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 2010-04-04

00:00:55saratogahuh using ldm is actually a bit slower for me then doing single loads ont he clipv2
00:01:16 Join jacekowski [0] (
00:01:23saratogai wonder whats going on here
00:02:12CIA-5New commit by gevaerts (r25459): Add IO priority handling. Currently all IO has equal priority, except the dircache scanning thread which is lower. This fixes the slow boot problem ...
00:04:22saratogaalso why is memory so much faster on the e200v1 then on the clipv2
00:04:53kugelgevaerts: why was the IF_MD part needed exactl now?
00:05:03kugelsaratoga: do you test boosted?
00:05:10gevaertskugel: where do you mean?
00:05:22saratogakugel: no both unboosted
00:05:24 Quit xiainx (Quit: Good Bye!)
00:05:33kugelsaratoga: unboosted is horribly slow as far as I noticed
00:05:41*bertrik spots a 1:37 min build
00:05:52kugelgevaerts: "Lots of the changes are due to changing storage_(read|write)sectors() from macros to wrapper functions. This means that they have to be called with IF_MD2(drive,) again."
00:07:07kugelsaratoga: pressing up/down boosts and unboosts just like in test_boost
00:07:12saratogayeah i saw
00:07:18saratogawe need a higher resolution timer
00:07:42gevaertskugel: well, due to lots of interesting preprocessor magic, the drive argument was getting eaten without IF_MD2() in the macro case. That doesn't work for wrapper functions
00:07:43saratogaor bigger BUF_SIZE
00:07:44 Join xiainx [0] (
00:08:02gevaertsgah, sim...
00:08:14kugelsaratoga: why?
00:08:26saratogathe step sizes are quite large
00:09:36kugelgevaerts: the trade-off was to be able to do STORAGE_FUNC() in some places?
00:09:58saratogahah ldm is 2x slower on the 200v1, so i'm doing something wrong
00:10:05gevaertsSTORAGE_FUNC() was to avoid another lot of #ifdefs
00:10:29saratoga"ldmda %[ptr]!, {r0, r1, r2, r3};"
00:10:34saratogathat should decrease the ptr variable by 16 bytes right?
00:10:37CIA-5New commit by gevaerts (r25460): Don't define HAVE_IO_PRIORITY for the sim
00:11:06kugelsaratoga: not increase?
00:11:16saratogada = decrement after
00:11:24kugelah, ok
00:11:40gevaertskugel: in storage.h, there are things like #define storage_get_info(drive, info) ata_get_info(IF_MD2(drive,) info)
00:11:51kugelI saw that
00:11:54gevaertsHaving the IF_MD2() there is enough
00:12:05saratogasee anything wrong in there?
00:12:26gevaertsIf you then change storage_get_info() to be a wrapper function, the IF_MD() is pushed back to the caller, where it used to be before the macros
00:12:46kugelsaratoga: any reason you go backwards?
00:13:01saratogakugel: just how i wrote the loo
00:13:32gevaertsok, now the real fun...
00:14:10kugelsaratoga: I'm no expert in inline assembly, but line 11 looks strange to me
00:17:57 Join Buschel [0] (
00:18:17CIA-5New commit by gevaerts (r25461): Fix some forgotten storage_(read|write)_sectors() calls
00:18:30Buschelsaratoga: shouldn't ldmda be ldmdb (decrease before)?
00:18:59saratogaBuschel: does it matter?
00:19:56Buschelsaratoga: if i am not totally wrong you may call the loop 1 time too much
00:20:21saratogaso use "ldmdb %[ptr]!, {r0, r1, r2, r3};"
00:20:53Buschelsaratoga: yep
00:21:36Buschelsaratoga: hmm, i was wrong: the loop will end correctly. but you will access buf[] out of range with the first call of ldma.
00:21:51Unhelpfulsaratoga: that moves the pointer down by 4 words, and stores r0-r3 counting up from the new location
00:25:32Unhelpfulif you're looping backward starting from just beyond end-of-buffer you want ldmdb
00:25:56Buschelsaratoga: you are doing r0 = ptr, r1 =ptr+1, ..., r3 = ptr+3. ldmda will decrease ptr after r3=..., ldmdb will decreas ptr before the first r0=...
00:26:36saratogabut either way its the same thing right since i just discard the values?
00:28:52saratoga#define BUF_SIZE ((PLUGIN_BUFFER_SIZE-(10<<10)) / (sizeof(int)))
00:28:56saratogai don't understand this line
00:30:45kugelwell, the plugin buffer minus 10k (for the code of the plugin). / sizeof(int) because I made an int array
00:30:49saratogaah ok
00:31:11saratogayou couldn't just do PLUGIN_BUFFER_SIZE- 10000 :)
00:31:33kugelI guess I could :)
00:32:08amiconn10<<10 != 10000
00:32:10dockimblejust out of curiosity
00:32:40dockimblewould there be any way to program an mp3 player to be used as an external sound card?
00:33:09gevaertsdockimble: yes
00:33:11saratogayes but its very close and 10KB is only approximate anyway
00:33:13kugelsaratoga: I think you need to align the buffer properly if you do burst access, otherwise the loop might exceed the actual buffer
00:33:49CIA-5New commit by funman (r25462): Clipv2: read hold switch correctly
00:33:50 Join Darkknight512 [0] (
00:34:06*kugel isn't sure how to properly implement the scrollwheel irq without a useful timer
00:34:25dockimblewonder how....
00:34:40amiconnWhat does the scrollwheel have to do with a timer at all?
00:34:56saratogadoing 32 repeats works much better on PP
00:34:56gevaertsdockimble: have a look at
00:35:19 Quit evilnick (Read error: Connection reset by peer)
00:36:14jhMikeSgevaerts: a most strange patch, that one. Just and FYI: a shortcut for thread_get_current() is THREAD_CURRENT, which thread_id_entry respects.
00:36:17 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
00:36:42jhMikeS* "THREAD_ID_CURRENT" rather
00:36:44kugelamiconn: to tell the difference between 2 wheel changes
00:36:52gevaertsjhMikeS: oh, right. that might have been a bit shorter
00:37:07dockimbleokis, thanks
00:37:51amiconnThe wheel just fires interrupts whenever it's moved, doesn't it?
00:37:54kugelthe time difference between two wheel changes (i.e. two interrupts), to make BUTTON_REPEAT generation and acceleration work properly
00:38:13saratogathe top loop is half as fast as the bottom
00:38:15saratogaand i have no idea why
00:38:20amiconnOn 1st/2nd Gen ipod we're using wheel interrupts, independently of any timer
00:38:40dockimbleok, wow i am way too drunk to understand this now
00:38:41amiconnWe're not generating repeat or ding acceleration
00:38:46dockimbleit'll have to wait
00:39:53amiconnsaratoga: Ahem, the top loop reads 16 values/iteration while the bottom one only reads 8
00:39:54 Nick Strife1989 is now known as Strife89 (
00:40:06jhMikeSgevaerts: it's works with any function taking a thread id, since those all pass through thread_id_entry inside
00:40:47saratogaamiconn: but the number of iterations should change, correct?
00:40:55saratogaso that each does the same number of loads
00:41:22Buschelsaratoga: compiler optimization? "x" is assigned different values within the same loop.
00:41:43amiconnAh, there's two loops.
00:41:45saratogalet me see what objdump says
00:41:48kugelamiconn: I would like to, though
00:42:00amiconnIt depends on the value of BUF_SIZE
00:42:27amiconnAlso, is buf declared volatile?
00:42:43amiconnOtherwise the compiler will optimize away reads in the 2nd case
00:43:05saratogathe reads happen but the asm looks really weird
00:43:05 Join r2k000 [0] (
00:43:20 Quit GeekShadow (Read error: Connection reset by peer)
00:43:25saratogawhat the hell is gcc doing
00:43:32dockimblethe usb external sound card code just needs some sort of automatic script, so that those of us with mp3 players with busted screens might use them for that
00:43:33amiconnOn arm9 using ldm should give the same speed as plain ldr (slight variation due to instrcution caching is possible)
00:43:38CIA-5New commit by gevaerts (r25463): use THREAD_ID_CURRENT instead of thread_get_current()
00:43:40roolkuugh, what happened to the last build round?
00:44:58saratogahow the hell is doing a bunch of ldr and add opps faster then doing an LDM
00:47:02amiconnBUF_SIZE = 8?
00:47:09amiconnDisassembled the other variant?
00:47:25amiconnBtw, stating "memory" as a clobber is incorrect. You're just reading
00:48:20amiconnAlso I'd use \n instead of ; . Otherwise you might run into the line length limit of the assembler for long asm() blocks
00:48:21saratogaBUF_SIZE is 10KB less then the plugin buffer
00:50:00 Part froggyman
00:52:26saratogahuh just doing a single ldmdb without unrolling gives a huge speed up on PP
00:52:50saratogaalmost 3x as fast
00:53:45saratogaso i must have been miscounting before
00:56:04amiconnThat's what's expected on arm7
00:56:33amiconn4x ldr -> 12 cycles. ldm to 4 registers -> 6 cycles
00:56:57amiconnI guess in the former case gcc also does its silly additions
00:57:00saratogayeah i'm just confused why it didn't work the first time i tried
00:57:39amiconnDisassembling is often the first thing I do in order to understand such things
00:58:00amiconnEven if the loop body is asm, gcc is still able to place some silliness around it
00:58:09saratogai skimmed the dissassembly and didn't see anything obviously wrong, but my arm is pretty rusty so i probably missed something
01:00:02amiconnNot arm related, but when I did the mmx asm for libdemac, I tried using gcc intrinsics first. The result was horrible...
01:00:13saratogakugel: using ldm is much faster on arm7 and no faster on arm9
01:00:17amiconnSo I went for asm()...
01:00:20saratogado you want me to commit it?
01:02:02***Saving seen data "./dancer.seen"
01:02:37 Join jennifur [0] (
01:04:20 Quit jennifur (Client Quit)
01:08:52kugelsaratoga: maybe that also explains the little speed ups from your tremor patch on the test platforms
01:09:19saratogathats definitely some of it, lots of what we did was just using ldm properly
01:10:00saratogabtw when using test_mem, increasing the number of repetitions helps the accuracy a lot
01:11:14 Join merbanan [0] (
01:14:25saratoganow i'm getting the ldm is faster on the clipv2
01:14:27CIA-5New commit by Buschel (r25464): Make test_mem.c and test_boost.c compilable on targets without BUTTON_UP.
01:14:29saratogai don't know whats going on
01:15:21 Part geertvdijk
01:16:47saratogathats my patch if someone else wants to test
01:19:11 Quit merbanan (Ping timeout: 246 seconds)
01:19:20Buschelsaratoga: why don't you just set ptr = &buf[0] outside the inner loop and use ldmia?
01:19:46saratogaBuschel: because I've been randomly trying things until i got something that worked at all :)
01:20:01saratogai'm actually building ldmia right now
01:22:02saratogai get about about 900KB/s and 9000KB/s on the clipv2 with that patch btw
01:24:49 Quit jd (Read error: Connection reset by peer)
01:25:07 Join jd [0] (
01:25:07 Quit jd (Changing host)
01:25:07 Join jd [0] (~jd@Wikipedia/HellDragon)
01:26:23Buschelsaratoga: btw, you can use stmia for writing as well
01:28:50 Quit amiconn (Disconnected by services)
01:28:52 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
01:29:10 Quit pixelma (Disconnected by services)
01:29:11 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
01:29:14 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
01:29:30 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
01:32:08 Quit roolku (Ping timeout: 246 seconds)
01:35:35 Quit anewuser ()
01:40:14 Quit Buschel ()
01:40:54 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
01:41:11 Quit ender` (Quit: All general statements are false. -- The Ultimate Law)
01:42:34 Quit angelwolf71885 (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
01:43:08 Quit kugel (Remote host closed the connection)
01:50:27 Join _deXter [0] (~dexter@
01:55:16 Join kugel [0] (~kugel@rockbox/developer/kugel)
01:55:31Tornegevaerts: well, the io priority thing gets me to the menu ~1 second faster, and to playback ~4 seconds faster, good stuff!
01:56:28Torneinit_tagcache seems to take 400ms regardless of whether load to ram is enabled, with no autoupdate
01:56:55Tornethat happens regardless of the priorities though
01:59:44Torne13 seconds total from pressing power to playback resuming is not too bad :)
02:00:12gevaertsslow bootloaders!
02:00:46gevaerts7 seconds on my gigabeat F
02:01:24gevaertsthat one has the bootloader in flash
02:01:44Tornedunno that it's *all* the bootloader.
