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 2010-05-15

00:00:03funmandfkt: do you think rockboxed clip+ has lower volume than clipv1 ?
00:00:11Blue_Dudegevaerts: Just for my own edification, is this sort of problem in Cygwin itself, or a consequence of compiling with MinGW, or what? I'd like to know better what's going on behind the scenes.
00:00:53dfktfunman, lower than rockboxed clipv1? i will check
00:01:08CIA-5New commit by 03gevaerts (r26030): Move include/sys along with libc/, so hosted (sim/RaaA) builds use the proper files for their OS
00:01:39gevaertsBlue_Dude: no, just the includes being messed up
00:01:42CIA-5New commit by 03funman (r26031): fuzev2: prevent button light flickering when accessing µSD
00:01:51gevaertsThat's not the only problem unfortunately, there seems to be more
00:02:10funmanFlynDice: fyi, we can modify B5 just after MCI_COMMAND has been written (it's fast enough to not see the flickering)
00:03:42Blue_Dudegevaerts: Just curious why it showed up for my build but not on the build server.
00:04:10 Join polobricolo [0] (
00:05:15gevaertsBlue_Dude: is "system headers are messy and unpredictable" good enough? ;)
00:05:30 Join Schmogel [0] (
00:05:57bertrikI think no-one is running a build client on cygwin
00:06:29polobricoloWhat are bmp2rb_mono and bmp2rb_native in the configure file ?
00:06:31CIA-5New commit by 03Buschel (r26032): Fix FS #11261 (slow seek forward in mpc files). Since the introduction of the sv8 library mpc did not use efficient buffering when seeking forward. ...
00:12:33Blue_Dudegevaerts: that's as good an explanation as any. :)
00:12:59CIA-5New commit by 03funman (r26033): cast ssize_t to long
00:15:12 Join Kitr88 [0] (
00:16:41CIA-5New commit by 03funman (r26034): another cast ssize_t -> long
00:17:09*gevaerts thinks that more coordination is needed :\
00:17:40 Quit Kitar|st (Ping timeout: 248 seconds)
00:18:06gevaertsfunman: please warn if you're going to fix things that someone else has just caused. Changing a file, test-compiling, and then getting commit errors is no fun
00:18:25gevaertsespecially if your fix is wrong
00:18:38funmanis it?
00:19:29gevaertswell, it's not the best fix anyway
00:19:31 Quit Kitr88 (Ping timeout: 240 seconds)
00:19:36 Join Kitar|st [0] (
00:19:49CIA-5New commit by 03gevaerts (r26035): Fix various size_t related warnings and errors
00:20:00gevaertsprintf has modifiers for size_t
00:20:14funmani checked rockbox' format.c for %z but it's not there
00:21:15gevaertsThese are all in DEBUGF(), so it might not matter
00:21:26gevaertsThe problem is that size_t can have various sizes
00:21:44Blue_DudeThought size_t was an unsigned int.
00:21:46funmantrue, but here it was used together with (unsigned) longs
00:22:03gevaertsBlue_Dude: no. It might be, but it's not always
00:22:06funmanBlue_Dude: not on 64bits i think where int is 4 bits
00:22:09funman4 bytes*
00:22:27Blue_DudeInt seems to be 4 bytes anyway.
00:22:28funmangevaerts: isn't DEBUGF usable over (usb) serial ?
00:22:36Blue_DudeSame size as long.
00:22:38Blue_DudeI think.
00:23:17gevaertsBlue_Dude: exactly. That's why they added the %z thing :)
00:24:20Blue_DudeYeah. All things considered I like fixed size types for bit twiddling.
00:24:40CIA-5New commit by 03gevaerts (r26036): Hopefully the last warning
00:24:55funmangevaerts: debugf() use rockbox formats with HAVE_GDB_API
00:25:12funmanor SH debug (but who uses SH)
00:25:44gevaertsfunman: hm, right. There are other %zs already though (but not since very long)
00:26:03gevaertsI'd actually say that implementing %z is the right fix if this is a real problem
00:26:35gevaertsNow the rest of the mingw errors...
00:27:15Blue_DudeWarning in codecs/flac.c line 505...
00:27:58Blue_DudeBut it compiles so far.
00:29:49Blue_DudeMore warnings, all the same.
00:30:45Blue_DudeYeah, they're LOGF format string mismatches.
00:31:32 Join halmi [0] (
00:32:15Buschelbuilding under cygwin still quits with error
00:32:34Buschel(ps sim only)
00:32:41gevaertsfunman: exactly, not very big, obviously correct, and solves a lot of problems :)
00:33:17gevaertsAfter that someone needs to go over all the casts to check if they're really needed of course
00:33:23 Join joeyg [0] (
00:33:35funmani know someone who would really like to do that
00:34:03gevaerts"warning: format ‘%zd’ expects type ‘signed size_t’, but argument 6 has type ‘ssize_t’
00:34:15gevaertsYou just can't win!
00:34:24joeyghaha, how about ptrdiff_t?
00:34:50*gevaerts ignores that warning!
00:35:17Blue_DudeAgh. Back in a while. Maybe we oughta just put a cast on all of them and be done with it.
00:35:32Blue_DudeJust a thought. Lots of work though.
00:35:50 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
00:37:20 Quit Buschel (Ping timeout: 260 seconds)
00:37:41 Quit funman (Quit: free(random());)
00:38:29 Quit cdb (Ping timeout: 276 seconds)
00:39:17 Quit merbanan (Ping timeout: 276 seconds)
00:45:37 Part toffe82
00:51:38 Quit petur (Quit: router upgrade)
00:54:19 Quit stripwax (Ping timeout: 240 seconds)
00:54:38 Join Forsaken_Boy [0] (~chatzilla@
00:57:33 Quit bertrik (Remote host closed the connection)
00:57:54CIA-5New commit by 03gevaerts (r26037): Make the sim buildable with mingw again
01:01:30 Join petur [0] (~petur@rockbox/developer/petur)
01:01:43gevaertsBlue_Dude: can you check if it also works for you?
01:04:55 Quit pamaury (Ping timeout: 252 seconds)
01:06:15 Join solexx [0] (
01:09:04 Quit solexx_ (Ping timeout: 260 seconds)
01:23:15 Quit ender` (Quit: Inside every cynical person, there is a disappointed idealist.-- George Carlin)
01:26:29 Quit petur (Quit: Zzzzz)
01:26:36 Join Strife89 [0] (
01:31:29 Quit flydutch (Quit: /* empty */)
01:37:26 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
01:38:30 Join bluefoxx [0] (
01:46:05 Quit jd (Read error: Connection reset by peer)
01:47:31 Quit Schmogel (Ping timeout: 240 seconds)
01:50:31***Saving seen data "./dancer.seen"
02:00:28 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
02:01:56 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
02:03:22 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
02:05:18 Quit DerPapst (Quit: Leaving.)
02:06:39 Quit kisak (Ping timeout: 245 seconds)
02:11:19 Join kisak [0] (
02:12:14 Quit halmi (Ping timeout: 276 seconds)
02:18:16 Join ollebe [0] (
02:19:39 Join Rob2223 [0] (
02:23:00 Quit Rob2222 (Ping timeout: 246 seconds)
02:29:25 Join merbanan [0] (
02:33:21 Quit n1s (Quit: Lmnar)
02:37:21 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
02:51:47 Quit merbanan (Ping timeout: 248 seconds)
02:56:22 Join huelk_ [0] (
02:59:56 Part Boldfilter
03:00:07 Quit huelk (Ping timeout: 265 seconds)
03:04:04 Join ischeriad [0] (
03:06:38 Quit ischeria1 (Ping timeout: 240 seconds)
03:08:37 Quit Forsaken_Boy (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
03:12:06kisakoh ... topic / conflict ... is it Ghent or Gent?
03:18:12 Join intrados [0] (
03:18:26 Quit togetic (Quit: WeeChat 0.3.0)
03:27:18 Quit powell14ski (Ping timeout: 240 seconds)
03:33:22 Join powell14ski_ [0] (
03:35:09 Quit GeekShadow (Quit: The cake is a lie !)
03:39:52 Join Blue_Dude [0] (~chatzilla@rockbox/developer/Blue-Dude)
03:40:47Blue_Dudegevaerts: Sorry to vanish like that. Anyway, it compiles, but I do get warnings. I have logf enabled for my sim build, so that's why I'm getting those warnings.
03:44:15 Join TillW [0] (
03:44:36 Quit TillW (Client Quit)
03:45:00 Join CGL [0] (~CGL@
03:50:32***Saving seen data "./dancer.seen"
03:54:22 Quit adnyxo (Quit: Leaving)
03:57:33CIA-5New commit by 03Blue_Dude (r26038): Fix logf lines in codecs (type mismatches)
03:59:29 Join ischeria1 [0] (
04:03:26 Join Forsaken_Boy [0] (~chatzilla@
04:03:30 Quit ischeriad (Ping timeout: 276 seconds)
04:04:35 Join anewuser [0] (anewuser@unaffiliated/anewuser)
04:09:21 Quit Forsaken_Boy (Read error: Connection reset by peer)
04:12:06 Quit TheSeven (Ping timeout: 240 seconds)
04:15:32 Join Boldfilter [0] (
04:16:21 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
04:17:44Blue_DudeFound other problems in some plugins. warning: signed size_t format, ssize_t arg. Well.
04:18:05Blue_DudeThat won't be fixed with a cast.
04:20:01Blue_DudeUnless, casting a ssize_t to signed size_t makes sense.
04:20:18 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:20:18 Quit pixelma (Disconnected by services)
04:20:23 Quit amiconn (Disconnected by services)
04:20:24 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:20:33 Join stormdragon2976 [0] (~stormdrag@
04:20:38 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:20:44 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:23:19stormdragon2976I wrote a bash script in Linux to generate .talk files. I found out only after finishing the script that the files aren't just MP3s with modified extentions. Does anyone know a command line way to generate the files?
04:24:10stormdragon2976Preferably using sox?
04:26:00Blue_DudeNope, a cast to "signed size_t" doesn't work at all.
04:29:16Blue_DudeAll the lines I can't fix with a cast use "%zd" in them. I can change that but it won't fix the formatting problem in the library.
04:33:30 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
04:36:03saratogastormdragon2976: on most devices the talk files are speex, not mp3
04:36:15saratogathe makefile should be able to generate them for you, or you can use rbutil to do it
04:41:13 Join zak1111 [0] (
04:41:36 Part zak1111
04:42:28stormdragon2976The Utility doesn't work with Orca (the screen reader for Gnome) It's written in qt4 I think, and Orca doesn't work with QT apps.
04:42:53stormdragon2976Unless there's a commandline way to do it?
04:44:58saratogayes, the makefile
04:45:54stormdragon2976How would I use it?
04:47:44saratogasame way as making a normal rockbox build, except select "voice" as the build type in the configure script instead of "normal"
04:48:51stormdragon2976Lol I am behind in the learning process. I bought my player with Rockbox pre-installed.
