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

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

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

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

#rockbox log for 2010-05-29

00:01:56***Saving seen data "./dancer.seen"
00:02:58 Join esperegu [0] (~quassel@
00:08:41 Quit bertrik (Ping timeout: 260 seconds)
00:09:39kugelfunman: can't you do .code 32 in inline assembly?
00:10:33funmanyes but if you try that you will see thumb code before & after the asm()
00:10:47funman-mthumb specifies if the full file is built in 16 bits, you can't select functions
00:11:16funmanbut i think not a lot of files use asm() (if i print which files aren't built with thumb)
00:11:24kugelI thought gcc doesn't care about inline assembly
00:12:09kugelit just passes the inline asm to as (possibly with allocated registers) so it shouldn't break -mthumb
00:12:35funmanall the surrounding C code is translated to thumb asm so it can't work together
00:13:02kugeldoesn't .code 32 switch to arm mode on the fly?
00:13:05funmangcc does either thumb either arm for each time it's invoked, never both together
00:13:09kugelno matter of the preceding code?
00:13:18funmantry .code 32 bla .code 16 bla and assemble it
00:13:56kugelwhat happens?
00:14:15funman1 ARM instruction, 1 thumb instruction; no switch
00:14:30kugelthat sounds stupid
00:14:42funman-mthumb-interwork add 'veneers' to be able to make *calls* between ARM & thumb functions
00:14:46kugelwhat are these directives for then
00:14:57funmanwell when you write asm
00:16:25funmanthen, you write the switches yourself
00:16:46kugelthat doesn't answer my question
00:17:05funmanif you want to switch between thumb & arm in gcc, put the functions in separate files and call them but they're not inline anymore
00:17:34kugelwhat are .code X for, what do they do?
00:18:17funmanthey output 16 or 32 bits code, "mov r0, r0" worsk in both but doesnt' give the same output
00:18:56kugelah ok, but they don't generate a switch?
00:19:28kugelthere's not much inline assembly (in the core), could we insert the switch for them?
00:19:39 Join MethoS- [0] (~clemens@
00:19:41funmanswitch = bx/blx
00:19:55funmanif you want to switch, make them real functions (not inline) and put them in their own file
00:20:08 Quit TheSeven (Ping timeout: 240 seconds)
00:20:13funmananyway first step would be to figure why it doesn't boot / crashes
00:20:13kugelsub pc, pc, #7 also, no?
00:20:34kugelI'm fairly sure bx/blx are not the only switches
00:22:52funman'DDI0100E' says BX/BLX, or setting the T bit in spsr and restoring it
00:23:05funmanor ldr/ldm pc
00:23:30 Quit funman (Quit: free(random());)
00:23:50kugel might be interesting
00:25:52amiconnIirc ldr/ldm pc won't work on armv4t though, only on armv5+
00:33:42 Quit kugel (Ping timeout: 265 seconds)
00:34:59 Quit Barahir_ (Ping timeout: 265 seconds)
00:35:54 Join kugel [0] (~kugel@rockbox/developer/kugel)
00:36:31 Join Barahir [0] (
00:37:53 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
00:47:45kugelAlexP: ping
00:49:06kugelAlexP: this is the list of PLA plugins and their actions:
00:51:12XStatikHow do i make a playlist in rockbox
01:04:04*Unhelpful finds it difficult to believe that arm/thumb code absolutely must reside in different source files?
01:04:48 Quit wincent (Ping timeout: 240 seconds)
01:07:42kugelUnhelpful: makes sense to me as far as gcc is concerned
01:07:49kugelbut it should be possible to mix with inline asm
01:08:37Unhelpfuldoes it make sense? i thought you could add attributes to tell gcc to compile a function as arm or thumb? and surely a pure-asm function in thumb mode in a C file compiled for arm mode is also valid?
01:09:17kugela pure asm function in a c file?
01:09:25 Join merbanan [0] (
01:10:55kugelUnhelpful: mixing is going to be problematic with function pointers
01:12:48Unhelpfulkugel: that depends on which arm revision is involved, i'd think... armv5 and up should do the right thing on branching to a pointer that goes to thumb code, pretty much no matter how the branch is done.
01:13:19kugelthe function pointer will not have the lowest bit set
01:13:26 Part toffe82
01:13:28kugelwhich you need to switch to thumb
01:15:43 Quit leavittx (Ping timeout: 240 seconds)
01:16:08 Join leavittx [0] (~leavittx@
01:16:33 Quit GeekShadow (Read error: Connection reset by peer)
01:17:37 Quit MethoS- (Remote host closed the connection)
01:19:38 Join MethoS- [0] (~clemens@
01:21:05Unhelpfulkugel: surely if the function is compiled as thumb code pointers to it will have the lowest bit set, in the same way that branches to it would?
01:21:41Unhelpfuloddly i *don't* see any attributes for making individual functions arm or thumb... i wonder why i thought there were?
01:21:44kugelyou'd get in trouble with the C standard here I think
01:22:07kugelthe lowest bit is just a trick for ARM, the actual function start is at a different address
01:22:47 Quit epicfailguy ()
01:23:26kugelyou can do a bit more with function pointers than just calling it, I'd expect problems if the lowest bit is just set unconditionally
01:23:43Unhelpfulproblems like what?
01:24:35Unhelpfuland i'm trying to imagine what you can do with a function pointer besides store/retrieve/call... none of which will care if the low bit is set. you could always *try* writing code that does this. ;)
01:24:40kugelalso, this is more of a linker thing. gcc doesn't know the addresses anyway
01:25:32Unhelpfulyes, the compiler *doesn't* know the addresses, but it puts in placeholders with the symbol name, which should be replaced with actual values or offsets at link time
01:27:18 Join anewuser [0] (anewuser@unaffiliated/anewuser)
01:29:20 Quit DataGhost (Ping timeout: 240 seconds)
01:41:15Kevoneso does anyone think that it will be possible to put rockbox in the future?
01:41:20CIA-5New commit by 03kugel (r26366): Upgrade the ARM eabi toolchain to gcc 4.4.4
01:41:21 Quit MethoS- (Remote host closed the connection)
01:41:32Kevoneon a 120gb classic
01:42:08 Quit DerPapst (Quit: Leaving.)
01:44:52kugelalmost to late now but did we ever consider gcc 4.0.4 for the old arm toolchain?
01:45:27 Join halmi_ [0] (
01:48:04 Quit halmi (Ping timeout: 258 seconds)
01:52:16 Join saratoga [0] (~9803c6dd@gateway/web/freenode/x-niwiiveldqaypsrx)
01:58:17 Quit kugel (Remote host closed the connection)
02:01:59***Saving seen data "./dancer.seen"
02:04:06CIA-5New commit by 03bieber (r26367): Theme Editor: Got code generation tentatively working along with a solid C++ tree structure for WPS parse trees
02:21:19 Quit efyx (Remote host closed the connection)
02:30:13bieberbluebrother: I've been playing around a little bit with the model, and I did manage to get some rudimentary tree display that would expand for a while and then crash when I went off the end of the tree
02:31:20bieberToday I just focused on getting a solid, working tree built out of ParseTreeNodes, and I've got them doing code generation. Now that I've got the data structure solid and read up some more on the documentation, I think I should be able to get the interface implemented without too much trouble
02:33:13 Quit Curtman (Remote host closed the connection)
02:37:39 Quit tmzt (Read error: Operation timed out)
02:41:05 Join tmzt [0] (
02:41:46 Quit domonoky1 (Read error: Connection reset by peer)
02:49:14 Quit halmi_ (Read error: Connection reset by peer)
03:09:11 Join keanu [0] (~keanu@unaffiliated/keanu)
03:21:37 Quit merbanan (Ping timeout: 258 seconds)
03:34:06 Join CaptainKwel [0] (
03:51:56 Join moparx [0] (~moparx@unaffiliated/moparx)
03:55:13 Join jhMikeS [0] (
03:55:14 Quit jhMikeS (Changing host)
03:55:14 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
03:56:29Unhelpfulkugel: i'm building 4.4.4 now, also trying the binutils patch from to see if it gets rid of the need for the lds workaround we use for plugins/codecs
04:00:28 Quit xavieran (Ping timeout: 260 seconds)
04:01:00*jhMikeS finally got arm-elf-eabi-gcc compiled after all sorts of bs and find it does just exactly what it should with the stubbing (checked disassemlies too)
04:02:00***Saving seen data "./dancer.seen"
04:02:59 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
04:04:32 Join xavieran [0] (
04:28:51 Quit komputes_ubuntu (Ping timeout: 245 seconds)
04:37:00 Join Barahir_ [0] (
04:39:31 Quit TheSeven (Ping timeout: 265 seconds)
04:40:21 Quit Barahir (Ping timeout: 252 seconds)
04:41:23 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:41:23 Quit pixelma (Disconnected by services)
04:41:25 Quit amiconn (Disconnected by services)
04:41:26 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:41:39 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:41:47 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:42:19 Quit moparx (Ping timeout: 248 seconds)
04:43:28 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
04:55:29UnhelpfuljhMikeS: "what it should" == only as many stubs as needed?