02:01:57Torneit's maybe 2.5s from power to logo
02:02:15Tornethe chart shows 394 ticks to get to root menu
02:02:46gevaertsthats about 150 here
02:03:20gevaertsit's a faster CPU of course
02:03:29Torneand a less crappy disk
02:03:31jhMikeSon the beast the bootup is quite a bit quicker. it used to stick on the splash for awhile. I have no times handy though.
02:04:08kugel<5s on a sansa fuze :)
02:04:41gevaertskugel: a sansa fuze is only two seconds faster than my disk-based player? That's *bad*!
02:05:08kugelI haven't tested after your commit actually
02:05:17 Join r2k000_ [0] (
02:05:20AlexPjhMikeS: Beast is pretty slow these days
02:05:31gevaertsit won't change much on flash
02:05:42AlexPEarlier I measured 16 seconds from bootloader to music resuming (before gevaerts' commit
02:05:44 Quit r2k000_ (Remote host closed the connection)
02:05:47AlexPIt is now 8 seconds
02:06:00 Quit dockimble (Quit: Leaving)
02:06:07LloreanAlexP: And surely most of that is OF now.
02:06:08 Join r2k000_ [0] (
02:06:10 Quit bertrik (Ping timeout: 258 seconds)
02:06:14 Quit TheSeven (Quit: ChatZilla 0.9.86 [Firefox 3.6.2/20100316074819])
02:06:17AlexPLlorean: This is from RB bootloader showing
02:06:20jhMikeSAlexP: compared to itself or to new targets that appeared on my absense?
02:06:23LloreanWow, mine never got that slow
02:06:30AlexPAnd the OF section is very quick with single boot
02:06:38LloreanAh, mine's still dual boot
02:06:55 Quit r2k000_ (Client Quit)
02:06:58AlexPjhMikeS: Itself, from what I remember in a fuzzy haven't bothered to measure way
02:07:08AlexPBut I'm not the only one with that impression
02:07:16gevaertsLlorean: this is to music actually playing, not to the wps showing
02:07:28Lloreangevaerts: True, I just boot to the menu then manually resume
02:07:29AlexPUsing headphones :)
02:07:33LloreanSo my experience is surely quite different.
02:07:43AlexPIt was 6 and 9 seconds IIRC to menu appearing
02:07:53AlexPi.e. 6 now, 9 without gevaerts
02:08:16AlexP6.5s with dircache 9.6 s with patch + dircache 6.5 s
02:08:24gevaertsLlorean: me too normally, but I switched it to autoresume for testing
02:08:25jhMikeSAlexP: I use single boot and didn't notice much change w/o dircache from earlier revisions. I've no times either, so long as it feels ok, it's ok imho.
02:08:36AlexPer, that paste was misleading
02:08:54AlexPjhMikeS: Well gevaerts halving the time tonight certainly helps :)
02:09:13 Quit r2k000 (Ping timeout: 264 seconds)
02:09:20AlexP"6.5s with dircache 9.6 s with patch + dircache 6.5 s" should have had "without dircache
02:09:36kugelgevaerts: what theme?
02:10:08AlexPThe measurements I just gave were with settings reset
02:10:19 Quit CaptainKewl (Quit: ( :: NoNameScript 4.22 :: ))
02:10:19jhMikeSAlexP: 'deed. it was easy to tell. I kept dircache off for awhile because of the sheer number of files on my drive.
02:10:34gevaertskugel: I use my own
02:10:52AlexPjhMikeS: Yeah, mine is bound to be a bit slower too - 120 GB disk nearly full of lossy files
02:11:31 Quit jd (Quit: Ω)
02:11:41gevaertshm, why is switching themes so slow?
02:12:48kugelon my fuze it's fast, on my e200 it's rather slow but horrible slow with music playing
02:13:10gevaertsWell, slow as in it's been two minutes now
02:13:21LloreanThat's... that's pretty slow.
02:13:29*gevaerts suspects a bug
02:13:29 Quit bmbl (Quit: Bye!)
02:14:26 Join jd [0] (
02:14:26 Quit jd (Changing host)
02:14:26 Join jd [0] (jd@Wikipedia/HellDragon)
02:15:14kugel5.09s messured
02:15:34gevaertshm, it's not slow. It freezes...
02:15:54 Quit jd (Read error: Connection reset by peer)
02:16:20*gevaerts installs a clean build first
02:18:27jhMikeSgevaerts: I just tried it on beast because that's what on-hand. It took a second or so.
02:18:53gevaertsCan you try while audio is playing? That seems to make it freeze here
02:19:54jhMikeSgevaerts: no problem here
02:19:56*kugel never understood why theme loading is so horrible slow on his e200v1 with music playing
02:20:02kugeleasily 30-40s
02:20:42jhMikeSI just replaced the bin only over a build from a couple days ago
02:23:05 Join r2k000 [0] (
02:25:06gevaertsIt's 100% reproducible here
02:27:11gevaertshm, not 100% maybe, but still
02:27:31jhMikeSany settings that might affect it? maybe I should be more agressive about it?
02:27:54gevaertsdircache doesn't change things
02:29:03gevaertsapart from that I don't see much special
02:30:30 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
02:31:09 Join digitxp [0] (
02:31:19jhMikeSdoes it depend on buffering? also, playback and the disk use the DMA but that probably would crash all the time if that were it. :\
02:31:26digitxphello peoples
02:33:19 Quit pixelma (Ping timeout: 265 seconds)
02:33:20 Quit amiconn (Ping timeout: 265 seconds)
02:33:23 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
02:33:45 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
02:33:51 Join pixelma [0] (quassel@rockbox/staff/pixelma)
02:34:27jhMikeSgevaerts: nope, I can't get a hang.
02:34:47gevaertsI'm now testing r25458 to see if it's my fault
02:34:56gevaertsIt's now stuck
02:35:21digitxpanyone notice that some of the daily manuals return 404's?
02:35:45digitxp(namely sandisks and toshibas)
02:36:46gevaertsjhMikeS: the lcd just turned white and striped, so I'm pretty sure it's not just waiting slightly longer than normal...
02:37:01jhMikeSgevaerts: the core isn't supposed to run above a certain voltage for more than a year accumulated time and so dropping the core voltage has to get done. I hope it isn't just CPU decay from that. :o
02:37:19 Quit Darkknight512 (Quit: ChatZilla 0.9.86 [Firefox 3.5.9/20100315083431])
02:37:45gevaertsjhMikeS: this is on an F. Anyway I doubt if you'd get this sort of effect from that
02:38:09jhMikeSgevaerts: oh! I thought you had an S. ??
02:38:17gevaertsI have both :)
02:38:22jhMikeSme too :)
02:39:35gevaertsWhat I'm doing is just going to Browse Theme Files, and then go through the list one by one while the music is playing. Sooner or later it freezes
02:42:03gevaertsdigitxp: weird. The people who can have a look at it have gone to sleep though
02:42:15jhMikeSok, then I'll get on the same page here and check that out. :)
02:48:00digitxpI want to praise the devs
02:48:15digitxpI get all the thrill of a broken computer
02:48:34digitxpwithout any of the actual problems/risk :)
02:48:42digitxpscratch problems
02:48:53 Quit _deXter (Quit: Lost terminal)
02:49:28 Quit r2k000 (Quit: r2k000)
02:53:03 Quit amiconn (Disconnected by services)
02:53:05 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
02:53:13 Quit pixelma (Disconnected by services)
02:53:14 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
02:53:27 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
02:53:34 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
02:58:56 Quit pixelma (Ping timeout: 265 seconds)
02:59:02 Quit amiconn (Ping timeout: 260 seconds)
03:00:22 Join amiconn [0] (quassel@rockbox/developer/amiconn)
03:00:26 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
03:00:34 Join pixelma [0] (quassel@rockbox/staff/pixelma)
03:02:06***Saving seen data "./dancer.seen"
03:04:27 Quit SirFunk (Read error: Connection reset by peer)
03:04:28jhMikeSgevearts: I went around the list about 8 times on GBF without any problems. With/without dircache.
03:12:19 Join flyback [0] (
03:14:01 Quit amiconn (Ping timeout: 246 seconds)
03:14:24 Quit pixelma (Ping timeout: 265 seconds)
03:14:51 Join SirFunk [0] (
03:17:15 Nick flyback is now known as f1yback (
03:21:35 Quit GeekShadow (Quit: The cake is a lie !)
03:21:35 Join pixelma [0] (quassel@rockbox/staff/pixelma)
03:21:50 Join amiconn [0] (quassel@rockbox/developer/amiconn)
03:25:42 Quit SirFunk (Remote host closed the connection)
03:36:39 Join CaptainKewl [0] (
03:39:03 Quit kugel (Remote host closed the connection)
03:46:56 Join fyrestorm [0] (
03:48:25 Join adnyxo [0] (
03:48:34 Quit adnyxo (Read error: Connection reset by peer)
04:06:15 Quit kramer3d (Quit: Leaving)
04:08:01 Quit z35 (Ping timeout: 246 seconds)
04:09:28 Join z35 [0] (
04:15:22 Quit z35 (Ping timeout: 246 seconds)
04:18:56 Quit Strife89 (Quit: See ya.)
04:19:26saratogalooking at the arm9e reference manual, a lot of our ASM code is probably slower then it needs to be on 9E because of the various pipeline interlocks
04:19:42saratogafortunately most of the code looks pretty easy to fix, just rearranging ops
04:20:25saratogafor instance, the synth filter for mp3 does lots of load then multiply operations over and over
04:20:58saratogathe loads should probably be grouped so that their result isn't needed on the next instructions, which will save a pipeline stall
04:22:23saratogathe mp3 imdct does a lot of that too
04:27:16jhMikeSsaratoga: that is an issue with the s3c and imx31, I know that much. it doesn't seem to hurt the lesser arms to do it.
04:27:43saratogayeah they're not pipelined enough to matter
04:28:20saratogalooking at the timings, arm9E is probably no faster for typical load-mul-store code we do in all our codecs then arm9 because the pipelining isn't taken into account
04:30:14 Join Rob2222 [0] (
04:33:06saratogathe new mdct actually looks pretty good though
04:33:22saratogai guess since its pretty light on multiplies anyway
04:34:17 Quit Rob2223 (Ping timeout: 265 seconds)
04:37:10 Join InterestedRocker [0] (
04:39:07 Quit InterestedRocker (Client Quit)
04:50:16 Part digitxp
04:58:58 Join Barahir_ [0] (
05:02:07***Saving seen data "./dancer.seen"
05:02:13 Quit Barahir (Ping timeout: 240 seconds)
05:07:22 Quit antil33t (Read error: Connection reset by peer)
05:12:59 Join antil33t [0] (
05:20:18 Join phanboy4 [0] (
05:43:01 Quit CaptainKewl (Quit: ( :: NoNameScript 4.22 :: ))
05:47:28 Quit Adubb (Quit: "The mind and body must be subjected to extreme stimulus, by means of drugs and music." - HST)
05:54:40saratogathe D2 is also really, really fast like amsv2
05:55:12 Join Adubb [0] (~aldubuc@
05:56:17saratogathe D2 has SRAM for its internal memory and performs the same or maybe a little slower then the AMSv2
05:56:29saratogastrongly suggests that at very least the IRAM on the AMSv2 has become faster
06:00:32saratogahuh does the D2 actually use its IRAM?
06:05:00saratogait sets it up in the lds file, but it doesn't seem to get linked in the codecs
06:05:26 Join froggyman [0] (~me@unaffiliated/froggyman)
06:07:00 Quit saratoga (Quit: Page closed)
06:07:43scott666is there anyone on who can help me compile a version of rockbox that will work with a 240GB hard drive? I have the dev kit set up already
06:11:48 Join saratoga [0] (~9803c20d@gateway/web/freenode/x-vmdvurtumaznhphy)
06:12:01scott666im going to have to go back to the apple firmware if i cant get it to work :-(
06:12:46scott666ive been using rockbox since i got my archos in like 2005, i really dont want to go back to stock firmware
06:13:15saratogawas there a question in there somewhere?
06:13:30scott666yeah you quit right before it
06:13:31scott666is there anyone on who can help me compile a version of rockbox that will work with a 240GB hard drive? I have the dev kit set up already
06:13:47LloreanWhy not say specifically what problem you're having
06:13:58LloreanThere's no way for anyone to know if they can help you without knowing what the problem is.