04:49:46 Join dys` [0] (
04:52:05 Quit dys (Ping timeout: 276 seconds)
04:54:47 Join zak1111 [0] (
04:55:35 Part zak1111
04:55:46 Quit Barahir (Ping timeout: 240 seconds)
04:57:39 Join Barahir [0] (
04:57:57saratogastormdragon2976: if you're just running stock rockbox, then maybe its easiest to just use our voice files
04:58:53stormdragon2976Maybe so, but I would like to be able to have the song and folder names read instead of spelled. I am only one step away from making that happen.
04:59:11stormdragon2976If only sox supported speex files lol
04:59:16 Part Boldfilter
05:03:56 Join togetic [0] (
05:03:58 Quit togetic (Changing host)
05:03:58 Join togetic [0] (~togetic@unaffiliated/ibuffy)
05:05:39Blue_DudeMore logf drama: /rockbox/firmware/logf.c:216: warning: implicit declaration of function `vuprintf'
05:16:22stormdragon2976I think I found the answer. There is a speex encoder in the Ubuntu repository.
05:17:14CIA-5New commit by 03Blue_Dude (r26039): Fix include problem
05:20:15saratogaBlue_Dude: have you seen FS #11270 - PictureFlow - WPS Integration patch
05:20:21saratogait looks vaguely related to hotkeys
05:20:37Blue_DudeI haven't. I don't even know what pictureflow is.
05:20:57saratogaits a plugin that lets you browse albums via their album art instead of pure text menus
05:21:08saratogaits semi-integrated into the actual gui
05:24:46saratogai only mention it to you because it talks about hotkeys
05:25:40Blue_DudeI grabbed the word hotkey out of thin air.
05:25:48Blue_DudeDoes it mean the same thing there?
05:34:32CIA-5New commit by 03Blue_Dude (r26040): Eliminate %zd tag in printf format strings, replace them with %ld. The %z formatter kept generating type mismatch warnings.
05:42:41CIA-5New commit by 03Blue_Dude (r26041): Fix yellow: missed a cast
05:47:08CIA-5New commit by 03Blue_Dude (r26042): vuprintf does not belong in stdio.h, causes problems with other versions of stdio.h
05:48:15Blue_DudeI feel like the sim police tonight.
05:48:20 Join rasher [0] (~rasher@rockbox/developer/rasher)
05:48:43Blue_DudeThat should be the last compile warning though.
05:50:34***Saving seen data "./dancer.seen"
05:51:23 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
05:52:05 Quit ollebe (Quit: Leaving)
05:52:15 Join RankFrank [0] (
06:01:24 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
06:01:45 Quit kramer3d (Ping timeout: 240 seconds)
06:05:04 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
06:08:59 Join Forsaken_Boy [0] (~chatzilla@
06:11:12 Quit kenguest (Read error: Connection reset by peer)
06:14:08 Join umrain [0] (
06:17:00 Nick umrain is now known as umrain_ (
06:17:25 Nick umrain_ is now known as umrain (
06:17:49 Join kenguest [0] (
06:21:18 Quit leavittx (Remote host closed the connection)
06:29:15stormdragon2976Do the speex files have to be a specific bitrate?
06:30:30 Quit arbingordon (Quit: `)
06:31:20 Join zak_ [0] (
06:31:32 Quit zak_ (Remote host closed the connection)
06:33:59 Quit kramer3d (Quit: This computer has gone to sleep)
06:36:06saratogastormdragon2976: see this link:
06:36:22saratogathe tools take care of things like that automatically
06:37:13saratogahmm though thats Windows only if that matters
06:42:22stormdragon2976I don't have access to a Win box.
06:47:19 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
06:52:03 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
06:59:26CIA-5New commit by 03jethead71 (r26043): wm8978: Clean out silly macros. Use 'POS' convention instead for shifted bitfields. Additionally, use volume update latching for all volume settings.
07:00:14ranmaAny usb experts here?
07:00:37 Join arbingordon [0] (~w@unaffiliated/arbingordon)
07:00:53ranmaI'm getting the config request from the host, pass that on to the usb_core, and try to send the response back.
07:01:26ranmaI can see that the response is received by the host, but something is wrong on a low level.
07:01:58ranmaSomehow the receive seems to timeout (5 seconds), but still gets the data correctly.
07:02:16ranmaBut since it timed out or so the host ignores the data it got.
07:02:36saratogathis is on AMS?
07:02:51 Quit arbingordon (Client Quit)
07:03:57saratogaah didn't realize you were working no that :)
07:04:53ranmaI've been working on that since thursday :)
07:07:17 Quit evilnick (Read error: Connection reset by peer)
07:07:18saratogawhats actually involved in porting a new USB controller to the current stack?
07:07:43 Join evilnick [0] (
07:07:44saratogai guess at the high level some kind of send and recv functions, but I'm guessing USB is a little more complicated then that
07:08:19ranmaWell most of the low-level stuff is handled by the controller.
07:08:28ranmaThe main problem is figuring out how the controller works.
07:09:39ranmaReceiving packets was the easy part. Getting it to send was a bit more tricky and obviously something is still going wrong.
07:19:57 Quit BHSPitMonkey (Quit: Ex-Chat)
07:20:54 Join Horschti [0] (~Horscht2@xbmc/user/horscht)
07:21:21 Quit Horscht (Ping timeout: 268 seconds)
07:21:37 Quit CGL (Ping timeout: 240 seconds)
07:25:21 Quit Xerion (Quit: )
07:29:36 Join Schmogel [0] (
07:29:50 Nick dys` is now known as dys (
07:36:06 Join CGL [0] (~CGL@
07:40:28 Join mikroflops_ [0] (
07:44:59 Quit mikroflops (Ping timeout: 276 seconds)
07:46:25 Quit Forsaken_Boy (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
07:50:00 Quit huelk_ (Ping timeout: 240 seconds)
07:50:38***Saving seen data "./dancer.seen"
08:00:31 Join Highlander [0] (
08:05:30 Join Buschel [0] (
08:14:41stormdragon2976lol my computer recognizes and can play these .talk files but they still aren't working on the MP3 Player.
08:15:22stormdragon2976I am using a speex encoder and still no luck.
08:19:04saratogastormdragon2976: if your computer can play them, then they're probably not in the right format
08:19:23saratogai don't think anything outside of rockbox will decode rockbox formated speex files
08:19:34saratogathough i'm not certain how this works
08:19:49*S_a_i_n_t suggests using VoiceBox in a VM
08:20:00S_a_i_n_t(if you're not on a WinSys)
08:21:03saratoga"On Archos players, .talk files are normal MP3 files. On other players, they are in a Rockbox-specific Speex format produced by the rbspeexenc utility, which is distributed with Rockbox."
08:21:07 Quit joeyg (Ping timeout: 252 seconds)
08:22:28stormdragon2976ahh, I'll go find that.
08:22:35 Join joeyg [0] (
08:27:19 Quit saratoga (Quit: Page closed)
08:33:25 Quit Buschel (Ping timeout: 260 seconds)
08:33:36 Quit Schmogel (Ping timeout: 240 seconds)
08:36:41 Join zak1111 [0] (
08:39:26 Quit CGL (Remote host closed the connection)
08:39:36 Part stormdragon2976 ("Leaving.")
08:41:33 Join ssorgatem [0] (~ssorgatem@
08:49:33 Quit joeyg (Quit: lions and tigers and bears, oh my!)
08:54:50 Join flydutch [0] (
08:55:41 Quit Strife89 (Quit: Bed.)
09:10:29 Quit RankFrank (Ping timeout: 245 seconds)
09:13:26 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
09:16:41 Join RankFrank [0] (
09:29:54pixelmaor he could just use the Rockbox Utility... (regarding the .talk files generation)
09:36:25pixelmaok, just read his problem using the Utility because of the Orca screenreader
09:36:41pixelma+about and +sorry
09:37:46 Part zak1111 ("Leaving")
09:50:40***Saving seen data "./dancer.seen"
09:51:26pixelmaI recently had the idea to make the "Now playing" in cabbiev2 a text string, adding it to the lang files so it is possible to localise it (and use something else for e.g. the radio screen etc.). The problem is that it could only be a smaller font which probably won't have a lot of non-latin characters and can't have. Would it be acceptable to add a comment for the translators "Leave it untranslated if your string can't be displayed with small-font
09:54:38pixelmathen it will stay as it is for those in the WPS and still will be more flexible for use in the radio screen, or the menus if someone attempts an SBS.
09:57:50 Join webguest79 [0] (
09:59:55 Quit webguest79 (Client Quit)
10:05:52 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
10:06:41bieberJdGordon: In your new WPS format, how many tags are there with optional parameters at the beginning of the list?
10:07:03JdGordon2 or 3?
10:07:15JdGordonwe can make it 0 if you would prefer?
10:07:31bieberThat would definitely make things a lot easier to parse
10:07:56JdGordonok, so totally forget about optionals for now...
10:08:37bieberThey're easy enough if they come at the end of the list, it just gets really hard to discern what's what if they can come at the beginning or the end
10:09:25JdGordonfair enough
10:10:33bieberSo for the tag specifications, I'll leave out the ?, and just keep the | to mark the beginning of optionals, along with prefixed numbers to specify groups of the same type
10:11:10bieberI'll add a patch with what I've got so far to FS before I go to sleep
10:11:17 Join Orion7 [0] (
10:11:59JdGordonhow far have yoiu got?
10:13:23bieberRight now, it can reliably assemble a parse tree from plaintext, comments, and newlines, and I've got as far as retrieving the param list for a tag name, but I've been stalled trying to figure out how I'd do those optional params :P
10:14:44bieberAnd while the code to parse them isn't there yet, it does recognize when there are sublines coming, so that will be straightforward once I've got the tag function finished
10:14:46JdGordondepends how you pass the params in... I was thinking just split up the param list and determine if the optional one is there based on the number of params listed
10:15:15bieberThat works for simple cases, but the problem comes when you have, say, one optional at the beginning and two optionals at the end
10:16:09bieberAnd then if the end optionals don't have to be grouped together, you have to analyze the types of the parameters to figure out which one is going to be the optional
10:16:22JdGordonleave that to the tag handler
10:17:15bieberSo you don't want me to check types at all?
10:17:56JdGordonits up to you, of course I'm happy to say that optionals must be at the end also
10:18:52bieberLets just go with optionals at the end, then. That'll give some more useful debug output from the parser
10:25:42JdGordonjhMikeS: you have to be in the recording screen to record?