04:57:21*Unhelpful is glad somebody on binutils got back to him about it, but their patch doesn't fix the bug that r23674 works around
05:03:01 Join CGL [0] (~CGL@
05:04:49 Quit CaptainKwel (Quit: Ex-Chat)
05:21:17 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
05:33:20 Quit ps-auxw (Ping timeout: 240 seconds)
05:33:53jhMikeSUnhelpful: yes, just a few of them and 8 bytes each - ldr pc,=function_address
05:35:03jhMikeSwe could work this out with -mabi=aapcs-linux which keeps enums 32 bits iiuc
05:35:22 Join ps-auxw [0] (~arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
05:36:12Unhelpfulwe could. there is also -fshort-enums. in either case we need to make sure libgcc etc is built w/ the same flags, as things *will* error on mixing long/short enums.
05:37:30Unhelpfuli'd rather fix code that makes bad assumptions, though, especially if the alternative is making our compiler even more different from mainline gcc. at present we don't really change much except which multilib targets are built and disabling exceptions in libgcc
05:40:35jhMikeSyeah, I'm building now and it's spraying warnings about it everywhere
05:43:06jhMikeSfinding the assumptions is the problem, not easy to track in all that. I hope stuff like int *p = addr; *p = enum_value; isn't a problem (that would be crazy)
05:45:20jhMikeSworse yet removing -fno-strict-aliasing doesn't help :\
05:56:25Unhelpfulyes, assigning an enum value into an int, or int*, is absolutely valid.
05:56:47Unhelpfulpointing to an enum with an int* is not, and comparing random things to an enum may break.
06:02:04***Saving seen data "./dancer.seen"
06:03:48jhMikeSrandom things...?
06:04:44UnhelpfuljhMikeS: for example, using -1 as a flag value for an enum that only enumerates small positive values.
06:05:51Unhelpfultesting for equality is probably ok, but if you do if (enum_var > -1) the result will not be right, and casting either side of it is unlikely to help.
06:06:26 Join Boldfilter [0] (
06:06:28 Quit Boldfilter (Remote host closed the connection)
06:06:34jhMikeSwouldn't that obey exention conversion rules?
06:06:59Unhelpfulcasting -1 to your enum type gets you 255 most likely, and the comparison is always false. casting enum_var to int is no better, if -1 was stored into it it has a value of 255 and casting it to int means the comparison is always true.
06:07:35Unhelpfulwhat extension conversion rules? enum signedness depends on the enumerated values :/
06:09:05jhMikeStrue, and extended an unsigned value to a larger signed type is a zero-extension, unless you go like (int)(char)c or something
06:11:02Unhelpfulbut that involves making an assumption about the size of the enum. :)
06:14:31jhMikeSI know, just saying. :) kugel said it's mostly the settings array. is there a patch up that gets that much done at least?
06:17:48Unhelpfuldon't know, i prodded a few of the obvious wrong bits in doom, but i never got around to trying to find *all* of them. the quick fix would be to make the "bad" variables ints, or to have their enum types enumerate *all* values we might store in them. i'm not sure which is going to be better in terms of compiled code, because the latter will also force us to write switch statements that test for values that we already know shouldn't reach
06:17:48Unhelpful them.
06:24:04 Quit shai (Ping timeout: 265 seconds)
06:30:32jhMikeSwell, I certain have no intention to rewrite all of idsoftware's code. this doom shouldn't really actually live up to that name here. :)
06:32:32jhMikeSI'm thinking to take the typedefs, and re-typedef as int
06:33:02jhMikeSthere's some enum xxxx {}; enum xxxx var; stuff but not much
06:34:43 Join shai [0] (
06:49:18 Quit panni_ (Read error: Connection reset by peer)
07:05:58 Quit Horscht (Ping timeout: 260 seconds)
07:06:56 Quit CGL (Ping timeout: 265 seconds)
07:11:58 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
07:16:09 Quit pyther (Ping timeout: 276 seconds)
07:18:41 Join CGL [0] (~CGL@
07:19:10 Nick CGL is now known as [CGL] (~CGL@
07:42:02 Quit Kevone ()
07:52:54 Join BHSPitMini [0] (
08:02:05***Saving seen data "./dancer.seen"
08:05:16jhMikeS:0 :)
08:06:09 Join S_a_i_n_t_ [0] (S_a_i_n_t@
08:07:10 Quit S_a_i_n_t (Ping timeout: 276 seconds)
08:09:35Unhelpfulwhat, it works?
08:11:07 Quit shai (Quit: Leaving)
08:19:25 Join shai [0] (
08:20:23Unhelpfulwhat, *all*?
08:21:52*jhMikeS wasn't going mess around twiddling this and that
08:22:51Unhelpfulalso, should it maybe be int *always* instead of int-or-unsigned? i'm thinking of things that use -1 as a flag value, and cases where >/< are used to compare w/ -1 instead of simple equality tests... although i think i may have fixed those?
08:23:51jhMikeSgcc seemed to compile it exactly that way, otherwise I got signed/unsigned comparison warnings
08:25:22Unhelpfulprobably somebody should test *more* plugins for correct behavior? i only really bothered with ones that threw warnings on compile.
08:25:52 Quit esperegu (Read error: Connection reset by peer)
08:26:48Unhelpfuli mean, before we give serious consideration to switching to gcc-4.4/eabi for arm targets ;)
08:26:51 Join esperegu [0] (~quassel@
08:27:34Unhelpfulalso there were some codec performance regressions, but *most* of these went away if -O levels were tweaked - the bulk of codecs get -O1 iirc.
08:31:37jhMikeSI didn't get any compile warnings. Does the which get their optimization change under EABI?
08:32:38jhMikeS*the ONES that get their own optimizations
08:34:45 Join stoffel [0] (
08:44:28Unhelpfulnot that so much as some of them performed *really* badly if left w/ the core's typical optimization
08:47:45jhMikeSHmm...SPC might be faster
08:50:35 Nick S_a_i_n_t_ is now known as S_a_i_n_t (S_a_i_n_t@
08:53:37*Unhelpful ran all of them on new and old w/ our standard and the new eabi toolchain, on -O1-3, and made a chart, but that was a while ago
08:57:55 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
09:14:31 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
09:15:23*jhMikeS doesn't remember getting a warning before preferring 4.4.4 for the beast, hm
09:17:23jhMikeSah, there be the commit that came when I wasn't looking r26366
09:18:23jhMikeSafter all I went through the first time, kugel goes and makes me do it all over again. gah
09:24:00 Quit rvvs89 (Ping timeout: 240 seconds)
09:27:24 Quit BHSPitMini (Ping timeout: 260 seconds)
09:29:01 Join kugel [0] (
09:29:02 Quit kugel (Changing host)
09:29:02 Join kugel [0] (~kugel@rockbox/developer/kugel)
09:29:14kugelUnhelpful: all codecs were pretty similar
09:29:44kugelonly flac dropped dramatically, but that's solved by -O2 (which makes it much faster for both toolchains)
09:29:47Unhelpfulkugel: between old/new compilers?
09:30:02*Unhelpful would have to check his "notes"
09:31:04kugel too
09:33:30kugeljhMikeS: sorry, but running isn't too challenging, is it?
09:34:38Unhelpfulwow, codesourcery not so amazing as claimed, eh?
09:35:46 Join ucchan [0] (
09:37:19Unhelpfulucchan: you've been working on embedded album art, yes?
09:39:59 Join flydutch [0] (
09:41:42jhMikeSkugel: not that itself. it's was the missing prerequisited and evironment problems on the little vm.
09:41:51 Join DerPapst [0] (
09:41:53kugeloh I see
09:41:56 Quit Tuplis (Quit: säätö jatkukoon)
09:42:55kugelso commit FS #11328 and switch to eabi asap! :)
09:44:23ucchanUnhelpful: yes. my iPod works well
09:45:33ucchanIs there a problem?
09:46:56 Join Rob2223 [0] (
09:47:12Unhelpfulucchan: nice. i was thinking that instead of using the codec buffer, it might make sense to be able to call the loader with an alternate read function, which would read the data from the file and transform it if needed.
09:47:54kugeldo we need to to read from file?
09:48:10Unhelpfulthis would minimize the changes in the loader, and also get around any need to use the codec buffer
09:48:12kugelcan't the AA be loaded into mem in the course of buffering the id3 data?
