must be 1.xx.xx
21:38:11FarthenDigital_Pioneer: then i'd suggeset to just follow the manual here:
21:39:36kugelDigital_Pioneer: I highly recommend you to use our automated installer, Rockbox Utility, instead of downloading a suspicious package from someone
21:40:59 Quit gill0r_ (Remote host closed the connection)
21:41:11 Join gill0r [0] (
21:42:37AlexPbluebroth3r: To build rbutil I do qmake, make inside the rbutilqt directory yes? As I'm getting the same errors as (I'm also on Arch)
21:43:57AlexPOr anyone else who has any ideas (domonoky?) :)
21:44:03Digital_Pioneerkugel: It doesn't run. Something about a libpng?
21:44:35CIA-5New commit by Domonoky (r25094): try to fix checkwps again.
21:44:55Digital_PioneerAlexP: Yup, same thing here. Arch.
21:45:14Unhelpfulwhat are other eabi blockers? testing all codecs/plugins, testing all platforms, or at least all cpu architecture+version?
21:46:03 Quit gill0r (Remote host closed the connection)
21:47:50domonokyAlexP: hm, it misses some speex symbols. do you have libspeex-dev installed on this machine ? ( the rbspeex makefile can use both, system speex and rockboxs own)
21:48:18AlexParch doesn't do -dev packages, everything is included in the same package
21:48:29AlexPi.e. libusb is both libusb and libusb-dev
21:48:50AlexPI have various speek things in /usr/lib
21:49:09AlexPBut it looks like the Arch speex package is incomplete?
21:49:12domonokyyou could try if you can build rbspeex itself successfully.
21:49:13 Quit bluebroth3r (Ping timeout: 246 seconds)
21:49:36AlexPOK, I'll try to build rbspeex
21:49:58Digital_PioneerI built rbspeex no problem, but what then?
21:50:17LloreanUnhelpful: Bootloaders need to be checked for all targets affected too.
21:50:32AlexPdomonoky: Yes, rbspeex builds fine
21:50:53 Join bluebrother [0] (
21:50:53 Quit bluebrother (Changing host)
21:50:53 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
21:52:34domonokyother things to check: does the rbspeex.a really miss the symbols ? and you could try to modify the rbspeex Makefile to always use rockboxs speex lib, and try building rbutil again.
21:52:44UnhelpfulLlorean: that's scary on targets with crap recovery :/
21:53:26LloreanUnhelpful: I think it's mostly been tested on those already
21:53:40AlexPdomonoky: The rbutil makefile?
21:54:06Zagoramiconn: here? what does your triton and jupiter clients say? they are bouncing.
21:54:14domonokyAlexP: no the rbspeex makefile. rbutil just calls that to create the .a file.
21:54:21AlexPah, OK
21:54:48 Quit shaggy-h ()
21:55:18domonokymake sure to clean both rbutil and rbspeex before rebuilding :-)
21:57:12amiconnZagor: "2010-03-09 21:56:45 HELLO failed: error duplicate name!"
21:57:15Digital_PioneerAlexP: rbutil is on AUR, trying -svn now.
21:57:21amiconn...over and over
21:57:29AlexPDigital_Pioneer: Interesting
21:57:59 Join shaggy-h [0] (
21:58:17Zagoramiconn: that's so strange. on my side it says "Disconnect: client jupiter-amiconn reason connection lost"
21:58:28kugelUnhelpful: so, now to fix doom :'(
22:00:59 Quit Strife89 (Quit: Going home.)
22:01:01amiconnkugel: The glitch was occasional "defective" macroblocks
22:01:33amiconnBut it does indeed look like it was just a glitch. Works fine in another test after a full rebuild
22:01:45kugelI don't see any glitches on e200/fuze
22:02:02 Join Altreus [0] (
22:02:26Altreushelp! My ipodpatcher tells me "Firmware partition doesn't contain Apple copyright, aborting."
22:02:39AltreusIs there a way I can tell it to do it anyway?
22:02:44saratogahave you installed ipodlinux or something weird like that?
22:02:54 Quit Buschel (Ping timeout: 276 seconds)
22:02:54Altreusno, that's what I was trying to do in the first place
22:03:04saratogawhat ipod?