10:28:46bieberAdded a comment with new patch to FS #11265, I'm done for the day
10:30:25bieberDo comment if you spot any bugs, or have comments on architecture
10:30:38JdGordonill have a look if i get a chnace toinght
10:31:38bieberDon't worry too much about it if you don't, hopefully I'll have a lot more to show in the next day or two
10:35:51 Quit Orion7 (Ping timeout: 252 seconds)
10:37:26 Join ender` [0] (
10:47:39 Join funman [0] (~fun@rockbox/developer/funman)
10:53:46jhMikeSJdGordon: yes
10:54:16JdGordonah ok
10:54:20JdGordonI didnt know that
10:54:50pixelmaexcept on hwcodec radio targets ;)
10:54:52jhMikeSthough I do have some thoughts on letting it run independent of UI
10:57:20funmanare menus voiced when listening to FM on c200v1/e200v1 ?
10:57:59S_a_i_n_tcan anyone confirm soe things I'm seeing in "Viewer" for me please? Scroll down to th e bottom of a file being viewed, "show footer" in viewer options needs to be set to "page num" (I'm using "rockbox-info.txt" to reproduce), now scroll back up to the top of the file and note the page number is not 1.
10:58:41S_a_i_n_tit takes quite a bit more scrolling to reach "page 1" on my Nanos despite it actually showing the top of the page.
10:59:09pixelmafunman: yes
11:00:33pixelmaat least in a slightly older build. I don't know if FMS support could have changed something there but don't expect any problems there
11:02:24*JdGordon glares at pixelma for that accusation
11:04:21pixelmaI said I *don't* expect problems there but it's quite a big change and I haven't tried with my c200 so there still might be a small possibilty
11:04:22funmanvoices don't work on clipv1 at all
11:04:58pixelmafunman: did you check voice file size and free RAM?
11:05:28 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:05:30funmanno, i think i don't need to check to know that the problem is RAM ;)
11:06:15pixelmawell, the Clipv1 has 2MB of RAM, no?
11:06:16funmanFM + voicing works fine on fuzev1
11:06:56AlexPamiconn: ape insane on gigabeat S is now 112.62% realtime
11:07:11AlexP468.8 MHz
11:07:35pixelmafunman: how much of it is free then with current Rockbox?
11:07:47 Join nls [0] (
11:08:12AlexPheh, I see amiconn has just done the same tests :)
11:08:33AlexPor is the process of so doing given the wiki page editing :)
11:08:49funmanaudio buffer is 473040 bytes (r26027)
11:11:23JdGordonthat seems way too small
11:11:44funmanno clip, the fuze works fine with 8mb
11:11:59JdGordonok *phew*
11:12:20amiconnAlexP: Yes. I'm also making the CPU information a bit more precise
11:13:55 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
11:15:26CIA-5New commit by 03uchida (r26044): ID3 tags parser separates from metadata/mp3.c
11:18:46pixelmalook like a bigger change (at least something with the potential to cause problems at different place) - out of the blue
11:19:32CIA-5New commit by 03funman (r26045): as3543: output sum of line & dac to have voicing work while using FM
11:20:04nlsfunman: maybe you could look into the mechanism used for the archos ondios for voice loading, they don't load the entire voicefile to memor, which all other targets do (which obviously doesn't work if the file is too big)
11:20:32pixelmaumm, the voice file needs to fit on the Ondios too
11:21:17pixelmacompletely. The mechanism is there to make loading and playing the clip faster IIRC
11:21:54funman941kB is obviously too big
11:23:14nlspixelma: ah, yes, just saw that, hmm
11:23:56nlsbut loading the clips on demand should be fst enough on a newer flash target
11:24:40CIA-5New commit by 03uchida (r26046): fix yellow
11:26:38JdGordondoes voice work at all on the clips?
11:27:00nlsJdGordon: no, loading the voicefiel fails since it's too big
11:32:34 Join stoffel [0] (
11:35:18AlexPamiconn: Isn't the % needed to play realtime dependent on the cpu? i.e. even though the beast is 112% realtime is has ~ 65 MHz left, so should easily be realtime
11:35:42AlexPI'd just like the line to be entirely green :)
11:35:55amiconnThere is a problem anyway - somehow the anchor syntax for portalplayer doesn't want to work
11:36:42 Join dfkt [0] (dfkt@unaffiliated/dfkt)
11:36:52amiconnAh, wrong order
11:40:25JdGordonso here is a fun suggestion... can we make buffering allow you to request a handle which is a set size, guarenteed to fit (so it dumps handles if it needs to) and static so it doesnt move?
11:40:34funmanpixelma: n1s: the voice file is only loaded as needed because of slow access but it still reserves the full room for it
11:40:56 Quit nls (Quit: Lmnar)
11:40:59amiconnJdGordon: That would be malloc, so no
11:41:21JdGordon:p worth a shot anyway
11:41:36funmanload_voicefile() checks if it fits but not talk_init()
11:41:48pixelmafunman: that's what I meant
11:41:55n1sfunman: yes, i saw that, but maybe it can be adapted
11:42:42n1si think loading on demand should be fast enough on the clip, so we need only a buffer to hold one clip
11:44:16ssorgatemhi all
11:44:35ssorgatemI'm unable to use festival with rockboxutility
11:44:53ssorgatemit does work with "make voice", though
11:45:13amiconnn1s, funman: On the Ondios this would probably be a good idea as well, as it would allow better quality voices
11:45:48amiconnThe buffer should hold a bit more than just one clip though. It would allow loading the next clip while the current one is still playing
11:46:11*amiconn thinks an lru cache like for the font might be useful
11:47:02n1samiconn: yeah, an lru could be good
11:47:06n1seh mru
11:47:47 Join merbanan [0] (
11:48:00pixelmafunman: I knew swcodec needs quite a lot more RAM for itself but I am surprised that it is so much more
11:48:05pixelmaa bit
11:48:09bertrikoh that sounds like coming up with a solution before determining that there is a problem in the first place
11:48:19n1sah, no the cache algos' name use inversed logic
11:49:15n1sfunman: is the pcmbuffer smaller on the clip?
11:49:17pixelmabertrik: what are you referring to?
11:49:33bertrikbuffering voice clips
11:50:03funmann1s: it is 1s (176k)
11:50:06n1sbertrik: there is a proble −−- it doesn't work
11:50:12 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
11:50:24pixelmabut which problem isn't determined?
11:50:25n1sfunman: ok
11:50:36funmanshould i commit ?
11:50:41***Saving seen data "./dancer.seen"
11:50:57bertrikI'm questioning the need for a cache to load the next clip while the current one is still playing, suggested by amiconn
11:51:07funmanat least playback will continue to work when voicing is enabled & there is a .voice file
11:52:06pixelmabertrik: that will make it quicker
11:52:48amiconnbertrik: Caching would reduce the number of disk accesses, i.e. it should help saving battery
11:53:02amiconnIt also helps keeping the codec running
11:54:14polobricoloHey did anyone try to port rockbox to a ARM Cortex target ?
11:54:37polobricolo(eg an STM32 cpu)
11:54:52bertrikok, I see your point, but we have no idea if those things are really a problem right now, i.e. it might already be quick enough, battery savings may be negligible, etc.
11:56:13n1spolobricolo: don't think so, but as long as it's not one of the thumb-only cores it shouldn't be too much trouble
11:58:23 Join DerPapst [0] (
11:58:25pamauryAny LCD expert here ?
11:59:07bertrikseveral, I guess :)
11:59:12pixelmageneral rule: just ask
12:00:22polobricolon1s: ok in fact i'm trying to port rockbox to a stm32 evaluation board (it has lcd, speaker/jack, microphone, microsd card reader (no internal memory though)
12:01:39pamaurycool, how are lcd screens generally controlled, is spi a usually way or not ?
12:02:40polobricoloso all the threading part would still work
12:03:27bertrikmost of the ones I've looked at had a parallel bus, spi is serial so it needs a very high bit rate to get reasonable transfer speeds
12:04:09pamauryfrom the physical point of view, I can say that the lcd (of unknown type) has 96x32 pixels and is connected to the board by 8 pins
12:04:11n1spolobricolo: most of rockbox is independent of the port, you basically need setup code for your cpu, and drivers for your hardware (and some keymaps)
12:04:31funmandoesn't the ipod mini2g use a serial bus?
12:04:32pamaury(except if there are hidden one but I doubt it)
12:05:39bertrikif the display is small (few pixels and/or low color depth), then a serial interface like spi is possible
12:05:49CIA-5New commit by 03funman (r26047): talk_init() : don't try to load the voice file if it won't fit in memory ...
12:06:48bertrikpamaury, do you have access to the schematic, maybe you can tell something from the names of the signals
12:07:03pamauryIs there a kind of standard or is every lcd completely different from the others ?
12:11:10 Join dfkt_ [0] (dfkt@unaffiliated/dfkt)
12:11:23bertrikI think there's no formal standard, but a lot of interfaces look quite similar anyway, like there's usually some kind of clock signal, R/W signal and a signal to select between command and data mode
12:11:35 Quit pamaury (Ping timeout: 252 seconds)
12:13:47 Quit dfkt (Disconnected by services)
12:13:50 Nick dfkt_ is now known as dfkt (dfkt@unaffiliated/dfkt)
12:20:09CIA-5New commit by 03uchida (r26048): mp3: when ID3 tags are not found, search APE tags
12:21:29*pixelma wonders why uchida (ucchan?) isn't around
12:21:38 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
12:22:17pamauryok thanks
12:23:26 Join JohannesSM64 [0] (
12:23:38AlexPWhen did that get it?
12:23:47AlexPI thought there was a long argument about that
12:23:58*pamaury thought there was a kind of nodo AlexP about non-id3 tags
12:24:09pamaury*s/AlexP/about :)
12:24:21AlexPpamaury: Yes, something like that
12:24:42AlexPI really really really wish (for the 5th or 6th time now) that uchida was around when committing
12:24:43pamauryDid uchida asked before comitting ?
12:24:52pamauryHe is never here anyway
12:25:19AlexPI know, but I really think it ought to be a "rule" that when committing you are on IRC
12:25:28pixelmathat too, but I also don't like that uchida made some big changes without being around here and I haven't seen discussion about it. Also the split of the ID3 parser before - I'm curious how much it was tested
12:25:36AlexPIf you are at a computer to commit then you should be on IRC
12:25:46*AlexP adds it to the devcon page
12:25:55pixelmaalmost +3k on hwcodec targets :(
12:26:16Unhelpfulperhaps it ought to be reverted, then...
12:26:25AlexPAt least until the discussion is had
12:26:59pamauryuchida often do big commits without asking
12:27:01*Unhelpful has taken reading of non-native tag formats to be a NODO for as long as he's been involved, and imagines it goes back rather farther than that
12:27:03pixelmaI guess that's because APE tags didn't have to be parsed before at all
12:27:32AlexPpamaury: Yes, and is never here
12:27:41AlexPHe has been asked to be multiple times before
12:27:56 Join einhirn [0] (
12:27:57 Join JdGordon_ [0] (~jd@rockbox/developer/JdGordon)
12:28:07pixelmaI saw him here before committing TTA but it isn't often
12:28:50pixelmaI mean that he is around
12:30:37JdGordon_Someone want to revert that?