06:14:13scott666unsupported max_boot_sector size 4096
06:14:25 Quit fyrestorm (Quit: Ur skills' fireproof like a wooden panel -- U got feds talking leet on your IRC channel!)
06:14:29scott666tried this but it didn't work for me:
06:14:45LloreanIs that a message you're getting, or?
06:14:49LloreanYou need to be a bit more specific.
06:15:05LloreanPeople in here can't read minds, and "it didn't work" doesn't tell much.
06:15:39scott666yes since i installed the default rockbox firmware/bootloader it refuses to boot and gives the error i just said
06:16:15LloreanAnd it still gives the exact same message if you install a modified bootloader and build as per those instructions?
06:16:54scott666the first time i tried, yes. this most recent time it complained about the checksum not matching
06:17:22LloreanThe checksum not matching usually means you've built for the wrong device.
06:19:26S_a_i_n_tthere's only like 3 defines you need to change, its not rocket science.
06:19:26 Nick f1yback is now known as flyback (
06:19:26saratogaI've seen this explained a few times, maybe search the logs
06:19:26scott666well thats what I thought, i found the file to modify and recompiled just fine, but it still refuses to boot rockbox
06:19:26scott666ill try searchign
06:19:26Lloreanscott666: Did you install an updated *bootloader* and not just an updated build?
06:19:26DBUGEnqueued KICK scott666
06:19:26scott666with ipodpatcher.exe
06:20:52S_a_i_n_tthis is iPod 5.5G?
06:21:02scott666no, just 5th. formerly a 60GB
06:21:54S_a_i_n_tis that still called the "Video"?
06:21:57scott666heres a question: do i need to run Make twice (once normal once for bootloader)? or does this happen when you do a normal make anyway
06:22:03scott666yes, not the classic
06:22:17saratogayou can't make a bootloader and a normal build at the same time
06:22:18S_a_i_n_tI'll just send you the config file if you want...
06:22:25saratogaso you don't need to run it twice
06:22:36saratogaalthough you'll have to compile both obviously
06:22:46 Quit panni_ (Read error: Connection reset by peer)
06:22:52Lloreansaratoga: You need to do a separate configure and a separate make, probably best to do it in its own build folder
06:22:57LloreanEr, not saratoga, scott666
06:23:09scott666hmm, maybe thats contributiing
06:23:15scott666ive been doing it in the same folder
06:23:22scott666you need to do make zip for each as well?
06:23:32saratogazip is just the main build
06:23:46saratogait won't actually work for the bootloader since theres nothing to zip
06:24:11scott666it let me do it and it replaced
06:24:16saratogaso you're really sure you actually installed a new bootloader?
06:25:29saratogait sounds like you haven't actually compiled one from these questions
06:25:29saratogawhat does the bootloader version number say when your player boots up
06:25:29scott666yes im sure, ive run ../tools/configure twice to get a bootloader as well as a normal, ive just been doing it in the same build folder
06:25:48scott666and then ipodpatcher.exe with the i copied from the build dir
06:26:01S_a_i_n_tyou *can* build the bootloader/build in the same build folder.
06:26:17scott666i know, ive done it
06:27:00saratogayeah but if make zip didn't error out you probably didn't do it correctly
06:27:11scott666once of the times make complained about an error in fat.c, lemme see if i can get it to do that again
06:29:59scott666okay: In function 'ata_transfer_sectors': error: 'CMD_WRITE_MULTIPLE_EXT' undeclared (first use in this function etc.)
06:30:45scott666this is with #define HAVE_LBA48 in my ipdvideo.h config file
06:31:15saratogabut at some point it did compile with LBA48>
06:33:16scott666i dont think so, i think i just didnt notice the error the first time
06:33:45saratogaso you never actually compiled anything?
06:34:24scott666i commented out that line and just changed MAX_PHYSICAL_SECTOR_SIZE to 4096
06:34:26scott666that didnt fix it though
06:35:18LloreanMaking it clear you didn't actually just use the changes in the link is another one of those things we should've been told at the beginning.
06:35:24LloreanAgain, the whole "can't read minds" thing.
06:35:51LloreanWhat SVN revision are you trying to compile, and do you have any other patches applied to your code?
06:35:52scott666well, i tried to
06:36:06scott666first i tried on fresh svn, now im working with 3.5.1
06:36:11scott666no other patches
06:36:29saratogaalso when i asked if you're sure you isntalled the bootloader, and told you how to check, and you didn't check because you knew better
06:36:38saratogathats one of those things
06:36:44scott666i missed the you telling me how to check part
06:37:24scott666if you like i could take a screen capture of the cmd window i have saying the bootloader was written to the device
06:38:04saratogajust check the version on your player
06:38:07 Quit JohannesSM64 (Quit: WeeChat 0.3.2-dev)
06:38:14scott666forgot to mention that error was from firmware/drivers/ata.c:428
06:38:24scott666my player wont boot rockbox
06:38:36 Join tomers [0] (
06:38:45LloreanThe bootloader version comes before Rockbox boots, which is what he's asking for.
06:39:11scott666sorry, its 3.5.1
06:41:34saratogai think thats the version for the rockbox build, not the bootloader
06:42:19scott666well, maybe thats why it doesnt work
06:42:59saratogaactually if you can't boot rockbox, i guess it would have to be the bootloader
06:43:07saratogadid it say "boot ver" or something like that
06:44:19scott666no, Rockbox boot loader. Version: 3.5.1 Ipod version:0x000B0005
06:44:56saratogathats an odd version string, not sure what it means
06:45:56scott666im building a fresh bootloader
06:46:44scott666is there a way to copy paste out of cygwin?
06:47:07saratogaright clikc and mark, same as windows command lines i think
06:47:09LloreanPlease don't paste stuff into the channel if you do have to paste something
06:47:14LloreanUse or similar.
06:47:30scott666i know, wasnt planning to paste it in this window
06:53:34saratogaif you do an "svn diff" from the root of your check out, are any other files changed?
06:56:12scott666ummm, where exactly do i run that?
06:56:26scott666i got svn: '.' is not a working copy
06:56:34saratogafrom the root of your source check out
06:56:35scott666i meantioned im running 3.5.1 right?
06:57:18Lloreanscott666: Try with a current SVN checkout instead, it will make it easier to check things like this.
06:58:17scott6661 sec
06:59:35scott666the reason i switched from svn to the release waws to rule out any bugs that might have entered since that, i really have no idea how much activity theres been since then though
06:59:49scott666its downloading right now though
07:01:26scott666its amazing how many more targets there are since the last time i compiled
07:02:08***Saving seen data "./dancer.seen"
07:03:17scott666okay, fresh svn download, attemping to Make a new bootloader with the updated ipodvideo.h file
07:03:46scott666hey, it made it past ata.c thats a good sign
07:04:41scott666bootloader went okay, trying normal now
07:13:15scott666no errors, lets hope it works this time
07:16:46 Join r2k000 [0] (
07:18:48 Join nabule [0] (~nabule@
07:19:25 Part nabule
07:19:42 Part r2k000
07:20:26scott666*PANIC* Unsupported physical secor size: 4096
07:20:45 Join r2k000 [0] (
07:21:38saratogapost the output of svn diff and check that the bootloader version is updated
07:21:46S_a_i_n_tAnd you *definitely* changed it to 2048?
07:21:55scott666i used the file you sent me
07:22:04S_a_i_n_tErr..yes, what saratoga said.
07:22:17S_a_i_n_tI thought you said it didn;t come through right?
07:22:38S_a_i_n_tyou said you only had ~130 lines of it.
07:22:43S_a_i_n_tSo that was bound to fail.
07:24:22scott666you sent it a second time, that one worked
07:24:59LloreanOkay, those changes are different from the ones in the linked email
07:25:37saratogaremove all that crap except the LBA changes
07:25:41saratogathen try that
07:25:48LloreanI think the 4096 sector size is necessary
07:25:56LloreanDepending on the actual physical sector size of your drive
07:26:04scott666an email on the mailing list says it isnt but ill try anyway
07:26:10LloreanThat being said, 2048 almost certainly isn't right, and I don't know why you changed it to that.
07:26:19scott666youd have to ask S_a_i_n_t
07:26:25scott666i had 4096 in my file
07:26:31Lloreanscott666: You really shouldn't follow two sets of directions at once.
07:26:48LloreanEspecially making if you don't communicate in here and let us know you're doing other things.
07:26:52 Quit xavieran (Read error: Connection reset by peer)
07:26:56LloreanAgain, "can't read minds"
07:27:03scott666im not trying them at once, im trying them one at a time
07:27:12scott666S_a_i_n_t pmed me which is why you're out of the loop
07:27:18saratogai really think you should ignore everyone but me
07:27:29saratogaand just try the lba change
07:27:39LloreanAt the very least, you should pick one person to listen to and make it clear whose steps you're following.
07:27:53Lloreansaratoga: Those large drives have larger physical sectors, don't they?
07:28:09saratogai see no reason to assume they need that define
07:28:15scott666re-makeing with 4096 as the sector size
07:28:28saratogaafaik its just a work around for the 80GB ipod which doesn't apply here
07:28:56LloreanI think that's a workaround for the 80gb drive, but I think it also applies to the newer 120gb+ drives for 5Gs
07:29:15LloreanBut LBA48 on its own is worth a shot, at least.
07:29:37LloreanWouldn't the obvious answer be to simply find out what the physical sector size is on the drive first?
07:29:39scott666ive tried that, got the same error
07:29:55LloreanI would assume, though, that if the error message is that the size is 4096 and that it isn't supported, enabling support would be essential.
07:30:09scott666again, i assumed the same thing and tried that first
07:30:34scott666thats not to say i didnt make some error though
07:32:16scott666this is what i get when i install the bootloader:
07:32:33scott666says the sector size is 512 which seems odd
07:33:00LloreanLogical sector size and physical sector size are two different things.
07:33:05LloreanThat's probably logical.
07:33:10scott666ah, ok
07:33:14saratogayeah i wonder if it works if you just remove all those defines
07:33:15LloreanDon't know for sure, though
07:33:40scott666ill try that again next, waiting for it to recompile so i can try with 4096 again
07:33:59scott666is there a way to tell it not to compile all these plugins that obviously havent changed?
07:35:26scott666or do i need even need to bother putting a new .rockbox dir on until the bootloader works/
07:35:39LloreanYou should.
07:35:49scott666ok, ill keep doing that then
07:37:18CIA-5New commit by mc2739 (r25465): Add missing entry to project history
07:38:12scott666okay, so next you want me to compile with both the sector defines commented out but LBA48 still in?
07:39:11 Join Boldfilter [0] (
07:41:18 Quit xiainx (Ping timeout: 258 seconds)
07:41:40scott666haha scratch that, it works!
07:42:01 Quit Boldfilter (Client Quit)
07:42:20 Join Boldfilter [0] (
07:42:25scott666thanks guys
07:42:29 Part Boldfilter
07:43:00scott666now i just need my new headphone jack and click wheel to arrive from china
07:43:05 Join Boldfilter [0] (
07:44:17 Quit Boldfilter (Client Quit)
07:44:48 Join Horschti [0] (~Horscht2@xbmc/user/horscht)
07:45:14 Join xiainx [0] (
07:47:42 Quit Horscht (Ping timeout: 240 seconds)
07:51:16 Join FlynDice [0] (
07:54:11 Quit shai (Quit: Leaving)
08:04:12 Join shai [0] (
08:05:44 Part froggyman
08:12:32 Quit S_a_i_n_t ()
08:13:27 Join S_a_i_n_t [0] (S_a_i_n_t@
08:16:25 Quit xiainx (Ping timeout: 276 seconds)
08:16:44 Join xiainx [0] (
08:17:50saratogaAAC-HE takes 185.80MHz on the Nano2G (arm9) verses just 99.27MHz on the D2 (ARM9E)
08:17:59saratogaamazing difference
08:19:00saratogaalthough the Gigabeat F (same CPU as the Nano 2G) decodes it at 127.14MHz
08:19:09saratogaso something is clearly weird about it on the Nano2G
08:20:31S_a_i_n_tIs there any reason why an .sbs shouldn't update at the same rate as a WPS?
08:21:18 Quit pixelma (Ping timeout: 245 seconds)
08:21:25 Quit amiconn (Ping timeout: 265 seconds)
08:31:30 Join drostie [0] (
08:33:36 Join pixelma [0] (quassel@rockbox/staff/pixelma)
08:36:41 Quit pixelma (Disconnected by services)
08:36:42 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
08:37:01 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
08:37:19S_a_i_n_tGah! This is driving me mental! I have a disk access/playmode animation in my .wps that I've moved into my .sbs as well. It alternates at 0.1 second intervals. It works absolutley fine in the .wps, but in the .sbs the update rate seems *WAY* slower and causes drawing issues.