09:48:36jhMikeSeabi during feature freeze? :)
09:48:52kugelfeature freeze is over
09:48:58Unhelpfulkugel: do we actually buffer the id3 data as-is from the file? i thought we just stored the tags as strings into a pretty struct?
09:49:36kugelI think nothing should stop us with that doom fix. of course the release binaries will still be build with the old toolchains
09:49:39ucchanwhen mp3's aa is raw image data. but other tags (e.g., apes, ogg) are encoded by MIME64.
09:50:03jhMikeSboo...that came quick (haven't seen the topic because I was busy with arm-elf-eabi-gcc :)
09:50:03kugelUnhelpful: I don't know really, I was just wondering
09:50:17Unhelpfulif we actually buffer raw id3 data already, then yes, that would work too, the read-aa-id3 function would just transform data from memory instead of reading a chunk from a file and transforming it.
09:50:44 Quit Rob2222 (Ping timeout: 272 seconds)
09:51:03Unhelpfulthe jpeg loader is *very* kind in its use of getc/putc, it never rewinds more than one byte so you never have to keep any "old" data when refilling the read buffer
09:51:53Unhelpfulucchan: yeah, i know about the horrid schemes for AA in ogg. surely the same plan would work, though, reading a number of bytes in and transforming them?
09:51:57ucchanIs it really safe?
09:52:12kugelnaming random functions exactly like standard ones is bad :)
09:52:43S_a_i_n_tHmmm, if the switch the eabi is definite...and since is barfing on CygWin, some kind soul needs to put up an eabi package for CygWin on
09:52:59kugelwhat's wrong with it?
09:53:06UnhelpfulS_a_i_n_t: or you could fix your cygwin?
09:53:07jhMikeSkugel: so it's not too obscene a fix? it should preserve the way it has been compiled previously and what it expects
09:53:40kugeljhMikeS: I think that's OK and it also makes sense to do it that way
09:53:54S_a_i_n_tUnhelpful: I have ni idea what's wrong, I've installed/re-installed 4 times now...something's changed.
09:53:54kugeland I really don't care about doom, I just don't want it to block us anymore :)
09:55:24ucchanIf it is possible to convert it into bitmap without using buffer from the embed aa though I think the best.
09:55:41S_a_i_n_tthere's nothing wrong with CygWin, and the packages on work...but barfs...always in the same place.
09:55:41 Quit phanboy4 (Ping timeout: 240 seconds)
09:56:55Unhelpfulucchan: the bitmap loader has a pretty generous read buffer, too... and the same sort of transformation could be applied to it, of passing in a read function which treats its "file" argument as something other than really a file handle.
09:56:58 Quit yawny (Read error: Connection reset by peer)
09:57:12 Join elcan [0] (
09:57:31Unhelpfulmy idea here is that the read function would reserve the end of the read function for the real file handle, and for any state it needs
09:58:03 Join lpereira [0] (
09:58:10Unhelpfulend of the read *buffer* rather
09:58:20jhMikeSkugel: well, if it's a "works for me too" thing, I'll just shove it down the SVN tubes
09:59:38Unhelpfulalso should expand the read buffer for jpeg to >256B. even with read being a normal read this would be a win, a big chunk of the time in jpeg_getc is likely only when it needs to refill the buffer
10:01:08JdGordonLlorean: (re cloning the theme site) because the current thinking is to auto update every zip which would mean nothing will work on 3.6
10:02:09***Saving seen data "./dancer.seen"
10:02:41ucchan I will be displayed well using your idea.
10:02:55Unhelpfulwould base64 decode need anything special in the way of state? i imagine it can be done without maintaining any state provided that reads are done in multiples of 4 input bytes?
10:03:30jhMikeSok then...down it goes
10:06:04LloreanJdGordon: Shouldn't we let the authors decide whether they want to work on 3.6 or update?
10:06:38JdGordonI doubt any authors would be annoyed at the auto update
10:07:11UnhelpfulS_a_i_n_t: try checking out an older and see if you can identify the change that broke things?
10:07:26ucchanmaybe multiple 4 bytes.
10:10:23 Join rvvs89 [0] (
10:11:31kugelS_a_i_n_t: define barfs, what's going wrong exactly?
10:12:42CIA-5New commit by 03jethead71 (r26368): Doom plugin: In anticipation of moving to the EABI toolchain, because of its assumptions about enums being sized as ints, convert enum typdefs to int ...
10:14:02 Join Tuplis [0] (
10:16:20 Quit esperegu (Ping timeout: 258 seconds)
10:16:43 Join esperegu [0] (~quassel@
10:18:53 Join Navi [0] (~Lancelot@unaffiliated/navi)
10:22:39kugeljhMikeS: thanks for that!
10:28:29ucchanI try create the new jpeg decorder using the idea of Unhelpful.
10:28:30CIA-5New commit by 03jethead71 (r26369): Doom plugin: Missed properly making a few 'unsigned' on my second pass.
10:29:13Unhelpfulucchan: i'll try to keep an eye on the FS task for it... i understand it'll be easier to reach you there or via the mail list?
10:29:44*Unhelpful wonders if a read function for data-in-id3 could also be stateless
10:30:27Unhelpfulprobably not? it would at least need to store whether the data is sync-safe or not, although if it's not sync-safe we *could* just open the file, seek to the start of image data, and then use normal read(), right?
10:31:48JdGordongevaerts: did you get an answer to your q about settings?
10:32:45 Quit kugel (Remote host closed the connection)
10:35:41ucchanread only sure. the same time reading file is one.
10:36:10 Join ender` [0] (
10:37:57ucchanIf there is a problem, only the display of AA must be not correct, and it not influence original audio file and other Rockbox functions.
10:43:15*jhMikeS suddenly wonders about 64-bit sim builds
10:48:44 Quit bieber (Ping timeout: 248 seconds)
10:48:56 Join bieber [0] (
10:50:52ucchanS_a_i_n_t: for my cygwin's env, binutils build failure.
10:51:56ucchan../../binutils-2.20.1/libiberty/cp-demangle.c:116:21: error: alloca.h: No such file or directory.
10:52:39ucchanIs it the same as your error?
10:53:15 Quit esperegu (Ping timeout: 276 seconds)
10:56:31 Join esperegu [0] (~quassel@
10:59:40Unhelpfulhrm, alloca should be reasonably standard...
11:02:14ucchanalloca.h include cygwin's include files.
11:04:02 Quit bluebrother (Disconnected by services)
11:04:05 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
11:05:45Unhelpfulhrm, maybe a read function and offset to start of image data should just be stashed in struct mp3entry by the metadata reader? if both are set, albumart search can be skipped entirely, and anybody wanting to load the embedded album art opens the file, seeks, and passes the read function and the file handle to read_<format>_fd
11:06:00S_a_i_n_tucchan: Um...I posted a log of the error yesterday.
11:06:06S_a_i_n_tI'll try find it for you.
11:07:03Unhelpfulthis is assuming that adding a pointer and an int to the struct is not unreasonable, and that no state is needed other than the read function and the open file handle
11:07:41*Unhelpful would call a format that requires more than that just to read a chunk of embedded data unreasonable, anyway
11:07:49JdGordonjhMikeS: what do you wonder about 64bit sims?
11:08:05*JdGordon can test if you cant...
11:08:34S_a_i_n_tucchan: anyone else interested in telling me why is failing:
11:08:40 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
11:08:51S_a_i_n_tThat's the output from trying to build the arm toolchain.
11:09:46CIA-5New commit by 03jdgordon (r26370): fix error when a comment is on the last line of the file
11:10:13JdGordonS_a_i_n_t: hey, that skin updater app is now i think finished if you could help testing :)
11:10:20JdGordonjust run make in the utils/skinupdater folder I need is a working build environment now ;)
11:10:50ucchannow I add INCLUDE_PATH into the include file's path. building binutils...
11:10:52S_a_i_n_t(I'm working on it)
11:11:28JdGordonna, regular gcc is all that is needed But I'll still need a build to test it on. Guess I can just grab a daily, but Rockbox seems foreign to me without my pulled-apart/patched to hell version :P
11:13:25JdGordonah yeah, im more interested if the output looks right now :)
11:14:11S_a_i_n_t"looks right" doesn't really matter id it doesn't *work* right ;)
11:14:23S_a_i_n_tbut, I know what you mean.
11:14:30JdGordonno, looks right is more important now...
11:14:43JdGordonif the new parser is wrong then that can be fixed if we find issues
11:14:52JdGordonthe updater needs to dump correct output
11:15:29 Quit kramer3d (Quit: Leaving)
11:15:49 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
11:16:11ucchanINCLUDE_PATH does not add unnecessary usually. but missing include file's path my cygwin why?
11:16:27 Quit esperegu (Read error: Connection reset by peer)
11:17:34 Join esperegu [0] (~quassel@
11:19:18S_a_i_n_tucchan: Did you update CygWin recently? I'm thinking something changed in the current version that's breaking it.