12:31:08JdGordon_Ape in mp3 has been nodo for a long time
12:31:36AlexPGo for it
12:32:04JdGordon_I'm afk
12:32:15AlexPNo you're not!
12:32:27AlexPUnless it is ghost JdGordon_ typing :)
12:32:30funmanwell ask him to revert rather
12:32:45AlexPI've sent a mail to the dev list, people can chime in :)
12:32:51pixelmahaving to go through the ml
12:33:46funmanreverting commit of someone else is impolite IMO, i prefer asking (or be asked if i did the commit)
12:34:20AlexPfunman: I agree, unless they aren't around
12:34:29pixelmahe didn't ask before either and isn't around now
12:34:49AlexPI've added it to the devcon talking points list - that you should be in IRC when you commit stuff
12:34:59AlexPOr rather, whether we think people should be
12:36:49 Join b0hoon [0] (
12:37:27funmani think some people don't like irc so for me the mailing list is sufficient (discussions can be close to real-time)
12:38:19JdGordon_Its more about discussion than being online
12:38:24AlexPyeah, true
12:39:38*AlexP replies to his own e-mail
12:39:40 Quit stoffel (Remote host closed the connection)
12:40:13funmani also think that it's easier to discuss on the mailing list for uchida if he's using a translator (all content is on topic in a given thread)
12:40:33funmanirc talks often need a bit of filtering
12:40:40AlexPthis is also true
12:41:15funmanperhaps we should just ask him why he doesn't find irc more convenient to use
12:41:41*funman uses irc by laziness and wait for uchida to pop up
12:47:06JdGordonas for that commit.. it should only be swcodec
12:47:07ThomasAHfunman: talk work now while playing radio, but english.voice somehow does not
12:47:37JdGordonI dont actually have any problem with APE in mp3 if it is efectivly free (which it is on swcodec)
12:48:08funmanThomasAH: perhaps you need an up to date file? i used the english.voice from
12:48:13 Join dfkt_ [0] (dfkt@unaffiliated/dfkt)
12:48:30ThomasAHfunman: I always rebuild from svn to be sure it matches
12:48:53funmanso talking directories work but not menus?
12:49:15pixelmayou always build a voice file too?
12:49:24ThomasAHfunman: talking directories with .talk clips work, spelling directories and talked menus don't
12:49:54ThomasAHpixelma: yes, this way I don't have to check if it is really needed :)
12:50:17amiconnpamaury: There are several different LCD interfaces, all of them are used in some targets: serial (SPI), 8/9 bit parallel, and 16/18 bit parallel
12:50:58amiconnExamples for serial are the archoses, the iriver and iaudio remotes, and the mini G2 (but the latter uses the PP built-in bridge to do the conversion)
12:51:30pamaurydo they have something in command except spi ?
12:52:16 Quit dfkt (Ping timeout: 265 seconds)
12:53:01amiconnThe command set/ registers depend on the controller. The method is similar for all of them, but the register numbers and bit patterns are usually different
12:53:20*amiconn thinks r26048 should be reverted
12:54:17ThomasAHpixelma: (I used to not build it in the past, but when funman enabled additional features in the clip+, the menu was always off by one)
12:54:47funmanThomasAH: i didn't use .talk files, only voice
12:54:51ThomasAHpixelma: and the OF database refresh takes much longer than building the voice file many times
12:55:02pamauryok thanks
12:55:13ThomasAHfunman: I'll double check now
12:55:26funmanamiconn: we just discussed this 20 minutes ago :)
12:55:35amiconnyes, saw that
12:57:38 Nick fxb__ is now known as fxb (
12:58:29funmani am not sure i want to make voicing work on clipv1/c200v2 : it would shrink the playback buffer, and then i'd have to fix playback bug. i think i'm not up the task and don't want to drive crazy on this
13:02:46ThomasAHok, clean svn r26037, rebuild and english.voice, installed fixed.cfg and fmpresets, nothing else ... now waiting for the OF media refresh
13:02:55ThomasAHclean svn r26047
13:03:06*ThomasAH should use more copy&paste :)
13:03:55 Quit JdGordon_ (Quit: Bye)
13:08:19ThomasAHfunman: still no voice menus
13:08:55funmanwhat if you disable .talk?
13:09:23pixelmaThomasAH: and the setting is enabled? Might be silly but worth a check ;)
13:09:41ThomasAHfunman: does not help
13:09:56ThomasAHpixelma: yes, I tripple checked that :)
13:10:16pixelmaand you use english in the language settings too?
13:10:46ThomasAHpixelma: yes ... and the compiled english.voice file is binary identical to some previous versions, too
13:11:19ThomasAHswitching to a different language and back does not help
13:13:32ThomasAHbooting an older rockbox.sansa via file browser brings voices back
13:13:44ThomasAH(with everything else identical)
13:14:17funmanhmm yes i have no voice anymore too
13:14:27ThomasAHfunman: somehow this is good news :)
13:14:40funmanhow older is the working rockbox.sansa ?
13:14:49ThomasAHfunman: yesterday
13:15:21kisakthat's a good, small timeframe
13:15:46funmanOTOH, voice worked for me 2 hours ago ;)
13:16:41*funman slaps himself repeatedly
13:16:46 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
13:17:20funmanmy supposed clipv1 fix in r26047 is incorrect
13:18:38 Quit n1s (Ping timeout: 252 seconds)
13:19:31CIA-5New commit by 03funman (r26049): fix inverted logic in r26047: voice works again
13:21:01 Join stoffel [0] (
13:21:10*b0hoon says hi.
13:21:16funmanThomasAH: everything's alright now?
13:21:20funmanhello b0hoon
13:22:02b0hooni will promote pb vibe to the stable ports today if nobody still objects.
13:23:09funmanb0hoon: isn't it better to wait for next release?
13:24:12b0hoonfunman: why it would be better?
13:24:35funmanb0hoon: there would be a release build
13:24:44rasherLlorean: I really the idea of having the font/language loader check compatibility
13:24:57AlexPamiconn: So do others, now people need to reply to the list saying so :)
13:25:02 Join marc2003 [0] (
13:26:10funmanany plan on a release? we are now more than 3 months after 3.5
13:26:13b0hoonfunman: yes, but if i'll do it now it will be available in the next release anyway
13:27:00Lloreanrasher: It would eliminate one of the easiest remaining ways to get an indecipherable UI, at least.
13:27:03ThomasAHfunman: I'll check (at least the patch looks promising :))
13:27:20marc2003hey funman, this is the output after making that you suggested on ABI
13:27:22marc2003sd0: 15994652 + 1 > 15993856
13:27:33rasherLlorean: I almost have a patch ready for genlang to add a string simply consisting of all characters used in that language
13:27:46rasher(which is the easist part, to me, anyway)
13:28:19funmanb0hoon: we would need a new rbutil release too? it lists the vibe 500 as unstable
13:28:48ranmaHey, first time I got that far:
13:28:52ranmaMay 15 20:26:56 navi kernel: [207036.298943] usb 1-3: New USB device found, idVendor=0781, idProduct=7452
13:28:55ranmaMay 15 20:26:56 navi kernel: [207036.298943] usb 1-3: New USB device strings: Mfr=1, Product=2, SerialNumber=3
13:28:58ranmaMay 15 20:26:56 navi kernel: [207036.298943] usb 1-3: Product: Rockbox media player
13:29:01ranmaMay 15 20:26:56 navi kernel: [207036.298943] usb 1-3: Manufacturer:
13:29:04ranmaMay 15 20:27:01 navi kernel: [207041.299256] usb 1-3: can't set config #1, error -110
13:29:29funmanmarc2003: hey, what is the exact size of the drive reported by your OS ? (whole drive, not partition)
13:29:39funmanranma: \o/
13:30:04funman'refresh database' must die, with pain
13:30:22ranmaThe error is because it called usb_drv_send with len>64, which I don't handle yet :)
13:30:42funmanranma: do you use DMA ?
13:30:45 Join moos [0] (moos@rockbox/staff/moos)
13:30:48ranmaBut I think sending control responses works now.
13:30:53ranmaYes, it's using DMa.
13:31:22ranmaI don't think you can use PIO with this thing and it has it's own DMA controller...
13:31:33funmanaah so you're not using the pl081 ?
13:33:01 Join anewuser [0] (anewuser@unaffiliated/anewuser)
13:33:24marc2003funman: 7810MB according to disk management on windows 7. i'm not sure how to get more accurate. i have ubuntu in VMware if you can tell me what to do (i'm a linux noob)
13:33:29b0hoonfunman: yes, but it can be unstable i think in rbutil for the next rbutil release
13:33:42b0hoonfunman: please look at the last sentence in "Current status of supported targets" section, (TargetStatus on the Wiki)
13:33:57ranmaNope, it has it's own DMA descriptor format, flags etc.
13:34:08b0hoonfunman: "Whilst every effort is made to try and ensure..."
13:36:00funmanmarc2003: oops I see the bug it's my fault, sorry :D
13:36:46funmanb0hoon: ah ok
13:36:49ThomasAHfunman: works fine now
13:36:58ThomasAHfunman: (even while and after playing radio)
13:37:13funmanranma: ok i thought we would have to use pl081 channel 1 and give up any future hope for USB audio ;)
13:39:21funmanmarc2003: can you test ?
13:39:57ThomasAHfunman: thanks ... one question regarding patching the manual: Would you or other devs prefer one larger patch which fixes/improves documentation for clip1/2/+ buttons or smaller patches with one change per patch?
13:42:16 Join sadzior [0] (
13:42:49sadziorwhen will rockbox 3.6 be released
13:43:09funmanThomasAH: depends what you consider 'one' change
13:43:28funmansadzior: there's no immediate plan for it at the moment
13:44:11AlexPsadzior: Not confirmed, but I guess it will be a few weeks
13:44:28AlexPIt shoyuld be fairly soon though, no?
13:44:36AlexPDo we want it before or after devcon?
13:44:56 Join petur [0] (~petur@rockbox/developer/petur)
13:45:54ThomasAHfunman: currently it is one patch "improve the manual regarding buttons of Sansa Clip and Clip+" with subtopics "missing hold button on Sansa Clip+" and "reverts FS #11241, which confused "context of WPS" (button_context_wps) and "context of WPS context menu" (button_context_tree)"
13:46:11ThomasAHfunman: then I noticed that Home/Submenu are wrong in the WPS help, too
13:47:34ThomasAHfunman: preparing a series of patches with corrsponding descriptions for commit messages is no problem for me, but of course more work for the one who has to apply/commit in the end
13:49:10funmando what you feel is better
13:49:37ThomasAHfunman: ok, then I'll make single patches ... combining them is easier than splitting :)
13:50:42***Saving seen data "./dancer.seen"
13:51:39marc2003funman: i'm getting these errors applying the patch
13:51:42marc2003patch unexpectedly ends in middle of line
13:51:50sadziorare there any plans to improve battery life on fuze?