08:38:06S_a_i_n_tIn the wps its a fluid animation, in the .sbs I can see every frame chugging along.
08:39:46 Join amiconn [0] (quassel@rockbox/developer/amiconn)
08:40:13 Quit saratoga (Changing host)
08:40:14 Join saratoga [0] (~9803c20d@rockbox/developer/saratoga)
08:40:16 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
08:42:24 Quit arbingordon (Quit: `)
08:51:11 Quit xiainx (Ping timeout: 240 seconds)
08:55:09 Join xiainx [0] (
09:00:51JdGordonS_a_i_n_t: the sbs doesnt update anywhere near that speed
09:01:02JdGordononce per button press... not 10Hz
09:02:06LloreanIs the D2 software USB?
09:02:12***Saving seen data "./dancer.seen"
09:02:46LloreanAlso, any reason it doesn't have the same volume range fix as the iPod 5G?
09:03:33saratogaits software USB
09:04:08saratogai don't think its 100% working though
09:04:28LloreanWell, it's definitely not enabled
09:04:38LloreanI was wondering if it was disabled because we don't have enough hardware info, or because the main filesystem is read only?
09:04:54saratogaprobably because its not stable enough to be used safely
09:05:33S_a_i_n_tJdGordon: Bugger...well, at least its not some hole in my code though.
09:05:51S_a_i_n_tI take it the update speed of the .sbs is lower for a specific reason?
09:06:09JdGordonwell, its a waste to make it update more frequently
09:06:29S_a_i_n_tnot for my disk access animation it isn't ;)
09:06:52S_a_i_n_tfor the theme I'll never be able to uplaod ;)
09:07:09JdGordonalthough, we may be able to fix that up... let the sbs specify its update freq in the code
09:07:29S_a_i_n_tOhhhhh, sounds *purdy*
09:08:19JdGordonone minor problem is that each screen does its own timeout... in the menus its 1 update per second
09:08:27JdGordonI tihnk the browsers are 10Hz
09:08:34JdGordonthere is no single standard
09:08:43JdGordon(which would be simple to change)
09:10:38 Part scott666
09:15:11 Quit linuxstb (Ping timeout: 240 seconds)
09:16:13S_a_i_n_tAt a guess, it seems like the main menu updates every ~0.3 seconds or so.
09:16:31S_a_i_n_tDoesn't sound like much, but enough to mess up an animation pretty bad.
09:16:54S_a_i_n_tI thought all the screens updated at the same rate, and I was messing up the .sbs code.
09:17:35S_a_i_n_tWell, I know that *now* ;)
09:17:38JdGordonmain menu will update at least once per second
09:17:48JdGordonotherwise once per button press
09:17:55S_a_i_n_tI'm actually kinda dissappointed it's not my fault, it's be easier to fix :P
09:18:26JdGordonno, I lie
09:18:44JdGordonthe sbs does its own checks to not update too often also, which seems like overkill
09:19:08JdGordon#define DEFAULT_UPDATE_DELAY (HZ/7) WTF?
09:19:45 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
09:20:15 Join Highlander [0] (
09:23:48JdGordonS_a_i_n_t: apps/gui/statusbar-skinned.c line 138... comment that out and let me know if it looks better
09:25:46S_a_i_n_tOK, will do.
09:26:12JdGordonno rush of course.. I'll be afk till tuesday night :/
09:26:23JdGordonbut I tihnk getting rid of that update check makes sense
09:26:39JdGordonespecially if it doesnt hit the battery (which going by my bench's it wont)
09:27:13JdGordonyou'll notice a change if you scroll through the list and your animation seems to speed up
09:27:47S_a_i_n_tI think the assumption was that it probably *would* drain the battery updating too much, but if your theme is *full* of animation...(like mine), you probably don;t care much about battery life ;P
09:27:52 Quit xiainx (Ping timeout: 258 seconds)
09:30:18 Quit linuxstb (Ping timeout: 276 seconds)
09:31:30 Quit phanboy4 (Ping timeout: 265 seconds)
09:32:32 Join xiainx [0] (
09:33:43JdGordonS_a_i_n_t: sure, but it looks like a basic sbs with just the inbuilt bar uses no more/less batt life than a full cabbie with AA
09:33:55 Quit tomers (Ping timeout: 265 seconds)
09:35:50S_a_i_n_tWow, how the heck does that work?
09:36:07S_a_i_n_tIt just doesn't seem right,
09:36:54 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
09:39:35CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
09:39:35*JdGordon starts his last bench
09:40:12 Quit JdGordon (Quit: Leaving.)
09:47:26 Quit xiainx (Ping timeout: 268 seconds)
09:48:06 Join xiainx [0] (
09:52:13 Join flydutch [0] (
09:57:18 Join stoffel [0] (
09:58:15 Quit linuxstb (Ping timeout: 276 seconds)
10:02:36 Join ender` [0] (
10:08:50 Join linuxstb [0] (
10:08:56 Quit linuxstb (Changing host)
10:08:56 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
10:08:56 Quit xiainx (Ping timeout: 260 seconds)
10:13:48 Join xiainx [0] (
10:20:15 Join roolku [0] (
10:25:45 Quit xiainx (Ping timeout: 264 seconds)
10:26:35 Quit evilnick (Ping timeout: 265 seconds)
10:27:34 Join grndslm [0] (
10:28:48 Quit linuxstb (Ping timeout: 276 seconds)
10:29:47 Join xiainx [0] (
10:31:20 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
10:32:39CIA-5New commit by alle (r25466): Make the displayed hotkey settings scroll (fix FS #11170)
10:36:51 Quit linuxstb (Read error: Operation timed out)
10:40:50 Join Kitr88 [0] (~Kitr88@
10:41:57 Quit xiainx (Ping timeout: 264 seconds)
10:44:21 Quit Kitar|st (Ping timeout: 264 seconds)
10:45:20 Quit Kitr88 (Ping timeout: 260 seconds)
10:47:09 Join xiainx [0] (
10:50:29 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
10:50:52 Join Kitar|st [0] (
10:57:04 Join merbanan [0] (
10:59:35 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
11:02:16***Saving seen data "./dancer.seen"
11:02:57 Quit bluebroth3r (Ping timeout: 264 seconds)
11:08:24 Join stripwax [0] (
11:08:38 Join Xerion [0] (
11:25:19stripwaxsaratoga - dropping-in the patched libvorbisidec in an environment that had the unpatched one previously, seems to cause application errors for apps that use this - in particular, mpd complains with "symbol lookup error: /usr/local/lib/ undefined symbol: LIKELY"
11:25:35stripwaxa rebuild of mpd would probably fix, but is that expected?
11:25:48saratogastripwax: i have no idea
11:26:19stripwaxIs the new lib supposed to be compatible with the existing one?
11:26:26stripwaxHm, I guess not actually.
11:26:37saratogathat likely error sounds like a gcc problem
11:26:44 Join kugel [0] (~kugel@rockbox/developer/kugel)
11:26:50saratogasince gcc should understand what likely means, its one of its macros
11:26:52 Quit Xerion (Ping timeout: 246 seconds)
11:27:00stripwaxgcc? it's not a compile-time error, though ....
11:27:21saratogait sounds like it interpreted likely as a function define?
11:27:29stripwax"make example" worked fine though
11:27:30kugelS_a_i_n_t: the update rate of the sbs is limited in the browsers, otherwise browsing through the menus might be a pain in the arse
11:27:45stripwaxunless the compilation of the dynamic libraries is broken here
11:27:55 Quit xiainx (Ping timeout: 246 seconds)
11:28:20stripwaxin any case, LIKELY is #defined to be empty
11:28:36stripwaxI'll try a clean make/rebuild and see what happens ..
11:28:45saratogamaybe we should just remove the likely macros
11:28:53saratogai doubt they make a measureable difference and they're gcc specific
11:29:39stripwaxsaratoga - hm, but os_types.h *already* #defines LIKELY to be empty.
11:29:52 Quit r2k000 (Quit: r2k000)
11:30:00kugelsaratoga: LIKELY is not a gcc macro, it's a wrapper macro for something that resolves to a gcc built-in
11:30:05saratogais it possible your app was compiled against some much older version of tremor?
11:30:51kugelin any event, it shouldn't survive the preprocessor, so something is wrong when it's in the binary
11:31:18stripwaxnot sure what you mean by 'much older' - has there been any recent releases of tremor? the app was previously compiled against the headers from the nslu2 package, fwiw.
11:31:28stripwaxwhich is afaict the same as Tremor trunk
11:32:32kugelI suppose if os_types.h #defines LIKELY then it's not included properly in all places where it's used
11:33:24stripwaxah, good point...
11:33:51 Join xiainx [0] (
11:34:22saratogawe still need to try the original ASM'ed mdct too to see how much of a difference it makes
11:34:42stripwaxin the patched code, only fft-ffmpeg.c and mdct-ffmpeg.c use LIKELY −− and not clear if either of them #include os_types.h ...
11:35:02stripwaxthe original rockbox ASM'ed one?
11:35:47kugelstripwax: it should be included directly if it uses something in there anyway, imo
11:36:05stripwaxkugel - right, i'm not sure that it does
11:36:07saratogayes the original one, i want to see how it perfoms in tremor
11:36:35stripwaxsaratoga - i was actually running the original rockboxed one on the nslu2 for quite some time :) plus a few other bits and pieces
11:36:48stripwaxlet me try building the libtremor_example for the original code
11:36:49saratogafeel like benchmarking it?
11:37:06stripwaxI have no idea if it's the 'latest version' in any meaningful sense, but I will try
11:37:17stripwaxI'll use the same testfile as last time
11:39:47*stripwax wonders if his LIKELY problems are actually due to the *old* rockbox mdct on his nslu2!
11:40:00 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
11:40:42stripwaxmaybe my mpd build is actually *expecting* LIKELY to be defined (due to some cruddy bug on my side)... will know soon enough. But yeah, we should just remove the whole LIKELY stuff from the tremor patch
11:42:16kugelstripwax: no, it's libvorbis
11:44:06 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:44:12kugelthe dynamic linker doesn't know where symbols come from, so it can't show that a missing symbol is missing in one of the linked libraries
11:44:49kugelit can only show in which binary/lib the symbols are missing
11:45:18stripwaxright, so i think my mpd build is dodgy if it's trying to load a symbol named LIKELY from the lib
11:46:05stripwaxanyway, sorry, I'm veering offtopic a bit here. I'll get some benchmarks from the patched tremor vs rockbox mdct tremor in a sec.
11:46:21kugelthen it would show symbol "lookup error: /path/to/mpd: undefined symbol: LIKELY"
11:46:54kugeldid you try simply including os_types.h where LIKELY is used?
11:47:02stripwaxi removed LIKELY from the code and rebuilt ...
11:47:08stripwaxand it still happened ...
11:47:57kugelbut from what I've noticed, all mpd defines (the gcc macros too) are lower case and prefixed with mpd_ so it seems unlikely
11:49:30stripwaxright, that's why I think it's a problem with my local mpd, which was originally compiled against a manually-patched version of tremor with the rockbox mdct code. I probably screwed up, that's all.
11:49:32 Quit xiainx (Ping timeout: 268 seconds)
11:50:05stripwaxI originally thought it was a problem with the patch; now I think it's a problem with me instead.
11:52:32 Join xiainx [0] (
12:00:58 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
12:02:01 Quit roolku ()
12:03:17 Quit xiainx (Ping timeout: 248 seconds)
12:08:40 Join xiainx [0] (
12:12:17stripwaxsaratoga - i see the old rockbox mdct coming in at 38.42MHz and the new one coming in at 32.8MHz - but that's not quite a like-for-like comparison ; you don't happen to have a 'clean' patch against tremor trunk to add in the rockbox mdct and nothing else?
12:13:01stripwaxnemmind, not hard.
12:13:04 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
12:13:30saratogastripwax: the aatch is basically just the rockbox mdct and windowing code
12:13:38saratogaand i doubt the windowing code makes much difference on arm9
12:14:48stripwaxI think there must be something else kicking around in my local version, because 38.42Mhz sounds slower than the reports for the unpatched trunk version
12:15:27stripwaxunless the old rockbox mdct asm is slower than the trunk C, which I find hard to believe.
12:16:13saratogadid your old version have arm_assem defined?