11:19:36S_a_i_n_tbut, I can't remember the version I had previously that worked :/
11:20:02ucchanS_a_i_n_t: cygwin 1.5.25 gcc: 3.4.4
11:20:32S_a_i_n_tHmmm...I had 1.7.(something) and it was working fine.
11:20:46S_a_i_n_tI have no idea what's going on, but it's annoying.
11:24:48ucchancygwin 1.7... many problems exist on non C locale env...
11:27:57 Quit Navi (Ping timeout: 265 seconds)
11:35:07CIA-5New commit by 03jdgordon (r26371): be smarter about %xd so it only eats 2 letters if that image has subimages
11:36:12 Join aim_ [0] (~hamish@
11:39:19 Quit aim_ (Client Quit)
11:41:37ucchannow binutils build succeed. but my cygwin's env does not install GMP and MPFR. then gcc-4.4 can not build.
11:42:46Unhelpfulucchan: they're not available for cygwin?
11:46:35 Join merbanan [0] (
11:49:11gevaertsJdGordon: no. I found out some bits (such as needing a boolean as well, I need the callback :), and I know I need a few lang strings, so I need to look into those, but the main issue I have is where to store the string. I need it in firmware/, so it should be stored in firmware/, and as far as I can see that's not really straightforward
11:49:35JdGordonwhats it for?
11:49:35ucchanUnhelpful: I try install GMP, MPFR from source (binary does not exist)
11:50:07gevaertsthe name of a file to use as a disk image :)
11:50:33JdGordonany reason you couldnt just strcpy() the string somehwere more local?
11:50:58ucchanGMP can build on cygwin 1.5 and gcc 3.4.
11:51:27Unhelpfulyes, but a binary would be nicer :)
11:51:35gevaertscopy the pointer would be smarter I guess, but there's no callback...
11:51:37 Quit esperegu (Remote host closed the connection)
11:52:24gevaertsI'll need an enable/disable setting as well, so I could copy it from that callback
11:52:38JdGordongevaerts: this is for the mount iso image thing?
11:52:59gevaertsYes, basically
11:53:24gevaertsI'm implementing the storage_*() api for loopback
11:53:26JdGordonwhy not add a context menu option instead? you dont actually need it to be persistant do you?
11:53:42gevaertsgood point
11:53:58gevaertsso basically a context menu that says "Use this file, starting from now"
11:54:29JdGordonmore like "mount this file immediatly... "
11:54:50gevaertsYes, that's what "use" means :)
11:55:26gevaertsThat also ensures that the file is there when the thing starts
11:56:40JdGordonare you doing this for RaaA also?
11:56:52JdGordonI mean using a file for the disk
11:57:06gevaertsno, this is a different idea
11:57:37gevaertsOK, I'll look into that. After that, there's the slight issue that loopback is basically hotswap, and the hotswap code assumes that there's only one driver providing hotswappable drives, and the slight issue that the main usecase is USB and USB unmounts everything on connect so the loopback file is gone...
11:58:06JdGordonsounds like a nice helping of spaghetti there :)
11:58:36gevaertsThe good thing is that the end result should be some nice cleanups :)
11:59:28gevaertsAfter that, someone should add MMC support to MMS, so we actually emulate a real CD drive
11:59:48gevaertsAnd then, we can start thinking about exporting playlists as audio CDs
12:00:29gevaertsAfter that you'll finally be able to re-rip your mp3 collection without having the originals :)
12:02:13***Saving seen data "./dancer.seen"
12:06:03JdGordonisnt audio cd support completly redundant? I mean if something has usb it very probably has audio in
12:08:04gevaertsIf something has usb and is meant for audio, I actually assume that it probably might not support CD drives at all
12:08:35gevaertsThe audio CD bit is a bit silly, but general CDROM emulation might be useful so iso images actually work as expected
12:10:57JdGordononly if it supports dvd iso images also :)
12:11:16JdGordonand yes it would be fun to keep a bunch of boot cds on a clip
12:11:51 Join halmi [0] (
12:12:10 Quit liar (Quit: Verlassend)
12:15:30 Join Guest84887 [0] (
12:15:43 Quit Guest84887 (Client Quit)
12:38:48 Join petur [0] (~petur@rockbox/developer/petur)
13:03:03 Join moos [0] (moos@rockbox/staff/moos)
13:06:35 Join JdGord [0] (~jd@
13:12:07ucchanI quit soon. I report the build result tommorow.
13:12:21 Quit ucchan (Quit: Leaving...)
13:14:54*S_a_i_n_t wonders if anyone has an ARM-EABI toolchain built for CygWin that could .zip it up/upload it somewhere for him.
13:15:15S_a_i_n_tBlue_Dude, if you come in and notice this: ^
13:19:43AlexPpixelma: You about?
13:20:38 Join halmi_ [0] (
13:21:50AlexPI'm just about to fix the plugins now using pla keys in the manual, and think it'd be cleaner and easier on the long run to add the PLAs to the platform keymap files. Thoughts?
13:22:14 Quit halmi (Ping timeout: 260 seconds)
13:28:09 Join robin0800 [0] (
13:31:56 Join Kitar|st [0] (
13:34:35 Quit Kitr88 (Ping timeout: 265 seconds)
13:36:50 Quit Kitar|st (Ping timeout: 272 seconds)
13:39:10 Join Kitar|st [0] (
13:39:52 Quit shai (Ping timeout: 260 seconds)
13:44:12AlexPD2 is touchscreen?
13:45:04 Join wincent_balin [0] (
13:45:16 Quit robin0800 (Remote host closed the connection)
13:50:59S_a_i_n_tyep, AFAIK
13:51:07S_a_i_n_tre: D2
13:59:54 Join halmi [0] (
14:02:14***Saving seen data "./dancer.seen"
14:02:40AlexP{PLA_EXIT, BUTTON_PLAY|BUTTON_SELECT, BUTTON_NONE }, means you have to press both play and select?
14:03:23 Quit halmi_ (Ping timeout: 276 seconds)
14:03:25S_a_i_n_tI think so, I remember that from the Hotkey keymap mess
14:04:22 Join shai [0] (
14:05:41 Join halmi_ [0] (
14:07:50 Nick fxb__ is now known as fxb (
14:08:06 Quit halmi (Ping timeout: 240 seconds)
14:09:01 Quit flydutch (Quit: /* empty */)
14:10:23 Quit JdGord (Quit: Bye)
14:14:13 Join esperegu [0] (~quassel@
14:16:04 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
14:19:57 Quit Rob2223 (Quit: Rob2223)
14:24:52 Join Rob2222 [0] (
14:30:53 Join anewuser [0] (anewuser@unaffiliated/anewuser)
14:34:54 Join funman [0] (~fun@rockbox/developer/funman)
14:39:27 Join pyther [0] (~pyther@unaffiliated/pyther)
14:42:56funman <- does it look like a 5 minutes improvement, or random variation ? (green line is patched)
14:44:31S_a_i_n_tlooks like an improvement to me, if its accurate.
14:45:03S_a_i_n_t*slightly* more than 5 min I'd say too...perhaps ~7?
14:45:34funman7'39'' :o)
14:46:00funmanbut the battery wasn't exactly at the same value when beginning
14:46:00S_a_i_n_twhat did I win? ;)
14:46:36funmanif you guess the milliseconds you get all my sansas
14:47:01S_a_i_n_tHmmmm... "31"
14:47:04 Join efyx [0] (
14:51:29 Join kugel [0] (~kugel@rockbox/developer/kugel)
14:51:36kugelAlexP: yes
14:52:00kugelorder or timing doesn't matter if the third column is BUTTON_NONE
14:52:31AlexPI'll just put that is as Play + Select
14:53:24 Join steve|m1 [0] (
14:54:27 Join teru [0] (
14:55:04 Quit steve|m (Ping timeout: 276 seconds)
14:55:45 Quit steve|m1 (Client Quit)
14:56:04 Join steve|m [0] (
14:57:16AlexPkugel: So for e.g. alarmclock on a scrollwheel target (e.g. ipod video), both scroll fwd/back and up/down (menu/play) will change the time?
14:57:24AlexPThanks for the list by the way :)
14:59:18rasherI wonder what that guy wanting to use ORCA is trying to achieve
14:59:40rasherIt just uses festival or espeak anyway...
15:07:01domonokyrasher: afaik ORCA is a screenreader not a TTS.