13:51:59marc2003Hunk #2 succeeded at 817 with fuzz 1.
13:52:34marc2003i'll try it anyway :p
13:52:41 Quit TheSeven (Remote host closed the connection)
13:53:11funmanmarc2003: just add a newline at the end of the patch (pastie removes it)
13:53:31funmansadzior: well it depends, do you have any plan to improve it ?
13:54:18bertrikI'm not aware of any obvious power wasters on the fuze
13:54:41 Quit stoffel (Ping timeout: 246 seconds)
13:55:41bertrikthere have been improvements in runtime by disabling the radio chip if unused and also recently a patch to stop DRAM clocks when idle
13:55:44funmanbertrik: playback wrt idle mode needs 10mA more on rockbox, and only 2mA more on OF (on c200v2)
13:58:05bertrikmaybe we can reduce some bias currents in the audio stages, but that would likely make audio quality worse
13:58:39sadzioryeah, but it would require some learning c and asm, so it makes it pretty much long termed, so to say
14:00:18marc2003funman: that looks to have done the trick. the database built with no problems. thanks.
14:00:32funmanlast test i made, rockbox lasted 66% of the OF time
14:01:53funmanmarc2003: thanks
14:01:55CIA-5New commit by 03funman (r26050): as3525: fix capacity sanity check ...
14:02:21 Quit marc2003 (Quit: ChatZilla 0.9.86 [Firefox 3.6.4/20100503122926])
14:02:33 Quit FlynDice (Remote host closed the connection)
14:03:09bertrikranma made some comparison of codec settings between rockbox and OF, right (IIRC)?
14:04:32funmanyes, saratoga committed some changes with this data
14:06:07funmani have an analog multimeter, perhaps i could try to use it on my clip
14:06:31funmanIIUC i need to remove the battery and provide another source of current ?
14:07:24funmanbertrik: i have tried to higher the burst size for pcm DMA transfers but it had a very little (if any) effect on battery
14:07:30 Quit krazykit (Read error: Connection reset by peer)
14:08:27bertrikfunman, I think if you measure the current drawn through the USB cable when the battery is fully charged, then that current is also a good way to compare power consumption
14:11:01 Join krazykit [0] (~kkit@
14:11:04funmanwhy is there 3 wires on the battery? -, + and N
14:11:41funmanis the ground - or N ?
14:15:35bertrikprobably one wire for a thermistor to measure the battery temperature
14:16:55bertrikI think - is the "ground", + is the positive terminal and N is the thermistor (maybe N for NTC?)
14:17:29 Join stoffel [0] (
14:17:48funmani think too, + is red, - is black and N is blue
14:20:15 Quit flydutch (Quit: /* empty */)
14:23:36 Quit bertrik (Remote host closed the connection)
14:26:08 Join n1s [0] (~n1s@rockbox/developer/n1s)
14:27:23ranmafunman: BTW, still haven't gotten around to do proper current measurements on memory reads.
14:27:24 Join einhirn_ [0] (
14:28:03ranmaI tried it with my cheap multimeter at home, but the voltage drop over the internal measurement resistor is too high.
14:30:07ranmaAt least I can say it looks like a tight memory read loop uses more current on iram than extram, maybe because iram is faster and it can do more read operations / less waitstates, meaning a less idle processor.
14:30:50ranmaBecause on the iram loop the player turns off with voltage dropping too low, with extram it stays on.
14:31:08 Quit einhirn (Ping timeout: 240 seconds)
14:37:14 Join Jaykay [0] (
14:38:41sadzior '<
14:41:27funmanbertrik: btw i have my clipv1 opened and it's hardware revision 1.4, is yours different ?
14:41:30 Join theli_ua [0] (
14:42:57 Join CFP [0] (
14:48:07 Part b0hoon ("GTG. Bye.")
14:53:50 Quit stoffel (Ping timeout: 246 seconds)
15:04:03 Join Trou [0] (
15:04:45 Quit funman (Quit: free(random());)
15:06:23 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:08:43CFPhello everyone :) Before I reboot to linux, is there anybody here who know how to modify kkurbjun gigabeat flashwriter to accept a new hash?
15:09:35*kugel grumbles a bit about the libc related committs yesterday
15:09:48CIA-5New commit by 03jethead71 (r26051): Gigabeat S: Fully enable access to hardware tone controls and 3-D effect feature. Under the hood, it's designated a hardware equalizer since it is ...
15:16:32 Part sadzior
15:17:14kugelgevaerts: we use our *printf on cygwin/mingw, I believe, which could be why %z fails for it. I think we should just add it to our implementation
15:17:55kugelOR we check again whether we really need our *printf there, that comment is old and in the meantime there's a new cygwin version
15:18:00 Part theli_ua
15:27:20 Join cdb [0] (
15:28:10 Join adnyxo [0] (
15:30:04CIA-5New commit by 03jethead71 (r26052): Hopefully get some green back from r26051.
15:31:32 Quit CFP (Read error: Connection reset by peer)
15:32:15 Quit moos (Ping timeout: 258 seconds)
15:32:23kugelJdGordon: I would also vote for optional parameters always at the end
15:32:51JdGordonI'm happy with using a differnet tag for those instead
15:33:18 Quit einhirn_ (Read error: Connection reset by peer)
15:34:15 Quit Highlander (Quit: Quitte)
15:35:33gevaertskugel: right. I'd prefer to do both think, i.e. support %z (that takes about 16 bytes according to my experiments last night), and use native printf in the sim
15:36:09*gevaerts *hates* casts in printf-like calls. Those should use the correct format string
15:36:32*kugel too
15:37:26JdGordonwhat is %z?
15:37:27kugelI'm wondering which cygwin version Blue_Dude is running
15:37:33pamaurythere are not enough modifiers ! Which one do you use to output a uintX_t for example ?
15:37:54kugelpamaury: c99 specifies tags for them but we do not support them
15:38:33gevaertsJdGordon: that's the one for size_t
15:40:13pamaurylet's use C++ then :)
15:40:26kugelwhy c++?
15:42:12JdGordonyes! and then real OO!
15:42:19JdGordonnone of this nonesense marco crap to do it
15:42:40*n1s grabs a pitchfork
15:42:49JdGordonand malloc! and xml!
15:43:05*n1s stabs JdGordon with pitchfork
15:44:43kugelJdGordon: not malloc, but new!
15:47:10 Quit pamaury (Ping timeout: 252 seconds)
15:50:14kugelgevaerts: btw, I didn't move firmware/include/sys/ to libc because it's not iso c and I didn't decide what to do with it yet, of course I didn't know it breaks weirdo cygwin again
15:50:40 Join Blue_Dude [0] (~chatzilla@rockbox/developer/Blue-Dude)
15:50:44***Saving seen data "./dancer.seen"
15:50:53*Blue_Dude grumbles about compile warnings. :)
15:51:06kugelBlue_Dude: what cygwin version do you run?
15:51:15Blue_Dudekugel: Using 1.7.
15:51:21Blue_DudeI think...
15:51:41Blue_DudeChecking where I can find that info.
15:51:47 Join cfp [0] (
15:52:06kugelI assume you could just run the installer again, that worked on 1.6
15:53:29Blue_DudeIt doesn't say.
15:53:31cfpGodEater: Refering to FS #7505, you did flash you Gigabeat, although it wasn't recognized by kkurbjun tool.
15:53:49kugelBlue_Dude: can you update?
15:53:51cfpcould you provide me with a little more info on how you did that?
15:54:06Blue_DudeIt is some variant of 1.7. I don't know the subversion.
15:54:21cfpgevaerts: I believe you did too. Could you explain how you modified the code?
15:54:30kugelJdGordon: that also works, remember you invented the dual use of %Vi :)
15:54:47cfpI'm not really sure I understand how it works
15:55:54cfpOr else, if anyone has a clue, I'm all ears :)
15:55:56kugelgevaerts: we probably should support more tags in *printf, e.g. to make logf and debugf more compatible
15:56:30gevaertscfp: see the top of gigabeat_flash.c, but make sure you know that there are some risks :)
15:56:36Blue_Dudecygwin1.dll has a version of 1.7.1.
15:57:05cfpgevaerts: I know =) But I'd like to try it anyway, although I don't quite get how the array is formatted, despite the comments
15:57:31gevaertscfp: don't touch the array, that's overkill!
15:57:39Blue_DudeThe problem with %z was compiler warnings. Signed size_t format, ssize_t argument. I know, they're the same thing, but the formatted didn't know that.
15:58:09gevaertsBlue_Dude: your fix is technically wrong on 64 bit systems though
15:58:12cfpgevaerts: oh? So I should just ENABLE_DEV ?
15:58:33gevaertsyes, but don't tell anyone! We don't want people to test this without thinking about the risks first
15:59:03Blue_DudeSo cast the ssize_t arguments as long with %ld tags and it was happy.
15:59:19gevaertsBlue_Dude: yes, but I'm not
15:59:19kugelBlue_Dude: have you tried #define ssize_t signed size_t?
15:59:21cfpgevaerts: I will be silent (do expect me to ask a lot of questions though if it fails =))
15:59:36gevaertsBlue_Dude: that's a workaround at most
15:59:51cfpgevaerts: I've seen that some people restored a backup posted by others that was recognized by the code ; would you recommend that?
16:00:01kugelBlue_Dude: I don't know if it's a windows or a msvc thing, but long is usually 32bit under windows environments
16:00:22kugelwhile size_t is 32bit or 64bit depending on the cpu
16:00:37gevaertscfp: I'd suspect that if one image works, other images compatible with that image will also work, so it probably doesn't matter much
16:01:15Blue_DudeAnd the Rockbox stdio.h had a "foreign" function that was not present in the other stdio.h libraries. format.h already had the prototype so I used it.
16:01:30kugelyea, I forgot to remove it from stdio.h
16:02:46jhMikeSdoes bass/treble cutoff already show up in the menus on D2? something's odd here.
16:03:07cfpgevaerts: Ok, I guess I'll give it a go then :)
16:03:09 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
16:03:32 Join Blue_Dude [0] (~chatzilla@rockbox/developer/Blue-Dude)
16:03:38Blue_DudeSorry. I could see your replies on the log but not in the IRC window.
16:04:06cfpgevaerts: What about those disabled lines though?
16:04:11cfp#if !defined(ENABLE_DEV)
16:04:11cfp splash_pause(true, "ERROR: Forcing backup restoration!");
16:04:11cfp do_flash_restore();
16:04:11DBUGSent KICK cfp to server
16:04:11Kick(#rockbox cfp :No flooding!) by logbot!
16:04:29 Join cfp [0] (
16:04:48cfpseems I shouldn't have posted those lines
16:04:57cfpsorry everyone :s
16:05:13*gevaerts thinks that while that should indeed have gone to a pastebin, logbot might be a bit too trigger-happy
16:05:48Blue_Dudekugel: I can indeed update. I think we're up to 1.7.5 now.