12:16:27saratogathat'll make other parts slower even if the mdct gets faster from your changes
12:21:37stripwaxhrm, -fomit_frame_pointer couldn't cause that sort of symptom, could it? (my old build had that defined, but the new one does not)
12:22:35saratogayou could also try commenting out the call to the mdct in each, just to see how many Mhz the mdct itself uses
12:23:21stripwaxi'll just add the old rockbox mdct into the same code as your patched version so I'm actually comparing apples to apples.
12:23:22 Quit xiainx (Ping timeout: 245 seconds)
12:25:59saratogamight as well put that online somewhere in case anyone wants to try it
12:26:42stripwaxi'll also make a copy of the mdctexp tremor patch so that it includes just the new mdct..
12:28:29stripwaxotherwise too many moving parts for a meaningful comparison of just the mdct
12:29:40 Join xiainx [0] (
12:31:21 Join Meekrab [0] (
12:31:59MeekrabOk so I have a few questions.
12:32:07MeekrabI havea Gigabeat S and a Zune.
12:32:42MeekrabHas anyone figured out the protection on the zune yet?
12:33:35MeekrabI'm pretty good at programming but I have no Idea the differences between the zune's bootloader and the GS's bootloader...
12:33:56saratogaone has a security flaw the other doesn't
12:34:33Meekrabfrom my settop box hacking days, I see that my zune will load an nk.bin and unpack it into ram... but if the signature is incorrect it quickformats the HDD and asks you to plug it in.
12:35:27MeekrabI may have found a way to wrap the NK.bin with some arbitrary code (ala xbox hack) and get rockbox working that way, but alas it may require dissasembly of an official microsoft firmware...
12:35:55saratogahow did you find a way if you haven't done any dissassembly?
12:36:57MeekrabI did
12:37:53MeekrabI know for a fact that the Gigabeat S runs a Wince5 kernel but the IMX31L will only run signed code...
12:38:58Meekrabsofar the only two ways to legitimately run anything non MS on the zune is by either jtagging it or trying to find a buffer overflow
12:39:31Meekraband thats like finding a needle in a haystack because the zune firmware is pretty nicely done, concidering its microsoft
12:39:35saratogaso what did you find?
12:40:01MeekrabA veeeerryy small exploit right as it looks for the signature of the NK.bin
12:40:22Meekrabsomething on the lines of a disc swap if you will
12:40:45Meekrabif we can find a way to load the SHA256 key on top of rockboxes it may load up fully
12:41:06 Quit stoffel (Ping timeout: 252 seconds)
12:41:52MeekrabOh yeah and another interesting note
12:42:12 Quit kugel (Ping timeout: 252 seconds)
12:42:37MeekrabI decided to plug in the MK3008GAL Hard disk from my dead ipod video into my Zune, it formatted it automatically and told me to plug it into the computer as if it were brand new.
12:43:27Meekrabso it seems that if it doesnt find a certian set of files it just executes a delete MBR and waits for input.
12:43:56n1sthe gigabeat s reformats the disk if it doesn't like the partition layout
12:44:11linuxstbMeekrab: You may want to speak to Torne next time he's around. He's done a lot of investigation into the Gigabeat S, to try to find exactly why it spontaneously decides to reformat.
12:44:16n1sand also sometimes for no obvious reason
12:44:33Meekrabit seems it has a hard coded SHA256 in the binfiles
12:44:43 Join kugel [0] (~kugel@rockbox/developer/kugel)
12:44:52linuxstbHe's made some notes here -
12:45:26Meekrabthe spontaneous reformats seem to be the results of fragmentation because the Kernel may be written to by accident under certain circumstances...
12:47:31 Quit xiainx (Ping timeout: 276 seconds)
12:47:40MeekrabThere should be a way to make the menu+up point to an NK.bin which points to another binary in the media partition....
12:49:05MeekrabToo bad I dont have a zif to 2.5" adaptor. otherwise I could take a look at the contents of my zune's HDD
12:50:25 Quit saratoga (Quit: Page closed)
12:51:51 Join stoffel [0] (
12:52:42 Join xiainx [0] (
12:53:10MeekrabGrr if i only knew more ARM stuff
12:53:29MeekrabI know how to run arbitrary on X86 wince, like settops or the like...
12:56:20MeekrabHmm, Is it possible that the zune uses Wince6?
12:59:37linuxstbMeekrab: I don't think anyone here has ever looked at the Zune, or has much experience with WinCE. Quite often, one person works out how to hack a device, then others do the actual work to write drivers and port Rockbox. That was the case for the Gigabeat S.
13:00:07MeekrabWell, anyone want to help me crack this nut?
13:02:14Meekrabtheres a 20pin header on the backside of the planar
13:02:19***Saving seen data "./dancer.seen"
13:02:27linuxstbMeekrab: The people who I think may be interested are not around at the moment. So please stay around, or check the logs later -
13:03:16 Quit stripwax (Quit:
13:03:27MeekrabWhy didnt I think of this beofe
13:03:32MeekrabThe wifi unit
13:03:44Meekrabsurely theres bugs there
13:04:01Meekrabor perhaps the RDS?
13:05:22MeekrabHowbout TIFF?
13:05:38Meekrabthey used it on the PSP to run hello world
13:05:48linuxstbDo you have access to everything for disassembly? i.e. the bootloader plus the main firmware?
13:05:58Meekrabnot everything
13:06:13MeekrabI can see the boot process and loading into ram but I cant poke anything in ram
13:06:26Meekrabjust over jtag
13:07:03Meekrabmain firmware can be downloaded from the net and dissasembled, but I cant see it in actual motion due to some protection
13:07:43 Join stripwax [0] (
13:09:51stripwaxsaratoga - on nslu2, the old rockbox mdct_arm.S doesn't seem to bring any advantages.
13:10:15Meekrabdoes the zune use the SGI implementation libTIFF?
13:10:16stripwax(over the non-asm stock C functions)
13:10:35stripwaxMeekrab - how would we know? ...
13:10:46*stripwax is guessing that we don't know
13:10:48Meekrabusually by looking in the credits menu...
13:11:00stripwaxMeekrab - do you have a Zune?
13:11:24TorneMeekrab: i think you misunderstand the security here, also ;)
13:11:27Meekrabmy zune is currently at the "plug into computer" screen
13:11:35Tornethe i.MX31 part in the Gigabeat S is not in secure mode
13:11:42Torneand I would be surprised if the one in the Zune was
13:13:16Torne11:45 < Meekrab> the spontaneous reformats seem to be the results of fragmentation because the Kernel may be written to by accident under certain circumstances...
13:13:20stripwaxsaratoga - I get 8m0s to decode a 60minute track with unpatched trunk tremor from; 8m1s (ish) to decode using the old rockbox mdct_arm.S; 7m24s (ish) to decode using the FasterMDCT patch (which includes non-imdct stuff)
13:13:20Tornealso this is nonsense
13:13:30Tornewe don't even mount the firmware partition
13:13:48Torneand fragmentation is irrelevant
13:13:49stripwaxsaratoga - trying a few runs to even things out of course
13:14:54Tornethe problem with the gigabeat S is that the bootloader implements the error handler as "reformat the drive"
13:15:07Torneso if absolutely *anything* goes wrong during boot, it will kill it, pretty much ;)
13:15:12Meekrabwhere did "Kernel may be written to by accident under certain circumstances..."
13:15:15Meekrabcome from lol
13:15:22MeekrabI must have pasted that
13:15:29Torne(including things which can't possibly be fixed by formatting the drive, e.g. flash corruption)
13:16:10Meekrabwhat I meant to say was the error handling would detect this as something up and trying to fix it and asking for a reformat.
13:16:26Meekrab*by asking
13:16:33Torneyes, but we *don't* write to the firmware partition
13:16:48Tornethough it gets exposed over usb, so someone could accidentally touch it that way
13:16:51Meekrabim sorry, its 5AM and I've been tearing apart my brain boneing up on ARM terminology
13:17:10Tornebut it's too widespread a problem for that to be it, really.
13:17:30Tornesomeone already linked you to the wiki page with my research so far on the S's loader
13:17:39Meekrabwhat about locking the drive in some sort of read only state?
13:17:54stripwaxwhat does that mean?
13:18:24Meekrabprevent the drive from formatting itself, to see what it would do in the event of a failed format.
13:18:48TorneWe know what it does
13:18:50TorneIt hangs.
13:19:01stripwaxbut readonly vs readwrite is a software thing not a hardware thing, surely. and if the gigabeat bootloader, which we didn't write, does something in software, we can't prevent it anyway.
13:19:05TorneThe format is controlled by a flag in the flash config area
13:19:14Torneyou can turn it off, and it won't format any more.
13:19:25Tornebut all it does is skip formatting :)
13:19:37Meekrabdoes it hang into some sort of wait or does it halt and catch fire
13:19:41Tornewhatever fatal error caused it to decide to want to still applies.
13:19:49Torneit still tries to jump to recovery.bin
13:20:19Torneso it doesn't work any better..
13:20:29Torneit just means you have to take the disk out and fix it externally.
13:20:39Torneso, that's unlikely to be a very interesting avenue to explore.
13:21:10 Join dacari [0] (
13:21:14 Quit dacari (Client Quit)
13:21:36Tornethe code for pmcboot_secure is not very complicated, really.
13:21:40 Join tredstone [0] (
13:21:47Tornethere's not a huge scope for them to have done something wrong.
13:22:05Tornethe version on the beast doesn't range check the load addresses for the chunks of nk.bin, which is how we run on that
13:22:17Torneyou just overwrite the instruction that fails the signature check with one that always passes :)
13:22:33Tornebut my understanding (though I haven't looked) is that the Zune's version of pmcboot does have proper bounds checks on this
13:22:52Torneso you can't overwrite any of the ram the bootloader is using.
13:23:03Meekrabthats what I was saying eariler
13:23:14MeekrabI cant look at any of that ram
13:23:28soapjhMikeS, could I have a clarification of context on your quote from ~12 hours ago: "<jhMikeS> gevaerts: the core isn't supposed to run above a certain voltage for more than a year accumulated time and so dropping the core voltage has to get done. I hope it isn't just CPU decay from that. :o"
13:23:58Torneyou have a jtag connection?
13:25:30Meekrabyeah, that 20pin connection between the planar and HDD, pins 4,5 are Serial RX and Serial TX, TRST - 8
13:25:31MeekrabTDI - 9
13:25:31MeekrabTMS - 10
13:25:31DBUGEnqueued KICK Meekrab
13:25:31MeekrabTCK - 11
13:25:31MeekrabTDO - 13
13:25:42 Quit tredstone (Client Quit)
13:26:07Tornecan you see the core?
13:26:38Meekrab6 is CMOS level voltage (3.3~3.2v)
13:26:56Meekrabground is 17-20
13:27:18Meekrabnope, its protected
13:27:31Torneright, so that's useless then :)
13:27:43Torneif icd is fused off then yeah, you can't do anythig useful with it
13:29:08Tornei assume you at least have a copy of pmcboot_secure.bin for the zune
13:30:05Torneso, get disassembling :)
13:30:38MeekrabPMCR0 = 0x00000000
13:30:42Meekrabthat doesnt make sense.
13:34:14 Quit xiainx (Ping timeout: 276 seconds)
13:34:28Meekrabin the boot sequence, it tells me the PMCR0 is blank, the Dword is 01 and then it inits the fm tuner
13:37:23Meekrabits weird
13:37:30Tornewhat's weird about it?
13:37:31 Join xiainx [0] (
13:37:39Tornezero is a valid value for that register
13:37:58Meekrabbut that doesnt seem right
13:38:00Torneif DVFEN and DPTEN are zero all the other bits are irrelevant
13:38:02 Quit stripwax (Quit:
13:39:43Tornealso this debug output is pretty arbitrary and random, you realise
13:40:15Torneif it's similar to the beast port, then what trace is enabled/disabled is fairly arbitrary :)
13:40:31Meekrabits like drawing a card
13:41:53MeekrabSysInit: GDTBase=811d40e8 IDTBase=811f5860 KData=811fd800
13:41:53MeekrabWindows CE Kernel for i486 Built on Jun 24 2004 at 18:23:42
13:41:53Meekrabg_pPageDir = 81200000
13:41:53MeekrabX86Init done, OEMAddressTable = 80223f98.
13:42:18MeekrabJe246, yeah its just the mamba bootloader
13:42:24Meekraboops wrong channel
13:42:40Torneer what?