15:07:50 Join DataGhost [0] (
15:07:50 Quit DataGhost (Changing host)
15:07:50 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
15:08:05rasherSo it makes absolutely no sense
15:10:45funman < clip+, here the difference is 10 minutes (over 14 hours)
15:12:41funmanFlynDice: i'm testing lcd invert on fuzev2 -> no plugins crash
15:15:42 Join panni_ [0] (
15:21:15 Quit kugel (Ping timeout: 260 seconds)
15:25:11 Quit petur (Quit: later)
15:25:20*AlexP summons kugel
15:25:48funmanFS #11329 : lcd_flip for fuze v1/v2
15:26:54 Quit n17ikh (Ping timeout: 240 seconds)
15:28:14funmanFS #11330 : disable peripheral clocks for Clip* LCD, please run comparisons to avoid measurement error
15:33:38 Quit merbanan (Ping timeout: 264 seconds)
15:34:22 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:34:32FlynDicefunman: fuzev2 plugins seem to work fine for me now with the same patch I used the last time for LCD upside down.
15:35:22funmanrandom bugs hit us again ? :o
15:36:00 Join kugel_ [0] (
15:36:19 Quit kugel (Disconnected by services)
15:36:22 Nick kugel_ is now known as kugel (
15:36:29 Quit kugel (Changing host)
15:36:29 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:36:31kugelAlexP: yea
15:37:13funmanFlynDice: i'll commit it then
15:37:44AlexPkugel: It seems that some keys have been lost on some targets - e.g. bubbles menu used to be save on the S, but it is now PLA_EXIT (or cancel, I forget) so there is no longer anyway to save it other than via the menu on the S (maybe for other targets too). Is this intentional?
15:38:00funmanone of you knows where the disk led icon is ?
15:39:32funmanbitmap_icon_disk in apps/recorder/icons.c :/
15:41:04 Quit kugel (Ping timeout: 260 seconds)
15:42:04 Join evilnick_ [0] (
15:43:20CIA-5New commit by 03funman (r26372): Fuzev1/v2 : lcd_set_flip() ...
15:44:36AlexPkugel: Next example - for clock the manual suggests that there used to be two buttons - one to go to the menu, and one to save and quit directly. It now (on the S at least) only seems possible to go to the menu
15:44:42 Quit evilnick (Ping timeout: 245 seconds)
15:46:24 Join liar [0] (
15:48:18FlynDicefunman: re lcd upside down I did update the fuzev2 bootloader several times yesterday looking at the firmware upgade difficulties Rob2222 was having. So if it causes problems for someone perhaps that's the first suggestion.
15:49:30JdGordonTorne: what do you want to do about the font "fix"? the patch looks ok to commit
15:49:41funmanhm now i remember ssorgatem mentioned problems with fuzev2 lcd (white screen) related to bootloader
15:49:51Rob2222FlynDice: any new patch for that aviable, yet?
15:50:42funmanFlynDice: can you give a try to on e200v2 and see if mpegplayer still works?
15:51:03FlynDiceRob2222: Don't hold your breath, I don't know what's going on there...
15:51:08FlynDicefunman: sure
15:51:14 Join evilnick [0] (
15:51:53Rob2222FlynDice: yeah, no problem. i'm still collecting information about that in different forums
15:52:22Rob2222but definitely im not the only one with that
15:52:37 Quit evilnick_ (Ping timeout: 245 seconds)
15:54:04 Quit bmbl (Ping timeout: 258 seconds)
15:54:30 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:54:45kugelAlexP: there's no save in bubbles
15:55:05AlexPThere is, you can quit or quit without saving
15:55:09AlexPthen resume next time
15:55:25kugelyes, that's in the menu
15:55:31kugelthere's no key for that
15:55:41AlexPSo the manual has always been wrong, as it suggest there used to be a key
15:56:48*AlexP adds a ? to the last sentence
15:57:12kugelAlexP: I assume some hotkeys got lost when the plugins were converted to have a menu
15:57:37kugelI remember there was a biggish commit, it added a menu for some plugins which can hold the playback controls
15:58:28kugelsame goes for clock.rock
16:00:02 Quit xavieran (Ping timeout: 240 seconds)
16:00:08JdGordonrasher: scorche: ping?
16:00:08AlexPkugel: demystify - you definitely used to be able to quit it using the h100 remote - I haven't tried, but I don't see that in the source now?
16:00:57CIA-5New commit by 03jdgordon (r26373): make the script zip the output
16:01:32kugelAlexP: should be possible
16:01:53AlexPkugel: As in possible to add, or currently there? :)
16:02:06AlexPah yes, sorry
16:02:15***Saving seen data "./dancer.seen"
16:02:15AlexPmy mistake
16:02:22AlexPI was looking in the wrong place :)
16:02:56kugelthe pla_remote_ctx actually only holds directions
16:03:05kugelsome remote keys are in the main_ctx
16:03:10kugelI assume that's confusing
16:03:25AlexPIt was actually something more silly :)
16:03:40 Join webguest54 [0] (
16:03:50kugelbut on some remotes you can only have a subset of the buttons, so the remote_ctx would be incomplete, that's why i did it that way. but it can be changed
16:04:49AlexPI'd done that fine - but when I added the pla to the keymap files, I had to add main unit and pla actions seperartly, so when I then looked at the demystify source I confused my self and was looking for PLA_RC_* (or some such), forgetting that the remote buttons were already defined in the PLA_* :)
16:05:28kugelah ok :)
16:06:01webguest54Anyone know that bootloader can compile with arm-eabi?
16:06:54Rob2222FlynDice: I think theres a new additional check performed in the firmware update routine. Now I just wonder if this routine (that flashs the *.bin to the cip itself) is within the firmware and so in the *.bin itself, or if this is maybe additional code that is in the player that gets not updated while flashing, like a hard installed firmware-upgrade-loader. But I have no idea how to figure that out atm.
16:07:11AlexPkugel: This is pretty cool actually, as it automatically adds quite a lot of remote buttons to plugins
16:08:10kugelyea, I'm not even sure anymore if the pla_remote_ctx should be separate
16:08:37 Join Rob2223 [0] (
16:09:23 Join S_a_i_n_t_ [0] (S_a_i_n_t@
16:09:41 Quit Rob2223 (Read error: Connection reset by peer)
16:09:52 Join Rob2223 [0] (
16:10:52 Quit S_a_i_n_t (Ping timeout: 248 seconds)
16:12:02 Quit Rob2222 (Ping timeout: 264 seconds)
16:12:58gevaertsJdGordon: how do I add something to the context menu?
16:13:13 Quit webguest54 (Quit: CGI:IRC (EOF))
16:13:20JdGordonwhich context menu?
16:13:27gevaertsfile browser
16:13:49JdGordon(that really needs to be renamed :p )
16:14:18FlynDicefunman: mpegplayer doesn't work with that patch. I get sound but no video.
16:14:18JdGordoncreate a MENUITEM_FUNCTION() and add that to the menu
16:14:28funmanFlynDice: ok
16:14:49 Quit Topy44 (Quit: No Ping reply in 180 seconds.)
16:14:57funmanbtw lcd inversion (colors) still works if you let the backlight go off and on again ?
16:15:26FlynDicelet me look
16:15:31 Join Topy44 [0] (
16:15:55FlynDicecolor inverted not screen upside down right?
16:17:31FlynDiceworks fine
16:17:39kugelfunman: why the changes to the window function?
16:18:13funmankugel: i want to move e200v2/fuze common code to a single file
16:18:57kugelthat doesn't really answer my question
16:19:18kugelI mean the s/R_HORIZ_RAM_ADDR_POS + 2/0x46/ sort of changes
16:19:25FlynDicefunman: I misunderstood your question I think
16:20:06funmankugel: i found it more clear that way in fact
16:20:14FlynDicefunman: on e200v2 with color inverted it does indeed switch back to normal(uninverted) once the screen goes dark
16:20:21funmanah ok
16:20:23kugelfunman: "more clever"?
16:20:34JdGordonAlexP: while you're playing with plugins... is the backdrop from the theme being displaed in the plugins menus like they should be?
16:20:48funmankugel: no, clear
16:20:52kugelI think if it has a reason it should be explained in the commit message. and it obviously wasn't related to lcd flip so it should be a separate commit
16:20:58kugeloh misread, sorry
16:21:14funmankugel: well it's just a revert of something i had made earlier
16:21:38kugelreverts should be even more commented on
16:21:40funmani should have mentioned it in the log
16:21:42 Join mischasworld [0] (
16:21:48funmanjust cosmetics anyway
16:22:18 Join merbanan [0] (
16:22:19funmanFlynDice: i'll revert 26353
16:24:14CIA-5New commit by 03funman (r26374): e200v2: revert r26353 ...
16:25:39 Quit wincent_balin (Changing host)
16:25:39 Join wincent_balin [0] (~wincent@rockbox/developer/wincent)
16:26:46AlexPJdGordon: Don't know, I'm looking at the code not a sim etc. I'll have a look in a bit :)
16:27:35JdGordondont worry :)
16:36:06 Quit wincent_balin (Ping timeout: 240 seconds)
16:37:01 Quit kugel (Ping timeout: 252 seconds)
16:37:32 Quit merbanan (Ping timeout: 248 seconds)
16:37:57FlynDiceis there an easier way to backport r26365 to 3.6 or do I need to checkout the branch with svn?