16:05:52jhMikeSodd, it defined the CAPS but the string is only designated for ipodvideo. somehow I woke-up the feature, probably throught a proper #include somewhere.
16:06:31gevaertscfp: that's fine too. If I understand the code correctly, it checks there if the end result has an expected md5sum, which it won't if the initial setup didn;t match
16:06:51cfpgevaerts: Ok, thanks =)
16:07:45gevaertskugel: did you see funman's %z patch last night?
16:07:54 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
16:08:30kugelgevaerts: yes, now. I don't think that's the correct way
16:09:41kugelor specifically it should cast to size_t in the body of that case
16:11:02kugelI'm not entirely sure if (s)size_t valid in va_arg(), I think there are some type restrictions for casting in va_arg()
16:12:08gevaertskugel: we know this code is for on-target rockbox use though, so assuming that size_t and long are the same there could be OK
16:12:15Blue_DudeAnyway, sorry if I stepped on toes with argument casts.
16:12:47kugelgevaerts: not necessarily. format.c is compiled for sims to, and is used for fdprintf and vuprintf (IIRC)
16:13:02gevaertsah, ok
16:13:33gevaertsIn that case the patch is indeed too simple and probably wrong
16:13:55*gevaerts wants to know where mingw's ssize_t comes from
16:15:27cfpIt works /perfectly/ \o/
16:15:48*cfp bows to kkurbjun and thanks gevaerts =)
16:16:18Blue_Dudegevaerts: \usr\include\mingw\sys\types.h
16:16:36CIA-5New commit by 03jethead71 (r26053): Hopefully finish off the red from r26051.
16:17:14Blue_DudeIt's a long there.
16:19:55 Nick dfkt_ is now known as dfkt (dfkt@unaffiliated/dfkt)
16:20:06Blue_DudeAlso \usr\include\sys\types.h
16:20:13gevaertsok, gives me the same warning. This is *wrong*!
16:20:31cfpDouble yeah :))
16:20:42gevaertscfp: happy? :)
16:20:46cfpMy second Gigabeat works too \o/
16:20:53cfpgeaverts: more than that =)
16:21:25kugelis format speed critical or why don't we just use a full-blown implementation of which a hand full exists on the net?
16:21:55kugelgevaerts: does #define ssize_t signed size_t work?
16:22:05gevaertskugel: I definitely hope that no speed critical code is mucking around with format strings :)
16:22:18Blue_Dudekugel: won't that throw away the sign?
16:22:19cfpgevaerts: I must say Karl did an impressive work there. And this is definitely convincing me /not/ to buy another DAP =)
16:22:20gevaertsA full-blown one will use more RAM though
16:22:42Blue_DudeOr at least make a negative number really, really big?
16:23:15kugelif we give up 10k on fonts then we can give 500 bytes on a widely used function IMO
16:24:31 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
16:25:19gevaertskugel: IIRC one of our committers has worked on
16:26:12Lloreankugel: And the fact that people use arguments like that is exactly why I'm against giving up RAM so often. Everything should be judged on its own merit, not how much RAM was used for other things (I have no idea if your thing is worth the RAM or not, I'm just opposed to that being used as an argument)
16:26:57*Llorean also dislikes the "if I free up X amount, I should then be able to spend <X amount for something else just because it's still a net gain" argument.
16:27:03gevaertskugel: "signed size_t" does not work at all
16:27:32 Quit n1s (Ping timeout: 245 seconds)
16:28:41kugelgevaerts: something is broken on the mingw side I think
16:28:49gevaertskugel: yes, I'm going to ask there
16:30:09gevaertscfp: amazing how having two players makes one more likely to try risky things, isn't it? ;)
16:34:02gevaertskugel: apparently the basic problem is that msvcrt (which mingw-produced binaries use) doesn't know about %zd
16:34:19kugeljep, I'm reading :)
16:37:03gevaertsI suspect we have to either provide our own stdio or stick to strick iso-C things
16:37:13 Quit mt (Remote host closed the connection)
16:37:45cfpgeaverts: Quite :)
16:37:47 Join Boldfilter [0] (
16:37:51jhMikeSnow, gigs won't let me power down when plugged, not just when charging. :\
16:38:43pamauryIf you really want a bullet proof, gcc specific solution, then you can use gcc builtins to compute at compile time the good printf modifier...
16:40:09cfpgevaerts: Actually, I thought my first one was dead when I came back from a travel last year, and then it happened that opening it and pluging the drive directly to my computer did the trick, so that's basically when I installed rockbox. I then got the second one when the first's screen broke, and eventually replaced the screen anyway =)
16:43:27cfpAnyway, I'd like to thank all the people behind the rockbox firmware. It's a great project, great to use, great to involve in, which has a wonderful dev team, lots of helpful people, and an impressive dynamic. Thanks everyone!
16:45:51cfpgoodbye everyone :)
16:46:07cfpgevaerts: thanks for your help =)
16:46:27 Quit cfp (Quit: Quitte)
16:51:58*gevaerts looks at trio
16:53:25kugelgevaerts: seems we should go for PRIdMAX
16:54:11kugeleven if we use our format; if we use the mingw headers we will still get that warning
16:55:35gevaertskugel: that's *ugly*
16:56:19kugelugly but at least iso c compatible
16:56:36pamaurywe don't write iso c...
16:57:04gevaertskugel: do you really propose to use that in *every* format string?
16:57:29kugelonly for the ones that use (s)size_t I guess
16:57:38gevaertswhich is a lot of them
16:58:04pamaurywhat is this PRIdMAX thing ?
16:58:17CIA-5New commit by 03jethead71 (r26054): Gigabeat S: There's no reason to pause the LCD DMA when changing the framebuffer address.
16:58:33gevaertspamaury: it's an ugly macro that contains the format specifier for intmax_t
16:58:36 Join funman [0] (~fun@rockbox/developer/funman)
16:58:42kugelor we make our own stdio, which includes our format.h on mingw/cygwin and the systems stdio.h on the rest
16:59:04funmanwhat about this: ignore cygwin/mingw completely
16:59:12gevaertskugel: yes, that's what I propose
16:59:28kugelbut we can't use native printf then
17:00:07gevaertsis that a problem? Native printf is broken
17:00:41 Quit pamaury (Quit: Page closed)
17:00:48funmankugel: - slows down scrollwheel updates on fuzev2 (minimum of 3 ms between handling a new interrupt)
17:00:51 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
17:00:58*gevaerts looks at trio
17:01:08 Join dfkt_ [0] (~dfkt@unaffiliated/dfkt)
17:02:00kugelfunman: does it work well? I have thought about a similar solution
17:02:03 Quit dfkt (Disconnected by services)
17:02:06 Nick dfkt_ is now known as dfkt (~dfkt@unaffiliated/dfkt)
17:02:12kugel3ms sounds a bit high though
17:02:27funmanseems to be fine, i have tried with 5ms to be like fuzev1 (1/(2*HZ)) but 3ms is fine
17:03:17CIA-5New commit by 03jethead71 (r26055): If HAVE_POWEROFF_WHILE_CHARGING is not defined and the charging configuration specifies CHARGING_MONITOR or greater, allow poweroff while plugged but ...
17:03:33funmankugel: gevaerts: i am not trolling when I propose to ignore cygwin/mingw/msvcrt, there are usable workarounds (running linux in a VM) for people using windows
17:04:02pamauryis the problem only about warnings ?
17:04:12gevaertspamaury: no, unfortunately not
17:04:20kugelfunman: we don't talk about compiling target builds from windows, but to compile the sim (or RaaA later) to run in windows
17:04:22gevaerts%zd and %zu print "zd" and "zu"
17:04:34pamauryand why do we need %zd and %zu ?
17:04:38funmani guess printf() can also crash ? variable arguments are shifted
17:04:44gevaertspamaury: for size_t and ssize_t
17:04:50funmanoh sorry ..
17:04:58gevaertsfunman: oh, right. Probably that too
17:05:20 Quit bmbl (Quit: Bye!)
17:05:34kugelgevaerts: trio looks messy, looks more sane
17:05:41funmancan you add a platform check to use specific cygwin code in this case?
17:06:15pamauryI don't see the point, if it's only for size_t and ssize_t, use a macro like PFSS for size_t modifier and go on
17:06:20funmana whole new printf() shouldn't be needed if the only thing which needs fixing is '%z'
17:06:39kugel7.7k lines seems overkill
17:07:05gevaertskugel: or just extend out format.c properly and use that
17:07:10 Join cfp [0] (
17:08:09funmanisn't mingw_printf() the c99 compatible version of mingw ?
17:08:30cfpfunman: Did you see my patch at ?
17:09:19cfpfunman: been testing this for a month now, works fine =)
17:09:45 Join GodEater_ [0] (
17:09:45 Quit GodEater_ (Changing host)
17:09:45 Join GodEater_ [0] (~bibble@rockbox/staff/GodEater)
17:10:37 Quit GodEater (Quit: leaving)
17:10:50 Nick GodEater_ is now known as GodEater (~bibble@rockbox/staff/GodEater)
17:11:25gevaertsfunman: yes, that's the sort of uglyness I don't want
17:11:33cfpfunman: could you commit it? The current version quite frequently gets calculations wrong
17:12:10funmancfp: looking at it
17:12:31funmangevaerts: well there has to be a workaround somehow
17:12:55funmanthe PRI* aren't needed as long as we don't try to print int??_t
17:13:52funmanif you don't want to use printf from mingw 'libc.dll' you'll have to redefine printf to something else anyway
17:14:01gevaertsand how do you avoid printing those?
17:14:39funmaneasy: you don't at all
17:15:12funmani suggest you ask j-b on #videolan as he went through this trouble already
17:18:03funmanhm i can't find any doc on mingw_snprintf, nor find it in the headers i have :o
17:19:53gevaertstrio handles ssize_t and %zd wrong
17:20:47pamaurythere is an alternative solution: under mingw, define printf as a macro which calls a converter function, that way, it will print everything correctly
17:20:48CIA-5New commit by 03funman (r26056): Fix alarmclock plugin time miscalculation ...
17:21:46pamaury(if it's just for the %zd and %zu)
17:22:24funmanpamaury: int*_t as well it seems
17:22:53funmanchanging format string in place would work if the format string is writable
17:23:17funmanand if new format specifiers are not larger (ex: %ld -> %lld)
17:23:19kugelwell, you could copy it to a spare buffer
17:23:21 Join n1s [0] (~n1s@rockbox/developer/n1s)
17:23:35funmanah right you could use the stack
17:23:53kugelbut a converter for all sorts of *printf seems a lot of work
17:23:57gevaertsIf you're doing that, why not just have the macro call another printf()?
17:24:23 Quit pamaury (Quit: Page closed)
17:24:33 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
17:25:02funmanwhich one?
17:25:26pamaurythe problem is only with mingw ?