13:42:57Torne(also please don't paste)
13:43:23Meekrabwas just explaining the i486 bootloader from the mamba hardware settop box (ip1101)
13:44:18MeekrabI'm tired. and I'm out of coffe,
13:44:53 Quit Meekrab (Quit: Want to be different? Try HydraIRC -> <-)
13:45:16 Join stripwax [0] (
13:49:09kugelhm, I have fuzev2 scrollwheel with interrupt working but it's too sensitive for my taste :(
13:50:14 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
13:52:05stripwaxsaratoga - yep, after a few more time trials, I don't see the rockbox old mdct_arm.S giving any significant speedup on nlsu2
13:52:08 Quit xiainx (Ping timeout: 264 seconds)
13:52:33stripwaxapprox 8m0s for both stock and rockbox old mdct_arm, to decode 60minutes at 266MHz
13:52:42 Quit stripwax (Quit:
13:57:09 Join xiainx [0] (
14:00:00 Quit linuxstb (Ping timeout: 258 seconds)
14:02:16 Quit Highlander (Quit: Quitte)
14:06:44pixelmaAlexP: what did you change \\ to? I removed it and there's now only a blank line between paragraphs in the code. The PDF output has paragraphs (indented) but no blank lines between them, is there some trick to get one (except \\ obviously) or should I just ignore it?
14:07:45 Quit shai (Quit: Leaving)
14:07:54AlexPpixelma: You have to change the paragraph rules in preamble.tex, but that changes the whole document, so at the same time you need to find/change all the \\ \\* needed in all the source
14:08:16AlexPpixelma: For now I'd use \\, and when I get a bit of time I'll do them all at once
14:08:21AlexPOr someone else can :)
14:08:22pamaurypixelma: what do you want to achieve ?
14:08:39AlexPpamaury: Blank lines in between paragraphs
14:08:39 Quit piotrekm (Read error: Connection reset by peer)
14:08:42pixelmaremove \\
14:08:55 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
14:09:19pamauryBetween all paragraphs ?
14:10:51pamauryYou can try \setlength{\parskip}{<size>}
14:11:25pixelmaAlexP: ok, I'll add them in the introduction text at the beginning again because it looks a bit unstructured without. But there were some completely unnecessary ones in the menu item description that had no effect at all
14:11:57AlexPpamaury: I know, but thatchanges the whole document so you have to go through and remove all the current forced breaks
14:12:09AlexPpamaury: It isn't difficult to do, it just needs a bit of time
14:12:19pamauryAh indeed ;) just do replace \\ by \par
14:13:31 Join Annaa [0] (~Annaa@
14:13:34Annaa do my breasts look to big?
14:13:35 Part Annaa
14:15:39 Quit pixelma (Ping timeout: 240 seconds)
14:16:29 Quit amiconn (Ping timeout: 258 seconds)
14:16:36 Quit stoffel (Ping timeout: 246 seconds)
14:17:49 Join pixelma [0] (quassel@rockbox/staff/pixelma)
14:17:55 Join amiconn [0] (quassel@rockbox/developer/amiconn)
14:19:32 Quit Rob2222 (Ping timeout: 265 seconds)
14:21:30 Join Rob2222 [0] (
14:24:18pixelma"debugging" the broken build was funny... I couldn't find out first because the error appeared later in the main rockbox.tex which is just a list, then changed something else which was unrelated but which I wanted to change anyway - then the error output was different and (even though not in my new changes) helped me find the mistake...
14:34:41 Quit pamaury (Quit: Page closed)
14:35:13 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
14:38:18S_a_i_n_tkugel: Sorry its so late. I've commented out the line that JdGordon told me to (sanity check for the .sbs refresh rate?), and nothing seems fishy with the menus so far.
14:38:27S_a_i_n_tAnimation is running a lot better though.
14:38:56kugelsure, but there are targets with slower cpus and displays where it's getting annoying
14:39:03 Join shai [0] (
14:40:58S_a_i_n_tAha...he (JdGordon) said it may be worthwhile looking into a way for the .sbs to specify its own refresh rate.
14:41:09S_a_i_n_tFor me at least, it could come in handy.
14:43:20 Quit merbanan (Ping timeout: 240 seconds)
14:43:50S_a_i_n_tI'd never actually noticed the refresh rate was different until I tried putting an animation that alternates frames at 0.1 second intervals. It made for some pretty ugly drawing issues.
14:45:46kugelwithout blocking some updates the sbs is updated on every button press, meaning to go through the whole lot if the skin engine and multiple screen updates for every selected item in the lists
14:46:07kugelin the wps, the sbs refresh block is removed so it updates always when the wps is updated
14:47:36kugelwe could consider allowing 10 updates per second but I would rather not remove the fresh rate limitation
14:48:42S_a_i_n_tten updates per second would do for animation I guess... the animation I had was at 0.1 second intervals, but seemed to update every ~0.3/4 seconds.
14:48:48S_a_i_n_tAnd looked really awful.
14:49:35kugelI wonder how the wps deals with such frequent updates, IIRC it only updates every 200ms (unless you press a button)
14:49:52S_a_i_n_tYourself and JdGordon seem to be the "GUI guys" so, its up to you guys if its worth it or not ultimately I guess.
14:50:12kugelI guess 0.2 timeout would not work for you?
14:50:26S_a_i_n_tIt seems fint (at 0.1second) on the WPS, unless its just taking the lowest value it can.
14:51:19S_a_i_n_t0.2 would be better than "at least once per second, or at button press"
14:51:32S_a_i_n_tI could just alter the WPS animation to the same speed.
14:51:57kugelthe wps button max timeout is 200ms, it only updates on timeouts (and a timeout happens if a button is pressed or the max timeout is exceeded). the only way for the wps to update more often is if it contains a peakmeter
14:52:34S_a_i_n_tHmmm, so even if I set it to 0.1, its still at 0.2?
14:52:48S_a_i_n_twell, if that works for the WPS, then the .sbs should be fine with it.
14:53:40kugelI don't know for sure, but I would be clueless if 0.1s actually works for the wps ;)
14:54:17S_a_i_n_tit seems to...but I guess the difference between 0.1 and 0.2 seconds is barely perceptable ;)
15:02:21***Saving seen data "./dancer.seen"
15:03:53pamauryis there a way to skip of database refresh on e200v1 ?
15:05:23pixelmaRockbox should already do this if the bit that needs to be zeroed is known for your exact OF version (I believe SanDisk changed it from time to time)
15:07:55 Quit parafin (Quit: So long and thanks for all the fish)
15:08:12 Join parafin [0] (
15:10:39 Join paulk_ [0] (
15:10:48kugelpamaury: you boot the OF?
15:11:21bertrikpamaury, pixelma I think this doesn't always work, IIRC you need to clear some configuration block to make sure the refresh bit is always in the same location
15:11:23paulk_Hello ! I can't found the archive with the rb 3.5.1's source code on . Is that normal ?
15:11:43pamaurykugel: I'm experimenting with usb, so sometimes yes, to recover :)
15:11:48kugelpaulk_: ^
15:11:55paulk_ok, but why ?
15:12:10kugelbecause there's svn
15:12:30paulk_kugel: it's better to direct download it
15:13:01kugelI don't disagree, and several people asked for it, but nobody ever put a source archive back online
15:13:56kugelsvn has several advantages though, so I'd go with it
15:14:44paulk_ok, but I can't find 3.5.1 on svg, just 3.5.
15:15:35bertrikpamaury, see rbutil/sansapatcher/sansapatcher.c line 917
15:16:25paulk_kugel: thanks, I was looking at
15:17:46kugelreleased stuff is always under tags
15:21:54 Join Schmogel [0] (
15:28:13pamaurybertrik: thanks, I'll have a look. that's not very important, however, I had forgotten that I had put 40000 files on my e200 to do some test and the OF database fresh is quite slow :)
15:39:47 Join liar [0] (
15:39:49 Join adnyxo [0] (
15:40:03 Join mitk [0] (
15:52:40CIA-5New commit by pixelma (r25467): Cleanup goban.tex. My main goal was to get the button table code into a readable state - following the structure I suggested on ManualGuidelinesTalk. ...
15:54:36 Join Sajber^ [0] (~Sajber^
15:54:47gevaertspaulk_: has the source
15:55:01gevaertskugel: we do have the source for releases
15:55:21paulk_gevaerts: ok, but they're not on the web page
15:55:38 Quit Sajber^ (Read error: Connection reset by peer)
15:55:53gevaertshm, looks like a bug
15:57:01CIA-5New commit by pixelma (r25468): Refine the file formats table which explains what plugins are used to open which files so that the table is only included in manuals of targets which ...
15:58:54CIA-5New commit by pixelma (r25469): Remove the Recorders and Players opt around the A-B part of the repeat mode tag because all targets can use A-B repeat now (and show this in the WPS).
16:02:30kugelgevaerts: oops, my bad
16:04:27 Quit xiainx (Ping timeout: 276 seconds)
16:06:39gevaertshm, I suspect that the freeze I was seeing yesterday is related to album art, or (more likely) sbs, or a combination of both
16:15:03CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
16:15:03*pamaury grumbles about control transfer handlers being called from interrupt handler but will soon be obliged to do the same with iso transfers :(
16:25:35 Join funman [0] (~fun@rockbox/developer/funman)
16:30:36 Quit blairb (Quit: Leaving)
16:32:37kugelfunman: FS #11172 :)
16:34:22 Join veeloc [0] (
16:36:16funmancool! (bookmarking it)
16:36:17 Nick krazykit` is now known as krazykit (
16:36:24 Join xiainx [0] (~xiainx@
16:39:24 Quit veeloc (Quit: is idle)
16:43:50kugelfunman: you're supposed to test it, not to bookmark :p
16:44:30 Quit pjm0616 (Quit: reconnecting)
16:45:07funmanright :)
16:45:08 Quit funman (Quit: free(random());)
16:46:15 Join pjm0616 [0] (~user@
16:56:32paulk_What file contains the Fm Radio's menu ?
17:00:12 Join robin0800 [0] (
17:00:41 Join robin0800_ [0] (
17:02:22***Saving seen data "./dancer.seen"
17:02:29 Quit robin0800 (Remote host closed the connection)
17:03:43gevaertspaulk_: I would guess apps/recorder/radio.c
17:04:13paulk_gevaerts: this file just contains : do_menu(&radio_settings_menu, NULL, NULL, false)
17:04:21gevaertsradio.h then?
17:04:45gevaertswhy don't you just grep for radio_settings_menu ?
17:05:10paulk_gevaerts: I think I'll do that
17:05:40*gevaerts wants lru_sort()
17:07:54paulk_I found what I was searching for !
17:14:36kugelgevaerts: isn't an lru sorted?
17:14:54gevaertskugel: well, yes, but not in a useful way for some operations
17:15:00 Quit xiainx (Ping timeout: 276 seconds)
17:15:21paulk_I just had an idea : I saw that the record button is unused for the radio screen, so I set-up rockbox to launch the record screen with radio as source when the record button is pressed. It's faster than "SELECT | BUTTON_REPEAT > Recording".
17:15:23gevaertshm, it's probably not that simple
17:16:42gevaertskugel: the problem I'm looking at is the slow font loading at boot, which is (at least partly) caused by the glyph cache being in lru order, which means that actually loading the glyphs from the font file is horribly slow on disk targets
17:17:11gevaertsSorting the lru cache by glyph id would fix that, but of course make the loaded cache be in the wrong order...
17:18:08kugelare sure?
17:20:17gevaertswell, it looks like it anyway
17:21:28kugelisnt the problem the lots of small reads rather than the order?
17:21:58gevaertsI don't think so
17:22:44 Join goffa [0] (~goffa@
17:23:14goffaloaded the new rockbox on my gigabeat s last night ...
17:23:34 Join xiainx [0] (xiainx@wpa106012.Wireless.McGill.CA)
17:23:36 Join froggyman [0] (~me@unaffiliated/froggyman)
17:23:45AlexPA slight difference to the OF
17:23:47goffawhen i try to generate a new dir list with the random folder advance plugin
17:23:52goffathe thing crashes
17:23:59goffasame if i try to make a playlist
17:24:08goffawondering if my hard drive is dying.. or if there's a bug
17:25:14goffaon the random folder advance it displays *panic* stokov main
17:25:34AlexPWhat version?
17:25:44goffalet me check
17:25:52gevaertshow deep is your directory structure?
17:26:10soapgevaerts, not that I'm suggesting it for commit - but would simply bumping up the size of the glyph cache prevent the slower loading (which IIUC is caused by loading specific glyphs out-of-order?)