22:03:08Altreus1g mini
22:03:26saratogayou could try a restore with itunes, that will fix the partitions if they're screwed up
22:03:43AltreusI don't have itunes, or even Windows or a Mac
22:04:05AltreusThe ipod is older than all the computers that have ever been used to access it
22:04:12Altreusnone of which exists any more :(
22:04:34stripwaxbanzairun - to reboot into the Apple firmware, you put the Hold switch to 'Hold' immediately after turning it on. (pressing left - that won't work - on an ipod 5.5g anyway)
22:05:00amiconnZagor: I wonder how the client script managed to get into this state
22:05:18saratogaAltreus: theres instructions for manually restoring an ipod on the wiki, search there
22:05:22amiconnDoesn't seem the first time that this happens, and also not just on my boxes
22:05:27Altreusthanks, I'll try
22:08:00 Quit stripwax (Quit:
22:11:38 Join froggyman [0] (
22:11:43Altreussaratoga: It worked! Thanks for pointing that out :)
22:11:50 Quit bmbl (Ping timeout: 245 seconds)
22:11:54saratogano problem
22:12:36Digital_PioneerAlexP: Any luck?
22:12:57AlexPDigital_Pioneer: nope, but I am poking around in the dark, so that isn't surprising :)
22:13:35saratogaDigital_Pioneer: i don't know why you couldn't use rbutil, but theres still the command line utils
22:15:50 Join endlesswonder [0] (
22:16:02endlesswonderwhy no support for Classic v6?
22:16:26AlexPencryption has only just been broken, nobody has done the huge amount of reverse engineering work needed
22:16:37endlesswonderis there a tool for Classic to be able to navigate by the Folders you have the Songs on your PC?
22:16:55AlexPDon't know, but that isn't a Rockbox question
22:17:18endlesswonderdoing a search for each single file is not all that intuitive nor quick
22:17:38AlexPI understand, but this is an on-topic channel :)
22:17:42endlesswonderesp. if you have a LOT of songs in different folders, that are not organized into albums by iTunes
22:18:40endlesswonderi mean come on now - just place my damn songs on the iPod as I originally had them on my PC!
22:18:57endlesswonderi went thru the meticulous time consuming effort to organize them via folers
22:19:28CIA-5New commit by Domonoky (r25095): try to fix search feature.
22:19:42saratogaendlesswonder: if you just want to chat, checkout rockbox-community
22:19:47saratogathats the offtopic channel
22:19:50 Quit efyx_ (Remote host closed the connection)
22:20:24Altreusyay now I have an ipod that refuses to work until I give it mains power
22:20:37endlesswonderkeep things as simple as possible
22:20:42endlesswonderbut no simpler
22:21:01endlesswonderunfortunately programmers get the idea adding fancy features & graphics, makes everyone happier
22:21:14AltreusThat's not programmers. that's managers.
22:21:44endlesswonderfunctionality and making things seen and unseen for the purpose and user's preference makes for far more Happy people.
22:21:46AlexPendlesswonder: Please, stay on topic
22:21:52AlexPi.e. stop
22:22:10*scorche settles down to get some GSoC stuff hammered out
22:22:21AlexPah, good work
22:22:31saratogagreat, looking forward to GSOC this year
22:23:14endlesswonderi doubt anyone would be a better software designer than me
22:23:24endlesswonderthis is not just fluff
22:24:02AlexPendlesswonder: For the last time - #rockbox-community for chat
22:28:02Stephen__gevaerts, ping
22:28:17Stephen__or domonoky
22:28:44 Quit arbingordon (Quit: `)
22:29:07 Join arbingordon [0] (
22:29:12Stephen__i acn't remember who but it ws about removing my basic blue/black art themes aqs the original author has upped his.
22:31:57 Quit Marctraider ()
22:32:19CIA-5New commit by zagor (r25096): Build failure blocks are lifted after 10 minues.
22:38:18CIA-5New commit by Domonoky (r25097): dont show themes multiple times.
22:39:41domonokyStephen__: done
22:39:49Stephen__thanks domonoky
22:43:01scorcheDevs: Please help this page out:
22:46:53 Join Casainho [0] (~chatzilla@
22:47:23amiconnBleh. For some reason eabi gcc doesn't build on opensolaris because 'patch' cannot find the files to patch?!
22:47:56amiconnThe patch works the same way as the gcc patches for other toolchains, so I wonder why
22:50:01kugelamiconn: maybe it doesn't accept/handle -p1 ?
22:50:14kugelthe eabi patch was made against 4.4.2
22:50:44amiconnIt does so for m68k, sh, and plain arm
22:51:10kugelmaybe it still looks at the patch paths (it says gcc-4.4.2 in the diff)
22:51:35 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
22:51:54kugelfunman, Torne: is there a way to get the fault address on arm7tdmi as well=
22:54:55 Join karashata [0] (
22:55:05 Join anewuser [0] (anewuser@unaffiliated/anewuser)
22:58:51 Nick krazykit` is now known as krazykit (
22:59:00***Saving seen data "./dancer.seen"
23:02:24Tornekugel: no
23:02:44Tornekugel: it's possible there is on PP, but I don't know of one
23:02:48Tornekugel: the cache controller might know it
23:03:21amiconnI doubt it
23:03:22Tornebut in general, no, ARMs have no way to know the address which caused the abort if they don't have an mmu/mpu
23:03:44 Quit Tuplis (Read error: Operation timed out)
23:03:52Unhelpfulkugel: i only said we should *test* doom!