17:25:39gevaertspamaury: who knows?
17:26:09pamauryI mean, does this problem arise on target ?
17:26:18gevaertson target we use our own formatter
17:26:40gevaertsThat one doesn't do %z right now, but we can fix that
17:26:54pamauryYes, so it's just for mingw basically
17:28:18pamauryAnd I don't see the problem with int*_t, it's about finding the good modifier on different hosts ?
17:28:36 Quit jordan` (Quit: Coyote finally caught me)
17:33:17funmangevaerts: btw, i think specifiers are only a real problem when using int64_t
17:33:58funmanpamaury: int*_t specifiers are in the form PRId64 defined in headers
17:34:12kugel__mingw_printf seems like the way to go
17:34:38funmanin fact they are defined to "lld" "ld" or alike
17:34:50kugelwe just need -D__USE_MINGW_ANSI_STDIO=1 on the cmd line it seems
17:35:10funmanI think the problem is that windows use special specifiers for them
17:35:26kugelgevaerts: can you test compile your test.c with that?
17:35:31funmanyeah, "I64d", which gnu functions don't know
17:36:09*pamaury always thought printf modifiers were broken
17:36:36funmanpamaury: well %d is a printf modifier and i guess it's working everywhere
17:36:47pamaurynot, the size modifiers
17:36:58funman%8d ?
17:37:00pamauryyou should be able to write %d for any integer
17:37:13pamauryand not fiddle with those horrible %ld, %hd, ...
17:37:14gevaertskugel: what exactly? Just adding -D__USE_MINGW_ANSI_STDIO=1 doesn't change anything
17:37:25pamauryand %lld now...
17:37:46kugelit should according to
17:38:30gevaertsMaybe with a newer version
17:40:55funmanworks for me, with ubuntu mingw (4.2.1-sjlj)
17:41:39funmanprintf() is redefined internally to mingw headers to use __mingw_printf()
17:42:02funmanand PRId64 works like expected, no redefinition needed
17:42:12gevaertsDoes %z work?
17:42:18kugelgevaerts: i586-mingw32msvc-gcc -D__USE_MINGW_ANSI_STDIO=1 test.c -o test.exe -Wall works for me
17:43:21gevaertsok, so my mingw is probably too old. It also claims to be 4.2.1-sjlj...
17:43:38kugelsame here... :)
17:43:41funmangevaerts: did you include stdio.h ?
17:43:51gevaertsfunman: of course!
17:43:58 Join jordan` [0] (
17:43:59cfpany commiter for ? small fix in francais.lang
17:44:23funmangevaerts: i'm using the 32bits version of mingw on 64bits host if it makes a difference
17:44:29gevaertsfunman: me too
17:45:28pamaurycfp: I would commit it if I could but I'm behind a proxy, however I can confirm it's a good french translation :) So anyone can commit it please ?
17:45:40ender`what's the problem with mingw?
17:47:52funmangevaerts: 0a6f49536b39cc1a6f733b5ea9e7b399 /usr/i586-mingw32msvc/include/stdio.h
17:48:06cfppamaury: :)
17:48:16funmancfp: pamaury: je vais le faire
17:48:36gevaertsfunman: different here. Oh well...
17:48:39cfpfunman: a marche =)
17:48:52gevaertsIf newer mingws can handle this, I think we're fine
17:49:23CIA-5New commit by 03funman (r26057): Small mistake in francais.lang ...
17:49:33kugelit also works with mingw32 from the mingw-w64 package
17:49:41cfpfunman: merci =)
17:49:46funmanno problem ;)
17:50:05 Join amr1 [0] (
17:50:21cfpgoodbye everyone =) (funman: remember about FS #11110 ;))
17:50:26 Quit cfp (Quit: Quitte)
17:50:39*kugel has the same file as funman
17:50:46***Saving seen data "./dancer.seen"
17:51:02kugelgevaerts: we are not fine if the cygwin shipped gcc is too old too
17:51:49CIA-5New commit by 03jethead71 (r26058): i.MX31: Issue some NOP's immediately after MCR WFI to prevent premature execution of subsequent code.
17:52:13kugelpixelma: can you compile with "gcc -D__USE_MINGW_ANSI_STDIO=1 test.c -o test.exe -Wall" under cygwin?
17:52:48gevaertspixelma: or (preferably)
17:53:09gevaertsThat one makes the ssize_t negative, so we can test if it *really* works
17:53:11funmanis our firmware/include/_ansi.h going to be used elsewhere than on target builds ?
17:53:14*kugel slaps gevaerts
17:53:50kugeloh, sst is -1!
17:53:55kugelthat it doesn't work here
17:54:01pixelmait's the same link... also..
17:54:16gevaertsoh, it edited in-place?
17:54:16kugelit prints 1 1 for me
17:54:35pixelmayou lot need to be a bit more descriptive about what I need to do
17:54:38kugelah wow
17:54:52kugelsince when does do that!?
17:54:58gevaertspixelma: put it in a file named "test.c" and run kugel's command
17:55:48kugelok, nevermind me, it does work
17:59:02 Quit merbanan (Ping timeout: 240 seconds)
17:59:08pixelmait just complained about no newline at the end of file
17:59:27gevaertsWhat output do you get if you run the resulting test.exe?
17:59:31pixelmaanything as a result I need to check - another file or so?
17:59:51pixelma-1 1
18:00:05gevaertsWait, cygwin... That will build cygwin binaries by default, not mingw
18:00:43gevaertsGood enough for cygwin sims I guess
18:00:56kugelthe difference seems to be a command line switch, I suspect -DXXX stuff is compatible anyway
18:01:05pixelmawhat about crosscompiled sims?
18:01:13 Quit JohannesSM64 (Read error: Operation timed out)
18:01:31gevaertspixelma: binaries built with mingw on ubuntu seem to be OK
18:01:39pixelmaoh, you said this version works for you before
18:01:51amr1kugel: ping
18:01:57*gevaerts is now confused
18:02:46kugelpixelma: now you could compile a sim in cygwin, but before running make edit the Makefile (in the build dir) and add '-D__USE_MINGW_ANSI_STDIO=1' to the GCCOPTS line
18:02:56kugelamr1: pong
18:03:02amr1I was testing diacritics support on simulator, but after r26019 I noticed that the last character containing diacritics appears as the first character in the line
18:03:08pixelmakugel: if you have some minutes ;)
18:03:16kugelsure :)
18:03:16amr1could that be related to the host headers ?
18:03:40kugelamr1: on the sim?
18:03:42kugelor both
18:04:20amr1till now, on the sim, i didn't install on target yet, but it was working till r26018
18:04:49kugelcan you delete the build dir and try again?
18:05:32kugelwhile it appears possible, I doubt it's related to my commit. I think the diacritics code doesn't use standard functions
18:05:39amr1I did that already, and even tried to download a new fresh tree from svn
18:05:50pixelmakugel: I guess I should update to recent SVN before or not?
18:06:47kugelpixelma: maybe to r26039, r26040 "fixed" warnings which should not appear with that line added to GCCOPTS
18:08:52pixelmasvn seems a bit slow to me today
18:10:20pixelmaI'll do a make clean just in case even though it'll take longer
18:11:07 Quit n1s (Quit: Lmnar)
18:19:34amr1kugel: diacritics.c includes only system.h, while lcd-bitmap-common.c includes <stdio.h>, <stdarg.h>, "string-extra.h"
18:19:40amr1 could it be a function in "string-extra.h" ?
18:21:49CIA-5New commit by 03funman (r26059): fuzev2: fix buttonlight flashing on µSD access (2nd try) ...
18:23:43funmankugel: about the fuzev2 scrollwheel should I put this on flyspray? you want to experiment with other delays than 3ms ?
18:25:19 Nick fxb is now known as fxb__ (
18:28:19 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
18:30:17kisaknovel question funman, is there any real penalty to that buttonlight flickering with uSD? (decreased battery life/premature death of the light/ -or- just because it's inconsistant with other player behavior?)
18:30:21 Join Strife89 [0] (
18:31:26funmani didn't measure (or tried to) any impact, it's just for consistance (same thing happens on fuzev1/e200v2)
18:31:52kisakfair enough
18:33:06kugelfunman: just go for it, if it's an improvement
18:34:16CIA-5New commit by 03funman (r26060): fuzev2: leave at least 3ms between scrollwheel events ...
18:35:55pixelmakugel: I got three warnings but I am not sure if they are just because of the "hacked" makefile
18:38:18pixelma with r26039 and an M5 sim with FM functions enabled
18:40:30 Quit kugel (Ping timeout: 264 seconds)
18:43:59 Join kugel [0] (~kugel@rockbox/developer/kugel)
18:44:37kugelpixelma: :/ that's the warning we try to resolve
18:49:31kugelpixelma: no other warnings, though?
18:49:41kugelexcept the buffering.c one I mean
18:50:31pixelmathe midiutil and a different ones (all of the same type) which have been there before though
18:50:56pixelmaI think they don't matter here
18:51:00 Join merbanan [0] (
18:51:10 Join CGL [0] (~CGL@
18:51:12*kugel wonders if r26058 also helps on as3525v2
18:51:15kugelfunman: ^
18:51:37kugelamiconn mentioned cpu_/core_sleep problems on PP
18:52:44funmani think i tried that already
18:52:58funmanbtw i wondered what would happen if we wrote CGU_PROC & CGU_PERI in one ldm instruction
18:58:03funmanlooking at i2c -> i can trigger INT_AUDIO interrupt just by reading IRQ_ENRD2
18:59:41 Nick CGL is now known as X_CGL (~CGL@
19:00:36 Nick X_CGL is now known as n00b2Hack_CGL (~CGL@
19:00:46 Quit kugel (Remote host closed the connection)
19:02:50 Nick n00b2Hack_CGL is now known as CGL (~CGL@
19:03:01 Nick CGL is now known as GGG (~CGL@
19:03:45 Quit funman (Quit: free(random());)
19:03:50 Nick GGG is now known as Gen_CGL (~CGL@
19:04:39 Nick Gen_CGL is now known as ZZZzzzz (~CGL@
19:04:39 Join Luca_S [0] (
19:04:45 Nick ZZZzzzz is now known as Z (~CGL@
19:05:08 Nick Z is now known as ZZZZ (~CGL@
19:05:14 Nick ZZZZ is now known as ZZZZZ (~CGL@
19:05:48 Nick ZZZZZ is now known as ZZZzzzZZZ (~CGL@
19:07:14pixelmacould you please stop that?
19:08:45 Nick ZZZzzzZZZ is now known as [CGL] (~CGL@
19:08:56 Nick [CGL] is now known as [CGL]_ZZZzzzZZZ (~CGL@
19:13:45 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:13:56amr1... or may the choice of compiling host libs or rb libs be a configure option before making ?
19:15:47kugelgevaerts: had a look at my sim-target-tree work?