17:26:12kugelgevaerts: Torne sounded like that
17:27:07gevaertssoap: as far as I understand it, that will prevent the slower loading for fonts that entirely fit the new size
17:27:18gevaertsin other words, it won't ever work for unifont
17:27:23goffagavaerts its a few deep... probably 5 dirs at times
17:28:00goffalike music\artist\album\cd1\possible subdir ... that's about max
17:28:02soapunifont is bigger than the max font cache size one could specify? Is that what you are saying?
17:28:06gevaertsgoffa: in that case I'd recommend (a) checking the filesystem, and (b) using tab completion in your irc client
17:28:27goffalol... sorry about that
17:28:42kugelI think the best would be to dump the cache as is along with a crc of the font file to invalidate the cache if needed
17:29:01AlexPgoffa: Works fine here with r25459
17:29:12gevaertssoap: unifont is 2MB
17:29:21goffarockbox have a disc scan util built in by chance?
17:29:23AlexPgoffa: And mine is organised in the same way as yours
17:29:27 Quit evilnick (Ping timeout: 265 seconds)
17:29:31AlexPgoffa: No
17:29:39goffayeah... i've been kind of suspecting this thing for a while
17:29:49goffasucks.. because its a 120gb
17:29:59gevaertskugel: that's one possible solution, yes
17:30:10 Join evilnick [0] (
17:30:15soapgevaerts, and... what IF I made the font cache 4MB?
17:30:25 Quit S_a_i_n_t (Ping timeout: 265 seconds)
17:30:36gevaertssoap: that would work. Do you also want some battery life?
17:30:40AlexPgoffa: Try a discscan then again, and let us know
17:30:40 Join S_a_i_n_t_ [0] (S_a_i_n_t@
17:30:54goffanow to figure out how to do fat32 in linux
17:31:07soapgevaerts, I wasn't trying to debate practicality - just possibility. Besides, I have 64MB of RAM. ;)
17:32:07AlexPgoffa: fsck.vfat
17:32:09goffawell.. i'll let that run
17:32:12goffayeah.. found it
17:32:30AlexPwith -a or -r if you want it to actually fix anything
17:32:37goffaah.. thanks
17:34:14goffayeah.. this thing would crash at odd times ... usually when deleting a file (not consistantly)
17:34:19gevaertskugel: I'm not entirely sure if the font cache has pointers in it or not
17:34:25kugelgevaerts: a single binary blob (especially if we dma align the buffer) is clearly the fastest solution isn't it?
17:34:27goffathanks for the help btw
17:34:33gevaertsoh, definitely
17:37:10 Quit flydutch (Quit: /* empty */)
17:39:07goffaah.. yeah... finding shit now ... time to hit ebay it's looking like
17:39:38goffa120 still about the biggest size for these without needing some kind of adapter?
17:40:10AlexPThe S has a problem with the OF
17:40:25AlexPBigger discs will physically fit, but the OF will reject them
17:40:36AlexPAnd Rockbox would be fine with them otherwise
17:40:59gevaertsgoffa: filesystem corruption usually isn't because of hardware issues...
17:41:17AlexPI can't remember the details though
17:50:30goffagevaerts: what usually causes it then
17:51:07gevaertsbugs, random cosmic rays, crashes, disconnecting at the wrong moment,...
17:51:08goffaAlexP: i'll probably wind up getting the same drive then... just had to ask
17:51:49 Join panni_ [0] (
17:56:01 Join veeloc [0] (
17:56:44AlexPgoffa: I'd fix the corruption and use the same disk first - as gevaerts says it is rarely due to hardware failure
17:57:17 Nick Beta2K is now known as Beta2K_ (
17:58:09goffayeah... i figure i have to start looking ... just so i can have one on hand
17:58:47goffaalthough i fully intend to try to keep this thing going
18:03:01 Join geertvdijk [0] (
18:05:15 Quit moos (Quit: ChatZilla 0.9.86 [Firefox 3.6.2/20100316074819])
18:05:46 Join moos [0] (moos@rockbox/staff/moos)
18:09:06 Join RadicalR2 [0] (
18:09:10 Quit RadicalR (Ping timeout: 245 seconds)
18:10:01 Join RadicalR [0] (
18:11:44 Join anewuser [0] (anewuser@unaffiliated/anewuser)
18:13:55 Quit RadicalR2 (Ping timeout: 265 seconds)
18:14:07paulk_I'd like to update a FM Preset on How can I do that ?
18:17:27 Join jeffp [0] (
18:18:55 Part jeffp
18:19:14 Join jeffp [0] (
18:21:31 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
18:25:44AlexPpaulk_: It is a wiki, you can just edit it
18:25:58paulk_AlexP: yes, I've done the modification
18:26:17AlexPI don't understand the question then
18:26:46AlexPOh, you mean you have done it since
18:26:49AlexPOK then :)
18:37:19 Quit xiainx (Ping timeout: 246 seconds)
18:39:17 Join stoffel [0] (
18:39:41 Part jeffp
18:46:28 Quit goffa (Read error: Connection reset by peer)
19:02:24***Saving seen data "./dancer.seen"
19:04:03 Quit kugel (Remote host closed the connection)
19:06:14pamaurybluebrother: I've update the French translation and put a new version on Flyspray (same task). If you want to commit it, feel free to do so. I'm really unsure about the remaining strings so I'm not sure I will translate them
19:07:34bluebrotherpamaury: ok. IMO it's better to have the translation somewhat updated than not at all, and I can't do anything about it as I don't speak french. So I'll commit it later. Unless you do it yourself before :)
19:08:17pamauryI can't commit from here so please do it ;)
19:10:16bluebrotherbtw, did you use the webinterface for translating? If so, is there anything that's problematic when using it? Or is it ok?
19:12:45pamauryFirst I used the webinterface but then I saved the file and used QtLinguist because it's easier. The webinterface is ok but it can't beat a dedicated tool :)
19:13:44 Join Highlander_ [0] (
19:13:57bluebrotherthat's definitely true. So at least it meets its intentional use :)
19:23:20*pamaury is really puzzled, iso transfers are somewhat mysterious, especially when exactly half of packets disappear !
19:23:50bertrikcould that be some problem with the toggle bit?
19:25:10 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
19:26:03pamauryI don't know, when I set everything to the limit (3 packets of 1024 bytes per frame) everything is ok but if I send less that that to the device (typically 32 bytes per frame) then half of them are lost. It could have to do with the number of packets per frame but the device can't forsee the data size !
19:29:36 Join xiainx [0] (xiainx@wpa062009.Wireless.McGill.CA)
19:30:48 Join Strife89 [0] (
19:36:18 Join mikroflops_ [0] (
19:36:24 Quit bmbl (Quit: Bye!)
19:37:51moospamaury: I will look at it if there are still missing strings
19:38:35pamaurymoos: ok thanks, there are missing ones, mainly about TTS that I don't know how to translate
19:39:17moosohoh, ok, will have a go at it in a few
19:39:37 Quit mikroflops (Ping timeout: 260 seconds)
19:39:40 Join jnss [0] (janes@gateway/shell/
19:45:30 Join flydutch [0] (
19:48:41 Quit geertvdijk (Remote host closed the connection)
19:49:02 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
19:52:19*pamaury is even more puzzled by the solution he found to this problem...
19:57:09 Join goffa [0] (~goffa@
19:57:18 Join funman [0] (~fun@rockbox/developer/funman)
19:57:30pamauryLadies and Gentlemen, usb audio WORKS !!!!!!!!!!
19:58:10pamaury\o/ awesome :) Much work to be done but at least it works
20:05:39 Join CaptainKewl [0] (
20:09:03 Quit funman (Quit: free(random());)
20:12:35 Join TheSeven [0] (~theseven@rockbox/developer/TheSeven)
20:13:15 Quit mitk (Quit: Leaving)
20:14:48 Quit stoffel (Remote host closed the connection)
20:14:52 Quit n1s (Quit: Lmnar)
20:16:10 Join bint [0] (~bint@
20:17:28 Part jnss
20:17:51 Join JohannesSM64 [0] (
20:18:12bintim fuckin bored
20:18:20binttravelling with this ugly girl
20:18:33bintshe's smokin all day long and stinks
20:18:36bintwhat should i do=
20:18:41gevaertsbint: this is an on-topic channel
20:18:47bintoh shit
20:18:50bintwhere can i go chat?
20:20:03gevaertswherever you like. Not here
20:20:50bintisn't there any like chat channel
20:21:50gevaertsfinding chat channels is off-topic here
20:23:38 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
20:24:13 Quit robin0800_ (Remote host closed the connection)
20:26:45 Quit kramer3d (Ping timeout: 258 seconds)
20:27:28bintu suck
20:27:30 Part bint
20:35:09 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
20:40:13Zagorwhat's with the name of apps/plugins/pdbox/PDa/extra/zerox~.c ?
20:40:23 Join kugel [0] (~kugel@rockbox/developer/kugel)
20:41:12 Join merbanan [0] (
20:42:11gevaertsZagor: iirc the pdbox files with ~ deal with waveforms
20:43:45Zagorhow ... creative
20:46:42gevaertsyea, it is that
20:50:03 Join r2k000 [0] (
20:55:20 Quit piotrekm (Ping timeout: 276 seconds)
20:57:19 Quit adnyxo (Read error: Connection timed out)
20:58:00 Join adnyxo [0] (
21:01:33 Join piotrekm [0] (~piotrek@unaffiliated/piotrekm)
21:02:25***Saving seen data "./dancer.seen"
21:02:44 Nick kramer3d_ is now known as kramer3d (~kramer@unaffiliated/kramer3d)
21:04:06kugelcan I combine the "Milestones and Project plan" section with the "Subdivide the working period in several parts" section? I'd repeat myself a lot if not
21:04:14kugel(re: GSoC application)
21:05:56CIA-5New commit by zagor (r25470): Added source and font links to manual download page. Cleaned up the links a little.
21:06:11gevaertskugel: I don't think that's a problem
21:06:36 Join geertvdijk [0] (
21:16:20 Quit Strife89 (Read error: Connection reset by peer)
21:16:45 Join Strife89 [0] (
21:17:15 Join phanboy4 [0] (
21:20:00pamauryFlynDice: if you have time, see for the modification I propose about the sd code. I hope this little piece of code is clearer than the explaination I gave last time. I didn't check this code because I don't have a SHDC card.
21:24:18*pixelma points out that the wiki page she's referring to is actually called LatexGuidelinesTalk :\\
21:24:32pixelmain the commit message that is
21:29:31 Quit Strife89 (Quit: Easter get-together.)
21:32:30gevaertspamaury: it's a bit hard to distinguish the API changes from the isochronous additions in your tree. We'll need to update other drivers to deal with the changes...
21:34:16 Quit guymann (Ping timeout: 240 seconds)
21:34:56pamaurygevaerts: all changes in the api are in usb-drv.h, clearly described, only usb-drv-arc is ported of course. We could have a new #define HAVE_NEW_USB_API :)
21:35:25 Join guymann [0] (
21:36:07jhMikeSsoap: Clarification from MCIMX31.pdf : 2 Supply voltage is considered overdrive for voltages above 1.47 V. Operation time in overdrivewhether switching or
21:36:07jhMikeSnotmust be limited to a cumulative duration of 1.25 years (10,950 hours) or less to sustain the maximum operating voltage
21:36:07jhMikeSwithout significant device degradationfor example, 25% (average 6 hours out of 24 yours per day) duty cycle for 5-year rated
21:36:07DBUGEnqueued KICK jhMikeS
21:36:07jhMikeSequipment. To tolerate the maximum operating overdrive voltage for 10 years, the device must have a duty cycle of 12.5% or
21:36:07jhMikeSless in overdrive (for example 3 out of 24 hours per day). Below 1.47V, duty cycle restrictions may apply for equipment rated
21:36:08***Alert Mode level 1
21:36:08jhMikeSabove 5 years.
21:36:56 Quit anewuser ()
21:38:24pamaurygevaerts: I don't know how much it represents to port all drivers but it's probably not negligeable, that's why I propose to have a switch for this new api
21:38:25 Quit linuxstb (Ping timeout: 258 seconds)
21:38:29pamaury*much work
21:38:32 Nick YPSY is now known as Ypsy (
21:39:06gevaertspamaury: well, it depends on how clean you can make the set of ifdefs needed
21:39:34gevaertsthe _nonblocking() change is fairly trivial I guess, so we could do that
21:40:04pamauryMy current code emulates the old api with the new one, it's only a matter of adding code within #ifdef (the usb_drv_{select_mode,allocate_slots}, whose name are probably not very well chosen)
21:40:36gevaertsYou seem to have a spurios change in logf.c
21:41:15pamauryOh, don't worry the changes in my repo are not up to date (I can't push from where I am) and I even changed usb-drv-arc today so the code is not clean
21:41:37 Join kugel_ [0] (
21:41:53 Quit kugel (Disconnected by services)
21:41:57 Nick kugel_ is now known as kugel (
21:42:01 Quit kugel (Changing host)
21:42:01 Join kugel [0] (~kugel@rockbox/developer/kugel)
21:42:11gevaertsApart from the other driver issues, I think the changes are fine
21:43:47pamauryOk I need to go, you can continue to comment the changes, I'll read them later.