16:39:41JdGordonyou need to checkout the branch
16:39:58FlynDiceok thanks...
16:43:08 Join solexx [0] (
16:43:33*JdGordon has a feeling noone particularly cares about the upcoming skin changes, but there will still be chaos when it happens :/
16:46:13 Quit solexx_ (Ping timeout: 276 seconds)
16:47:24 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
16:50:57 Join merbanan [0] (
16:55:57JdGordonbieber: im having a look at the theme editor code.. please stop using the name WPS... call it skin instead
16:56:13 Nick fxb is now known as fxb__ (
16:56:26CIA-5New commit by 03Buschel (r26375): Small changes to asm for better readability.
16:56:57 Join Buschel [0] (
16:57:37funmanFlynDice: can you give a try to and see if e200v2 lcd still worsk ?
16:57:49funmani changed lcd_enable() to be like the fuze
16:58:05funmanre activation is simpler and doesn't run the full _display_on()
16:58:11funmanhmm wait
16:58:20funmani didn't check if the register settings were identical :o
16:59:18*FlynDice waits patiently watches git svn clone branch 3_6 for amusement...
16:59:34JdGordonbad idea...
17:00:12CIA-5New commit by 03Buschel (r26376): Correction to clobber lists of several codec's inline assembly.
17:00:54gevaertsWhere's my menu item?
17:01:49funmanFlynDice: pastie updated
17:03:21JdGordongevaerts: on the couch where you left it!
17:03:31*gevaerts looks
17:03:36 Join kugel [0] (~kugel@rockbox/developer/kugel)
17:03:37gevaertsOh, yes, there it is!
17:04:09JdGordonhehe, did you add it to the right menu?
17:04:13JdGordoni.e not the wps menu?
17:04:47gevaertsI hadn't added it to the checker thing
17:05:47JdGordonthe checker thing?
17:06:10funmanafaics fuze/e200v2 send reversed BGR colors to the controller
17:06:23 Quit esperegu (Ping timeout: 264 seconds)
17:06:42funmanand the fuzev2 send the bytes in a swapped order, how more twisted can that be ?
17:06:48JdGordonwhy are you adding it to that?
17:06:56gevaertsbecause I copied code!
17:07:35JdGordonshow me a diff?
17:07:44JdGordonyou just want it on the file menu right?
17:08:36 Quit mischasworld (Remote host closed the connection)
17:08:59 Join Navi [0] (
17:09:00JdGordonand you shold be good
17:09:03 Quit Navi (Changing host)
17:09:03 Join Navi [0] (~Lancelot@unaffiliated/navi)
17:09:12gevaertsYes, but IIUC this way it also checks if it's actually a file
17:09:48JdGordonwell yeah
17:10:00JdGordonpresumably you only want certain file types though right?
17:10:08gevaertsno idea yet
17:10:21JdGordonok, if you do then create a seperate callback for it
17:11:22Buschelcan anyone with a Cowon D2 test the patch in FS #11240 ?
17:14:37 Join esperegu [0] (~quassel@
17:16:13FlynDicefunman: can't get that patch to compile, keep getting this error −−->
17:16:31funmanFlynDice: make dep should fix that
17:16:58FlynDiceok I'll try that
17:18:07FlynDicesame problem
17:19:06 Part Navi ("Leaving")
17:19:34FlynDiceI cleaned out the folder and started from scratch and still had the error
17:19:55 Quit kugel (Ping timeout: 252 seconds)
17:21:53 Join kugel [0] (~kugel@rockbox/developer/kugel)
17:22:00FlynDiceI'll clean it out & repatch and see what happens
17:22:11funmantry just editing make.dep and remove the .h from here
17:23:03CIA-5New commit by 03alex (r26377): Add the plugin lib actions to the manual keymap files.
17:23:12CIA-5New commit by 03alex (r26378): Add non-defined remote actions to the manual keymap files
17:23:17CIA-5New commit by 03alex (r26379): Correct typos in D2 keymap file.
17:23:24CIA-5New commit by 03alex (r26380): Correct manual keymaps for plugins recently converted to PLA, and use manual PLA defines for them.
17:23:57AlexPRight, that ought to go in 3.6 too I suppose
17:29:15FlynDicefunman: out of time for now, I'll try a bit later
17:31:17kugelAlexP: that was fast!
17:31:37AlexPkugel: I hope it is right :)
17:31:45AlexPThey build at least :)
17:31:54kugellots of "code" removed, it must be right :)
17:32:06AlexPAlthough I forgot to mention the flyspray task :/
17:32:09Rob2223FlynDice: Ive found a little difference in the furmware files that my Fuze2 accepts. The to he likes have 0x24 at offset 0x408, and all older files have 0x18 at offset 0x408, the RB-patched has code there already. can it be that this is a data section and the new FW checks for a 0x24 there?
17:32:47gevaertshm, interesting
17:33:05gevaertsFAT keeps working if you have a file open even if you unmount the drive
17:34:46funmanRob2223: this is normally code at this place
17:35:07funmannamely, software interrupt handler
17:36:02 Quit teru (Quit: Quit)
17:36:11Rob2223funman: but in the OF there is text at 0x420
17:36:34funmanyeah nothing prevents you to put ascii text in code
17:37:55funmanread mkamsboot.c or amsinfo.c if you want to see what we know about the file format
17:38:38kugelRob2223: can you downgrade to an older OF?
17:39:34Rob2223no, it just flashes 02.03.31-OF and .33-OF
17:39:53kugelmaybe they changed the checksum calculation
17:40:22Rob2223cause of that im researching whats even in 31 and 33 and looks whats change from them both to the older OF-files
17:40:29funmani don't think we check if their checksum is correct
17:41:01Rob2223yeah, i think checksum is fine, cause the _identical_ .33-rb-patched is working on other fuzes
17:41:22kugelI thought it was a general problem with 2.03.xx OFs
17:41:36 Join petur [0] (~petur@rockbox/developer/petur)
17:41:54Rob2223funman: is 0x460 code, too? this byte differs, too. and much 00's around
17:42:21Rob2223i will get the mkam* file and study it
17:42:53Rob2223kugel: it doesnt flashes 02.03.28, too
17:42:57funmanafaik 0x0-0x400 is the header, everything else is code/data, except the last 32bits of the file which are another checksum
17:43:25funmankugel: you have updated your OF ?
17:43:39funmanwhich one you have?
17:43:50Rob2223funman: yeah but if the data is same for much OFs' the could check a byte of that data like a flag, before allowing flash, or?
17:44:02kugel2.02.17 or .26 I think
17:44:55funmanRob2223: i have the same info than you
17:45:04Rob2223and much codein that area is identical in the OFs
17:45:24 Join halmi [0] (
17:45:31funmanno, add an instruction and all offsets change
17:46:04funmanthat said there are parts of the OF which we don't know what they are
17:46:31 Join phanboy4 [0] (
17:46:42Rob2223well, 0x400-0x600 only differ in 3 bytes, which 1 is the OF version numbe, in OF.31 and OF.33
17:46:59Rob2223i understand
17:48:28 Quit halmi_ (Ping timeout: 260 seconds)
17:48:43kugelfunman: I surely don't dare to update now :)
17:49:07funmanif you want to try and it doesn't work we can exchange fuzes at devcon
17:49:52Rob2223atm i think it could be that they (my fuze2) checks 0x408="0x24" or 0x460="0x44"... at least its even in .31 and.33 and differs in all rejected FWs
17:50:03kugelI can give you mine to play with at the devcon but I need a working one back ;)
17:50:35funmanthat works
17:55:13CIA-5New commit by 03alex (r26381): Add the plugin lib actions to the manual keymap files.
17:55:18CIA-5New commit by 03alex (r26382): Add non-defined remote actions to the manual keymap files
17:55:22CIA-5New commit by 03alex (r26383): Correct typos in D2 keymap file.
17:55:26CIA-5New commit by 03alex (r26384): Correct manual keymaps for plugins recently converted to PLA, and use manual PLA defines for them.