23:05:07kugelthe regs are untouched by the handlers or mode switching?
23:06:04Tornekugel: hm?
23:06:25kugelwell could I hack something together if I know the reg where the address is in?
23:06:35Torneyou could just dump theuser mode registers, sure
23:06:46Tornebut i expect we either stack or overwrite several of them :)
23:06:49Torneotherwise you can't get to the handler
23:07:40kugelI want to know what r4 contains
23:08:13 Join stripwax [0] (
23:08:18*amiconn *hates* this beast usb bug
23:10:22Tornekugel: well, we don't save any registers, and they will most/all have been overwritten by the time you get to UIE
23:10:57Torneif you want to handle exceptions properly the thing to do is push all the user registers onto the abort mode stack and pass a pointer to UIE
23:11:21amiconnOn the beast doom just crashes with eabi
23:12:27amiconnHmm, different problem this time...
23:13:13 Quit kugel (Ping timeout: 246 seconds)
23:13:22 Quit DerPapst (Quit: Leaving.)
23:14:03 Quit Digital_Pioneer (Ping timeout: 268 seconds)
23:20:34amiconnHmm, opensolaris 'patch' fails if the patch file is a multi-file patch
23:20:35 Join Digital_Pioneer [0] (
23:21:15*amiconn wonders how his sh patch worked back when he created it
23:21:18 Quit hebz0rl (Quit: Ex-Chat)
23:21:38 Quit tvelocity (Quit: Αποχώρησε)
23:25:06 Join kugel [0] (~kugel@rockbox/developer/kugel)
23:26:11 Quit Casainho (Ping timeout: 264 seconds)
23:34:49kugelUnhelpful: I'm getting unaligned access in M_LoadDefaults()
23:35:20kugelon this line: *defaults[i].location.pi = defaults[i].defaultvalue.i; (the lhs is unaliged)
23:36:48 Quit alexbobp (Ping timeout: 248 seconds)
23:37:52 Quit Stephen__ (Ping timeout: 260 seconds)
23:38:34kugeldefaults is a static array, that's why its strange
23:40:50amiconnghah! This is another gnu vs. non-gnu problem
23:41:01amiconnIt works when using 'gpatch' instead of 'patch'
23:41:18 Join Marctraider [0] (
23:42:13 Join planetbeing_ [0] (
23:42:18kugelamiconn: does doom start on the beast and an eabi build?
23:42:34amiconnYes, it starts, but then fails
23:42:53amiconnIt hangs after some time (time varies), then powers off
23:43:17amiconnIf it runs long enough to start the demo animation, there are graphis errors
23:44:48 Quit planetbeing__ (Ping timeout: 264 seconds)
23:45:48 Join Stephen__ [0] (~S@
23:45:48 Quit jgarvey (Quit: Leaving)
23:46:40 Join Hillshum [0] (
23:47:53 Quit stripwax (Quit:
23:48:24 Quit evilnick_B (Quit: Page closed)
23:48:26kugeldata abort as soon as I select the difficulty (after executing the .wad). on my e200 it creases as soon as doom.rock launches
23:49:52amiconnI think the reason is probably the same
23:50:20amiconnData aborts might be disabled on the beast, so unaligned accesses don't crash, but they still give wrong results
23:50:49 Quit piotrekm (Quit: piotrekm)
23:51:04Tornewait what?
23:51:19Torneunaligned reads work on armv6
23:51:27amiconnThey do?
23:51:47TorneYes, if you disable alignment aborts it does them properly, with multiple cache accesses
23:52:04amiconnAh, hmm
23:52:12Tornewe shouldn't disbale alignment aborts, though ;)
23:52:20Tornesince it's still slower
23:52:25Torneand code that works that way won't work on other chips
23:52:42amiconnOn earlier arms which allow disabling that check, unaligned accesses work, but not the way one would expect
23:52:52TorneYes, they rotate the aligned word
23:52:58Tornebut armv6 and later do it properly
23:52:59 Quit Zagor (Quit: Clint excited)
23:53:32amiconnThen there much be another problem. They're probably related though
23:54:16kugelerr, only the first array element is unaligned
23:54:56*kugel likes that doom has printf() :)

Previous day | Next day