21:44:07gevaertsI was actually planning to look for a bug elsewhere :)
21:44:30gevaertsjhMikeS: do you have a theme installed that includes an sbs? I think I suspect those
21:46:09***Alert Mode OFF
21:47:40 Join MaadMan [0] (
21:47:54CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
21:47:54*bluebrother grumbles at
21:48:41gevaertsbluebrother: that means you should ask them to fix broken usb enumeration in wine
21:49:22bluebrothergevaerts: I'm rather confident that after fixing that other issues will come up. Especially once we manage to add MTP support.
21:49:54bluebrotherso no, I _really_ don't want to run the windows binary in wine, and I _really_ want to be able informing the user that there's a linux binary of the program around.
21:50:21gevaertsbluebrother: I know, but if you ask them about these issues a few times, I suspect they might get the message :)
21:51:31gevaertsAh, they live in #winehq
21:51:59 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
21:55:19gevaertsjhMikeS: some more testing seems to point to the freezing occuring when switching *away* from
21:56:03gevaertshm, interesting. It's not frozen now, but since a few minites the "Loading
21:56:21gevaerts" splash is there, and I get a few seconds of music alternated with a few more seconds of silence
21:56:33*bluebrother wonders if it's a valid way to trust on portability differences between windows and linux:
21:56:46jhMikeSgevearts: I only have the RB "OEM" themes on that one
21:57:17gevaertsjhMikeS: if I skip that one theme, I don't see issues either
21:58:16gevaertsOh, it finished switching!
21:58:46jhMikeSgevaerts: I don't have that one there. Just: cabbiev2, DockPod Aqua, iCatcher, rockbox_default, Rockboxed, Unicatcher
21:59:45*gevaerts decides to try the sim
21:59:48jhMikeSiCatcher can hang a little though when loading
22:04:43jhMikeSsomething it takes a few seconds, but I can't get duppel, away to to it, to "hang" though I might hear music restart twice when going to it
22:09:15kugelbluebrother: couldnt you run wine from within rbutil and assume windows if it's not found?
22:10:02bluebrotherwine from within rbutil?
22:10:28bluebrotherah, by checking for the binary? Well, that won't work if wine isn't in PATH
22:11:42kugelthat's very unlikely I'd say (you couldn't just double-click a .exe then IIUC?). does it need to work 100% of the time?
22:12:20gevaertsI can't reproduce this in the sim
22:12:37kugelor detect if a wineserver process is running?
22:12:57*gevaerts thinks that kugel might know something about sbs loading
22:13:49kugelhm, I think that all won't work, since it would give false-positives if rbutil itself isn't run under wine but other apps
22:14:05 Join jd [0] (jd@
22:14:05 Quit jd (Changing host)
22:14:05 Join jd [0] (jd@Wikipedia/HellDragon)
22:14:29kugelgevaerts: uhm yes, but I have no idea why particular themes would load slower
22:14:46gevaertskugel: well, "slower" isn't the issue. Sometimes it just freezes
22:15:17kugelno idea about that either
22:17:13kugelhave you checked the disk/filesystem?
22:17:46gevaertsyes, several times during the past days :)
22:19:05gevaertsit's probably time for lots of splashes
22:21:54gevaertskugel: any idea where I should start looking?
22:22:17kugelhave you figured out if it's the wps or the sbs?
22:22:59gevaertswell, it seems to happen when I move away from the one theme that has an sbs
22:23:01*pixelma wonders why she gets a rockbox.iaudio and no rockboxui.exe when trying to build a win sim under linux using the "w" option in an advanced build
22:23:03 Join Bagder [0] (~daniel@rockbox/developer/bagder)
22:23:04kugelanyway, skin_data_load() is the function that handles the loading
22:24:08 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
22:27:35 Quit kramer3d (Ping timeout: 260 seconds)
22:29:58 Quit Zagor (Quit: Leaving)
22:31:31linuxstbbluebrother: Is your worry about wine just from or have there been others?
22:32:09 Nick Ypsy is now known as YPSY (
22:33:53gevaertsyay, now the screen is just corrupted...
22:34:13 Quit timaeus (Quit: Page closed)
22:35:16bluebrotherlinuxstb: it's the post that reminded me of the issue. I've seen people try to run it on wine in the past a few times
22:35:44gevaertsfun, now the screen refuses to show any "d"
22:35:58bluebrotherI remember a post somewhere (don't think it was our forum) where someone complained about autodetection and installation not working, and he tried running it with wine :/
22:36:20*gevaerts sees files like "test_coec_log_01.txt"
22:37:41kugelgevaerts: I'm not sure the skin engine is the cause here
22:37:59gevaertskugel: it goes wrong after the settings (?) code prints "Loading..." and before it enters skin_data_load()
22:38:08kugelbluebrother: doesn't Qt offer runtime detection of the os?
22:38:34*jhMikeS just wonders what's different on gevearts' F than his own here.
22:38:38gevaertskugel: I'm not sure either. All I know is that there seem to be some stray pointers every now and then
22:39:08bluebrotherkugel: that doesn't work at all. Runtime detection of the OS should return Windows when running on wine, obviously.
22:39:27bluebrotherand I guess you're referring to the OS macros Qt offers. But that's macros, so no runtime detection
22:39:44gevaertsjhMikeS: my config right now is
22:39:44bluebrotherhowever, the solution the guys in #winehq suggested works fine so I'm going for that.
22:39:55kugelno, my idea was runtime detection but implement different behaviors with the macros
22:40:22kugelit wasn't so obvious that it should returns windows to me
22:40:59kugelbluebrother: what solution?
22:41:11bluebrotherwell, IMO it's pretty obvious: wine is trying to emulate Windows, so a syscall retrieving the windows version should return a valid windows version
22:41:27bluebrotherkugel: the solution I haven't mentioned here yet ;-) Check for a registry key
22:42:16kugelsure you haven't mentioned it, that's why I'm asking ;)
22:48:08 Join cjcopi [0] (
22:48:51gevaertsjhMikeS: could you try with ? That's a new version that actually loads the correct font (the old version still used the old font names)
22:49:14jhMikeSgevaerts: duplicated! cleared my settings to stock, loaded your config + druppel, rebooted, started playback, crash when switching back to cabbiev2
22:50:46gevaertsSince I added splashes, the most common symptom is a corrupted font
22:50:49jhMikeSgevaerts: so now you want me to try that link to see if it stops? I duped your results before that.
22:51:43gevaertsjhMikeS: not sure. Does it use nimbus 14? You might also still have the old fonts installed in which case it won't make any difference
22:52:27 Quit kramer3d_ (Quit: Leaving)
22:52:34jhMikeSgevaerts: it seemed to be stuck with the stock system font
22:52:53gevaertsOK, that's useful to know
22:53:15gevaertsI was beginning to suspect that maybe it was something in the font loading, but that seems unlikely then
22:54:07*gevaerts thinks that this is a weird bug
22:54:21jhMikeSI have 19-Nimbus.fnt only. I don't think I have that installed
22:54:32jhMikeS*Nimbus 14
22:55:02gevaertsmaybe it's from the font pack. Anyway, if you can reproduce it, the font is not important
22:55:55gevaertsMaybe the antiskip setting makes it more likely to hit? I don't really see any other setting that should make a difference
22:56:16jhMikeSyou had the correct font I take it? I don't have it at all, so you're likely correct.
22:56:30gevaertsyes, I do have it
22:57:12kugeljhMikeS: you need the font package
22:57:13jhMikeSok, just sanity checking. settings seem relevant. I suppose I'll pastebin mine.
23:00:57jhMikeSgevaerts: what i just saved before changing to your config:
23:01:30 Quit paulk_ (Read error: Connection reset by peer)
23:02:29***Saving seen data "./dancer.seen"
23:03:23*gevaerts wonders why he has gather runtime data set to on
23:04:06 Quit xiainx (Ping timeout: 240 seconds)
23:05:02gevaertsjhMikeS: can you see if "gather runtime data" makes a difference? That seems to be the one that does it here
23:07:41jhMikeSwhere's that one located again? lol (i don't use it)
23:08:42CIA-5New commit by bluebrother (r25471): Try to detect Wine. ...
23:08:44gevaertshm, I have it disabled now, and it just froze
23:10:48jhMikeSI just switch from druppel to cabbiev2 and it did it again. First the fonts were trashed (metadata display trashed) then it poped, locked and the screen turned white.
23:12:22gevaertsI just had it send noise to my ears
23:13:01 Nick Highlander_ is now known as Highlander (
23:17:25pixelmaprobably too much smurfing...
23:17:37pixelmasorry, couldn't resist
23:17:38jhMikeSdruppel->cabbiev2 switch is pretty reliable
23:18:07jhMikeSsmurfs!! those wretched, vile, blue creatures!
23:19:06gevaertsOK, I've now reproduced it with default settings, i.e. an empty config.cfg
23:22:18 Quit liar (Ping timeout: 258 seconds)
23:22:51 Quit MaadMan (Read error: Connection reset by peer)
23:25:30FlynDicekugel: Do we have the same problem with the buttonlight for the fuzev2 with GPIOD being used for both uSD interface and general I/O?
23:25:59gevaertsSo what makes "druppel" special?
23:26:26kugelFlynDice: not the same problem, but the buttonlight pin is the same as the internal-external-storage select pin (gpiob5)
23:26:27jhMikeSgevaerts: the sbs and floating listview with metadata always visible I suppose
23:26:43kugelwhich is probably why the buttonlight blinks when playing something off the microsd
23:27:23kugelgevaerts, jhMikeS: still not reproducible in the sim?
23:27:35goffahmm ... this happened to me before don't remember how i fixed it ... i have a corrupt dir in the file system... its basically a mirror of root
23:27:46goffaif i attempt to delete that dir it'll delete everything out of root
23:28:30kugelFlynDice: that was a yes and no :) so far there's nothing special on gpiod except the power and home button (but that seems to work so far)
23:28:30jhMikeSperhaps when switching away, metadata is being accessed that is no longer present?
23:28:46FlynDicekugel: So There shouldn't be a problem if we just leave XPD configured as the uSD interface. Do you think that's accurate? If that wont cause a problem I've got the clip+ uSD functioning again.
23:28:59gevaertskugel: I haven't managed anyway
23:30:00kugelFlynDice: just go for it, I'll report failure :p
23:30:14gevaertsjhMikeS: could be. It does a rebuffer (due to different AA size)
23:30:23FlynDiceI'm sure you will ;-)
23:30:38gevaertsAlthough when I had the splashes, it looked like the crash happened before it even started on that
23:30:48FlynDiceI'll give it a try here shortly.
23:31:02kugelthe rebuffer happens at the very end of skin loading (but once per skin)
23:34:31jhMikeSis druppel supposed to switch to 14-nimbus? it seems to not load it but I can load that from Theme Settings->Font ok
23:35:21gevaertsjhMikeS: the "old" (i.e. until an hour ago) version referred to nimbus-14
23:35:52jhMikeSaha, ok. now I just get the system font
23:36:58gevaertsAh, news!
23:37:15gevaertsI managed to reproduce it by loading 15-Adobe-Helvetica while in druppel
23:39:36jhMikeShmmm...I can't seem to get results that way. did it take awhile?
23:39:53 Join liar [0] (
23:40:09gevaertsfirst time
23:40:43jhMikeSis there any protection against text output when overwriting the old font?
23:43:17 Quit Highlander (Quit: Quitte)
23:45:27gevaertshm, I can't manage to reproduce it just by switching fonts anymore, but I also don't seem to have it if I first switch to the right font and then switch themes
23:46:52*gevaerts *always* speaks too soon
23:48:33 Join xiainx [0] (
23:48:37jhMikeSno worry, I do that *alot*
23:49:17gevaertswhat can we try next?
23:51:36jhMikeSI don't know. Any nice way to make sure nothings writing into naughty places?
23:52:21jhMikeScheck all theme-related data makes sense when used, somehow

Previous day | Next day