18:02:18***Saving seen data "./dancer.seen"
18:03:08*gevaerts plays audio from a loopback-mounted FAT filesystem image \☺/
18:06:01AlexPgevaerts: What would this let us do
18:07:02kugelAlexP: play audio from a loopback-mounted FAT filesystem image :)
18:07:13AlexPum, thanks :)
18:07:49kugelprobably a proof of concept for mounting images, more useful for mounting iso cd images (going by the recent talk)
18:10:55gevaertsAlexP: as soon as the hotswap code has been cleaned up, and the USB code knows about exporting only selected disks, this will allow you to have bootable images on your DAP. Later on, if we ever add MMC support, that would extend to cdrom emulation
18:11:13AlexPOK, cheers
18:11:13gevaertsRight now it's just a fun experiment though
18:11:54gevaertsIt actually does work over USB, but that's an accident, and if you mount the underlying FS on your PC as well, you'll probably see interesting things :)
18:12:58 Join esperegu_ [0] (~quassel@
18:13:02 Quit esperegu (Read error: Connection reset by peer)
18:15:18 Quit TheSeven (Remote host closed the connection)
18:20:25funmanis thumb calling convention different from arm ?
18:20:44funmanI see a 'mov r7, r11' as first insrtuction after pushing registers on the stack
18:21:38 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
18:22:58funmanah never mind it's for stacking it
18:25:05kugelgevaerts: I hope I can get the #ifdef SIMULATOR cleanup done this week
18:26:12*kugel decides to ignore mpeg.c
18:29:57kugelfortunately apps/ doesn't have too many #ifdef SIMULATOR
18:35:22Buschelhmm, walking through the manual brings up a (minor) question: do we really have -72, -73 and -74 dB as lowest possible volume on different targets? or is it just an effect of different counting or interpretation (e.g. counting -74 dB as lowest step even it is equal to off) the same is valid for -89 and -90.
18:37:44 Join domonoky1 [0] (
18:38:27 Quit domonoky (Ping timeout: 245 seconds)
18:39:01funmanBuschel: which targets?
18:44:29 Quit panni_ (Read error: Connection reset by peer)
18:44:42 Join panni_ [0] (
18:46:49funmanah memset has an incorrect return for thumb/armv4
18:47:49funmani'm gonna try clipv2
18:48:20 Join thehodapp [0] (
18:48:29 Quit bmbl (Ping timeout: 258 seconds)
18:49:36 Quit thehodapp (Client Quit)
18:53:21funmanboots fine but plugins/codecs crash
18:53:42funmanram usage almost 100kB down
18:53:52funmanand i didn't have to use -mlong-calls :?
18:55:15 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
18:57:14funmani just fixed cpucache_invalidate and playback/plugins are fine
19:01:26 Nick S_a_i_n_t_ is now known as S_a_i_n_t (S_a_i_n_t@
19:04:54 Join toffe82 [0] (
19:06:30 Join orzech [0] (
19:07:19orzechhey, is there a possibility to change functions of keys of sansa_clip_plus player?
19:07:20CIA-5New commit by 03FlynDice (r26385): Backport 26365: sd-as3525.c: Do not wait for TRAN state after transfer is over. ...
19:10:09orzechso that for example. button 'volume up' changed volume to down and vice versa
19:11:48AlexPorzech: Yes, but you will need to edit the source then compile it
19:12:44CIA-5New commit by 03funman (r26386): cpucache_invalidate: use bx reg instead of mov pc, reg to return ...
19:14:57orzechAlexP: no problem, is there any doc? or maby somone with deaper insigth might tell what to look for in source code?
19:16:00AlexPorzech: Edit apps/keymaps/keymap-clip.c
19:16:14orzechAlexP: thx :)
19:16:14AlexPWell, I assume that it is that one and that the clip+ is the same as the clip
19:16:40orzechmost probably it is
19:16:41funmanit's teh same
19:19:14 Quit moos (Ping timeout: 240 seconds)
19:19:52 Quit z35 (Quit: Leaving)
19:26:02pytherIt seems lke it would be a pain to use a rockbox with the sandisk clip considering how small it is
19:26:39S_a_i_n_tmany people quite like it.
19:27:03 Quit liar (Quit: Verlassend)
19:27:29 Quit S_a_i_n_t ()
19:28:52 Quit Buschel ()
19:29:10funmanam I the only one to have test_codec never exit ?
19:31:20funmanthe goto is suspicious
19:31:44 Quit rvvs89 (Changing host)
19:31:44 Join rvvs89 [0] (ivo@pdpc/supporter/base/rvvs89)
19:33:28 Join efyx_ [0] (
19:33:48 Quit efyx (Read error: No route to host)
19:34:18 Quit phanboy4 (Ping timeout: 245 seconds)
19:38:20orzechpyther: I'm ipod shuffle (2gen) and what I dislike about clip is fact that it is hard to control without accualy looking at screen
19:40:39CIA-5New commit by 03funman (r26387): Fuze: don't change gate scan position if lcd_set_flip() is not used
19:41:32orzechand the fact that clip is designed to be used verticaly - I like to clip plyerts to my belt, and sansa clip ends up in horizontal position
19:42:17orzechwith rockbox I hope to overcome those limitations :)
19:42:54 Quit stoffel (Ping timeout: 265 seconds)
19:53:31funman -> files built with -mno-thumb
19:57:11r0b-orzech what clip?
19:57:51kugelfunman: is there no way to manually fix up inline asm?
19:58:09kugelI expect enable/disable_irq is the only reason most of those files contain inline asm
19:58:13funmanyes there is
19:58:24funmanmake them real functions and move them to separate files
19:59:05kugelwhat about blx pc, nop, nop?
19:59:26funmanwhat about my solution?
20:01:31funmancodecs are a bit slower except aac
20:01:50funmanand ac3
20:01:57orzechr0b-: clip+
20:02:15r0b-you have Rockbox on it?
20:02:20***Saving seen data "./dancer.seen"
20:02:28orzechyes, installed 1h ago
20:02:33funman <- and lame_256.mp3 fits in the buffer with thumb :P
20:03:07orzechand from the looks of it, it seems to be working
20:04:20kugelfunman: what about thumb versions of enable/disable_irq?
20:05:26funmannot possible
20:05:40kugelno mcr/msr in thumb?
20:05:49funmananyway my script just checks for 'asm'
20:06:10funmanif you have a variable named 'XXXasmYYY' it will be built with -mno-thumb (default)
20:06:52funmanof course we could also try to build the file first with -mthumb and fallback to -mno-thumb if it fails
20:07:10funmanit could turn to be faster than parsing the preprocessor output
20:08:05funmanwhat i would like to know is if modifying the returns to be thumb compatible is ok because it might mean add 1 instruction
20:08:55r0b-orzech how hard was the RB Install on a clip?
20:09:01r0b-clip +?
20:10:59orzechr0b-: almost 'copy&play' :)
20:11:17r0b-heh i rockboxed me e260 v1
20:11:43orzechI followed instructions form the manual - took like 15 mins - my first time to use rockbox who
20:13:01orzechwell, I do not know much about sansa player - I just whated some Ipod shuffle alternative :)
20:13:20*r0b- has a 1GB 2nd gen shuffle
20:13:24r0b-no charger
20:14:18orzechI alos have 1gb 2ed gen shuffle, and infortunatelly as far as design/usability goes shuffle>clip+ ;(
20:14:35funmanamiconn: <- works on clipv1. We could skip that on armv5+ but I guess for 1 instruction it's not a problem?
20:14:55orzechthat's why I decided to use rockbox
20:15:41 Join shodan45 [0] (
20:16:15funmanaudiobuffer 773kB on clipv1, voice almost works!
20:16:37 Join stoffel [0] (
20:18:05shodan45what eats more battery: playing oggs from internal flash or playing flacs from microsd? (fuze v1)
20:18:42shodan45the oggs are "q6" iirc, if that matters
20:20:27AlexPWe have two different chinese (simple) translations FS #11327 and FS #11290 - any thoughts on which to commit?
20:20:45funman-Os => another 30kB
20:20:59gevaertsAlexP: proof-read both and pick one :)
20:21:10AlexPgevaerts: er, yes.... :)
20:21:36gevaertsMaybe markun can help (or maybe he knows people who can)
20:23:01funmanperhaps suggest them to collaborate
20:23:10funman(both translations authors)
20:23:35AlexPyeah, I'll put a note on for them
20:31:17 Join XXStatik [0] (
20:35:41 Quit XStatik (Ping timeout: 276 seconds)
20:37:51 Quit kugel (Remote host closed the connection)
20:40:06ranmafunman: Anything against reducing pluginbuffer to 0x10000 on clipc200v2?
20:51:17 Join saratoga_ [0] (~463f90ed@gateway/web/freenode/x-syeugwqqlfrfyjxy)
20:56:41 Quit XXStatik (Ping timeout: 260 seconds)
20:58:15r0b-orzech might i PM you?
21:02:12CIA-5New commit by 03alex (r26388): Add a brief description of radio art to the manual. Closes FS #11279.
21:02:21CIA-5New commit by 03alex (r26389): Only show the section on radio tags if the target has a radio.
21:02:26CIA-5New commit by 03alex (r26390): Consolidate the information on album art into one place.
21:03:31r0b-is there a way for Rockbox to decrypt like id3 tags broadcasted by radio stations?
21:03:37r0b-for the Artist/Song?
21:04:12gevaertsYou mean RDS?
21:05:18orzechr0b-: sure, but I need to be going now, so I will answer you when I caome back
21:05:26orzecheta. 8h
21:06:23CIA-5New commit by 03alex (r26391): Add a brief description of radio art to the manual.
21:06:30CIA-5New commit by 03alex (r26392): Only show the section on radio tags if the target has a radio.
21:06:32AlexPr0b-: Sort of
21:06:38CIA-5New commit by 03alex (r26393): Consolidate the information on album art into one place.
21:07:02AlexPr0b-: It is only implemented for the ipod FM accessory
21:07:09saratoga_so if the clip+ i switched to thumb it'll almost double the audio buffer?
21:07:13AlexPr0b-: Although the gigabeat S tuner has the capability
21:07:38r0b-i dont think the e200v1 has the ability
21:08:16r0b-oh well
21:11:01r0b-im about to purchase a Clip + as soon as i get to best buy
21:11:27 Quit stoffel (Remote host closed the connection)
21:12:20funmanranma: clipv1 is already no?
21:13:01funmanranma: ah, some plugins didn't build on c200v2 (ppmviewer at least)
21:29:22funmanranma: btw i have warnings in kernel.c with -Os
21:30:15tmztr0b-: what are those selling for?
21:32:17funman& rockboy
21:42:50 Join Jaykay [0] (
21:47:41r0b-for a 4G
21:50:07 Join Jaykay_ [0] (
21:51:26 Quit Jaykay (Disconnected by services)
21:51:32 Nick Jaykay_ is now known as Jaykay (
21:54:49saratoga_does the clip+ have a hold switch?
21:55:20r0b-no it uses a long home button press to trigger hold
21:57:08r0b-i dunno how rockbox handles it though'
21:58:15funmanand i see it's missing from the manual
21:59:20saratoga_a single button press might be a lot bettter for that
21:59:31saratoga_double button presses are fairly hard on the other clips because they're so small
21:59:48funmanthis combo is already used for go to wps
22:00:01funmanso 2 times the combo = go to wps + lock
22:01:25CIA-5New commit by 03funman (r26394): manual: add key combo for clip+ software hold
22:01:37r0b-have any of you other then orzech tried Rockbox on a Clip +?
22:01:47funmansome parts of the clip keymap need rework i think
22:01:49 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
22:02:10saratoga_i think holding home for lock is more useful then getting to the quick screen
22:02:22***Saving seen data "./dancer.seen"
22:02:53*funman never uses quick screen
22:03:13funmanfeel free to exchange the combos
22:03:26saratoga_i'll do it later when i'm at my main pc
22:03:36AlexPfunman: I think that wants to go in the branch as well
22:03:48saratoga_good point
22:03:53funmanAlexP: clip+ isn't in the release so it's ok
22:04:01AlexPah right
22:04:19funmanwell it could be IMO ;)
22:04:54saratoga_kugel: (for the logs): any chance you could take a look at "FS #11286 - PictureFlow segfault when showing tracklist or playing - fixed."
22:05:01saratoga_it sounds like something we would like to see committed
22:06:30saratoga_actually csavery has a few useful looking patches on the tracker, but I'm not familiar with anything they touch
22:07:41 Quit saratoga_ (Quit: Page closed)
22:08:33r0b-so rockbox is still very unstable on Clip v2/clip + right :?
22:24:59 Quit shai (Quit: Leaving)
22:27:48 Join stoffel [0] (
22:34:33 Quit Jaykay (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
22:38:34 Join wodz [0] (
22:39:04Rob2223r0b-: im using it since 2 days now
22:39:09Rob2223for playing MP3
22:39:14Rob2223works fine
22:39:53Rob2223even more you can boot the original firmware with a keypress if you want
22:40:24wodzcan someone look at . This doesn't work *BUT* if I substitute lcd_drawline with commented one it works (but don't use whole lcd width of course). I must miss something obvious
22:44:11CIA-5New commit by 03gevaerts (r26395): fix broken case handling in storage_present() and storage_removable(). Those were buggy for targets with a hotswappable drive *and* more than one ...
22:45:24r0b-Rob2223 what Clip + you got 2GB 4GB or 8GB?
22:45:40r0b-also rockobx has many things over the OF
22:45:50r0b-games and such
22:46:07r0b-its like 49$ for the 4GB
22:46:09r0b-is it worth it?
22:46:55wodzcould You move discussion to #rockbox-community please
22:47:17Rob2223well it has 8GB, is small and is rockbox capable. does that answer the question? :D
22:47:48 Nick Zarggg_ is now known as Zarggg (
22:48:17 Join Soap_Hotel [0] (
22:51:50funmanamiconn: Torne: please have a look at
22:57:53 Quit Soap_Hotel (Quit: CGI:IRC (Ping timeout))
22:59:03 Join Soap_Hotel [0] (
23:00:11 Quit stoffel (Remote host closed the connection)
23:06:18 Join wincent [0] (
23:15:53funmanmpegplayer code does the same
23:17:31 Quit wincent (Changing host)
23:17:31 Join wincent [0] (~wincent@rockbox/developer/wincent)
23:17:47r0b-i really should try to write an mpeg-4 player for Rockbox
23:18:46saratogago for it
23:18:47 Quit lpereira (Quit: Leaving.)
23:18:57saratogaanyone have an H10 handy?
23:20:07Rob2223why you ask?
23:20:30saratogai want something tested
23:21:23Rob2223ah ok. no sorry. i had an H140 and have a H340, but no H10
23:21:31gevaertssaratoga: which one
23:21:38saratogadoesn't matter
23:21:57gevaertsI have a small one, but I don't know if it's charged
23:22:00saratogai'm going to commit FS #10887 - Change poweroff to only check that the power button is pressed if someone can confirm that AB repeat mode still works with it
23:22:13saratogaeverything else is working
23:22:30gevaertshm, I'd need to find out what AB repeat mode is supposed to do then :)
23:22:56Rob2223repeat a part of a song from mark a to makr b continously
23:23:21saratogagevaerts: yeah I had to dig through the manual to figure out how to test it on the sansas . . .
23:23:37saratogai presume its one of those features that included because some OF had it
23:24:00gevaertsI can probably test later. The thing claims that the battery is too low to boot right now
23:24:23saratogawould you just try and use AB mode and let me know if it shuts down the player too easily
23:25:10gevaertsI'll have a go
23:25:15saratogaon the sansas its possible to accidentally shutdown while trying to use AB mode, but the extra time between when AB mode triggers and when the player shuts down is large enough that its not a real problem
23:28:00gevaertsI don't use the thing very often though, so my button-handling expertise might lead to misleading test results
23:30:00saratogai doubt anyone on the H10 uses AB repeat at all
23:30:03saratogaso its fine
23:31:04wodzcan someone test latest patch on FS #11324? Preferably on DAP with bigger screen.
23:33:14 Join CaptainKwel [0] (
23:35:43 Join shai [0] (
23:35:44saratogaare thre any devices without a hold switch besides the clip+?
23:35:57saratogai'm having trouble figuring out how the manual handles them
23:37:37 Quit esperegu_ (Read error: Connection reset by peer)
23:38:14saratogait seems like the manual doesn't cover software hold buttons
23:38:34AlexPsaratoga: It does, see funman's commit earlier
23:38:40 Join evilnick_ [0] (
23:38:50 Join esperegu [0] (~quassel@
23:39:04AlexPNot in great detail I'll admit :)
23:39:54gevaertssaratoga: seems to be fine
23:40:30saratogacan i change \ButtonHome+\ButtonSelect to "hold \ButtonHome"
23:40:37saratogaits been years since I used latex
23:41:11saratogahmm maybe noting that you can use hold in the getting started - clip+ section would be a good idea as well
23:41:43 Quit evilnick (Ping timeout: 240 seconds)
23:43:42AlexPsaratoga: Long is used in the manual
23:43:52AlexPsaratoga: As in "Long \ButtonHome"
23:43:59saratoga \opt{SANSA_CLIP_PAD}{Long \ButtonHome}
23:44:10AlexPThat would be fine
23:46:17CIA-5New commit by 03saratoga (r26396): Change the software hold button combo from select + home to just holding home. This is both easier to do on such a small player and more consistent ...
23:49:42 Quit Topy44 (Quit: No Ping reply in 180 seconds.)
23:54:16CIA-5New commit by 03saratoga (r26397): Commit FS #10887 by me. Change the check for software poweroff to be that the power button is held, not that the power button is held AND that no ...
23:54:25CIA-5New commit by 03gevaerts (r26398): No need to break after a return
23:58:42 Join Topy44 [0] (

Previous day | Next day