19:17:24gevaertskugel: a bit. I've tried to rebase it to current master so I could see the entire thing as one diff, but I failed
19:17:36kugelgevaerts: don't do that :P
19:17:51gevaertsWhy not? It was in a temporary branch!
19:18:07kugelgot co master && got co -b diff_branch && git merge sim-target-tree && git diff master..HEAD
19:19:04kugelmaybe update master and sim-target-tree so that they're on the same svn rev before
19:19:53kugelgevaerts: I uploaded the entire diff on the tracker, btw
19:20:22kugel(again, co==checkout, /me always forgets his aliases are not default)
19:22:35kugelgevaerts: don't do that because rebasing is not suitable for this situation
19:23:03kugelI don't even know how it handles the merge commits
19:23:10kisakkugel: what? git co is not common everywhere?
19:24:58 Join stoffel [0] (
19:25:42 Quit GeekShadow (Read error: Connection reset by peer)
19:26:38 Join mikroflops [0] (
19:27:49 Join LowRider [0] (
19:27:49 Join JohannesSM64 [0] (
19:29:19*LowRider Brand New!Notebooks and LCD TVs.Discounts up to 30%. The newest electronics only
19:29:36Mode"#rockbox +o gevaerts" by ChanServ (ChanServ@services.)
19:29:38Kick(#rockbox LowRider : ) by gevaerts!~fg@rockbox/developer/gevaerts
19:29:41Mode"#rockbox -o gevaerts" by ChanServ (ChanServ@services.)
19:30:21 Quit mikroflops_ (Ping timeout: 240 seconds)
19:32:41gevaertskugel: I'm going to assume that the patch indeed does what your description says it does :)
19:36:36 Join funman [0] (~fun@rockbox/developer/funman)
19:36:39kugelgevaerts: still problems looking at it?
19:37:06gevaertskugel: no, but, well, you know how big that diff is :)
19:37:13funmanpamaury: i have found the usb init code on clipv2!
19:37:30kugelgevaerts: you could look at the individual commit diffs
19:37:31pamauryfunman: great !
19:37:39gevaertskugel: not after that merge!
19:37:44pamauryfunman: what does it do ?
19:37:45kugelalthough they're somewhat hard to find with that merge commits
19:37:50funmanpamaury: it looks like the as353x one
19:37:52gevaertsThat's why I wanted a rebase :)
19:37:56funmanpamaury: dunno ;)
19:38:04funmani'm looking at i2c code for now
19:38:12pamauryat which address ?
19:38:17funmanit is executed in the AUDIO isr
19:38:25kugelgevaerts: I think searching for "Thomas Martitz" in the shortlog view on
19:38:43pamauryhuh, that's highly logical
19:38:59gevaertskugel: it doesn't really matter. Moving files around isn't *that* exciting to look at
19:39:49pamauryfunman: true, there are lots of references to usb registers, that's great
19:40:25funmanpamaury: i have just looked at reference to 0x804 bit 1 which is soft disconnect of dctl in linux
19:40:59pamauryI'll have a closer look at it later, I'm just wondering why it's in the audio isr...
19:41:17pamauryis it shared in some way with something else ?
19:41:26funmanwe detect usb this way on as3525v1
19:41:32funmanand soon, on as3525v2 too
19:41:47pamauryhow ?
19:41:58 Join webguest17 [0] (
19:41:58pamaurythe usb plug causes an audio isr ?
19:42:24webguest17I went to and filled in the spanish translations
19:42:35funmanpamaury: yes, look at ascodec-as3525.c
19:42:50webguest17Iḿ going to submit it to the patch tracker, is there anything else I should do_
19:43:49pamauryfunman: ok, I saw it once but wondered what it was doing here :)
19:44:15gevaertskugel: that looks OK I think. I guess one of the next steps would be to unify init() in apps/main.c?
19:44:45gevaertswebguest17: I think that's all. Make sure to provide your real name on the tracker
19:45:03kugelgevaerts: that could be done yes, but it's not very critical
19:45:08 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
19:45:41kugelI'd like to resync the pth work to the latest changes and submit that too
19:45:45webguest17ok, sounds good, just saw the message about the IRC chat on the tracker page and figured I should check in here
19:45:47gevaertsI think it should be done at some point, but you're right, no hurry
19:47:24 Quit stoffel (Remote host closed the connection)
19:49:32 Quit webguest17 (Quit: CGI:IRC (EOF))
19:50:03amr1kugel: I couldn't figure out a function that seems non standard specific to diacritics in lcd-bitmap-common, and idea ?
19:50:50***Saving seen data "./dancer.seen"
19:55:04funmanranma: i think reading ENRD has an effect on clearing interrupts
19:59:49 Join voRia [0] (
19:59:52ranmaYes, that's true.
20:01:39funmanif i add a read of ENRD2 at the end of ascodec_init() then the code works as expected on fuzev2
20:01:47 Part voRia ("Leaving.")
20:01:59funmanbut after backlight goes off it doesn't work
20:02:08ranmaInterrupt requests in ENRD[012] are cleared on read of the corresponding register.
20:02:51ranmaBetter designs have a separate interrupt_clear register instead of having the clear as a side-effect of read...
20:04:14funmanwell the interrupt just stop to happen and i should see a lot (for adc)
20:04:24ranmaHmm, I thought that was specified in the Datasheet, but apparently not.
20:04:41ranmaI guess I must have found that out while writing the as3525 audio_irq code...
20:05:13funmanstart -> dbg menu to see the counters of i2c -> can see it moving and stop at 11833 int_audio_ctr & 7 int_adc
20:05:22ranmaTo be sure to have cleared all interrupt sources you should read all 3 ENRD-Registers
20:06:04funmanwe don't use ENRD1 though
20:06:50ranmaThat's why my INT_AUDIO handler does a 3-byte read starting with ENRD0
20:07:00 Quit ssorgatem (Read error: Connection reset by peer)
20:07:16 Join ssorgatem [0] (~ssorgatem@
20:07:54ranmaIf you do single-byte reads for ENRD0 and ENRD2, you have to send the address twice
20:08:07ranmaSo in the end the three-byte read is not slower.
20:09:05ranmaADDR DATA ADDR DATA vs ADDR DATA DATA DATA should be pretty much take the same time
20:09:25 Join ssorgatem_ [0] (
20:10:39funmanhm but why interrupts stop all of a sudden ?
20:11:52ranmaMaybe check that no one else has overwritten ENRD?
20:12:03ranmaThat's why I had to change the charger control code.
20:12:15funmansupposedly it uses the same code
20:13:02 Quit ssorgatem (Ping timeout: 265 seconds)
20:13:03funmanhm but we use the special PMU registers on v2, it might have an effect
20:13:35 Quit JohannesSM64 (Quit: WeeChat 0.3.3-dev)
20:14:36 Join cr [0] (~cr@
20:15:16ranmaYou could try if it's lowactive instead of highactive on AMSv2...
20:15:41funmanbut in 10s i see like 20000 INT_AUDIO interrupts
20:16:55funmanif i comment IRQ_HIGHACTIVE it's teh same
20:16:57ranmaSo, you see INT_AUDIO interrupts, but only the adc finished path is not taken?
20:17:21funmannope it's taken 10 or 15 times
20:18:58funmanactually it seems to stop when i enter debug menu :?
20:19:10funman(view HW info)
20:20:01ranmaThen I can only assume something is writing to ENRD2 and setting bit0 to 0, disabling the interrupt source.
20:20:15ranmaOr the bitfield meanings are different.
20:21:51funmanoh i see a suspicious adc_read()
20:23:04 Join saratoga [0] (~463f90ed@gateway/web/freenode/x-muiafugysxcsulse)
20:23:15funmanbtw, thousands of INT_AUDIO interrupts per second doesn't look normal, does it ?
20:23:40 Quit Strife89 (Quit: Reboot.)
20:24:07ranmaHmm, good question.
20:24:53funmanUSB detection works fine (didn't try charger or rtc)
20:25:02 Join benji13 [0] (~benji13@
20:25:56ranmaYou could try just disabling the adc interrupt and wait.
20:26:20funmanthat's what i'm doing to avoid deadlocks when the INT_AUDIO stop
20:26:51ranmaDid you also disable the ADC bit in ENRD2? And it's still generating thousands of INT_AUDIO?
20:28:07funman~2000 per second
20:29:04ranmaDo you have a patch for debug info in the menu?
20:29:21pixelmakugel: I believe I could use a slightly newer gcc if that could make a difference? My cygwin install is quite old and I onbly updated a few tex packages maybe
20:31:07ranmaHmm, my current build has debug enabled, so I should easily get at the stats on my AMSv1...
20:32:10ranma30 int_audio in 10s
20:32:22 Part Trou ("Leaving")
20:34:30 Quit benji13 (Quit: Leaving)
20:36:51 Join Xerion [0] (
20:37:04funmanwriting 0 to all 3 registers & all combinations of pushpull/highactive -> same effect
20:37:43funmanit wants to tell us something
20:40:15ranmaSo, maybe it's connected to a different irq?
20:40:33 Join arbingordon [0] (~w@unaffiliated/arbingordon)
20:40:35funmanyou mean there's another unrelated condition which triggers it ?
20:42:14 Join DataGhost [0] (
20:42:14 Quit DataGhost (Changing host)
20:42:14 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
20:42:15ranmaI mean, maybe the audio block interrupt line is not connected to interrupt source 9 on the VIC
20:42:31 Quit kramer3d (Quit: This computer has gone to sleep)
20:42:37ranmaInstead source 9 is connected to something else that generates your 20000 ints per second.
20:43:31ranmaSo effectively you're polling the ENRD* status very often instead of reacting to the proper interrupt.
20:44:18ranmaAt that rate you may call the next async i2c read while the last is still unfinished.
20:44:31funmanaudio9 is used for enabling usb at least
20:44:45ranmaThat would screw up the reading and explain why you suddenly don't get the ADC notices anymore
20:46:00funmandoes the OF use interrupts?
20:46:06funmanon amsv1
20:46:14ranmaIIRC not
20:46:47 Quit cr (Quit: It's time to live and it's time to die!)
20:46:49funmandon't see it in the isr table
20:47:33funmanclipv2 use the same table than clipv1, but it has the interrupt 9 (supposedly audio) too
20:49:02 Join ssorgatem__ [0] (
20:49:37funmani'm gonna try the unused ones
20:49:45 Quit ssorgatem_ (Ping timeout: 246 seconds)
20:50:19ranmaAh, wait. I'm used VIC_INT_EN_CLEAR = INTERRUPT_AUDIO;, so async_read can't get called again until the read is finished.
20:50:35 Join Strife89 [0] (
20:51:21ranmaMaybe there's some ENRD3? Something else that uses the AUDiO_INT and is not acked so you have a 'screaming' interrupt line.
20:51:23 Join benji13 [0] (~benji13@
20:51:41ranmaAnd it's 'only' 2000 irqs per second because it's limited by the i2c read spead.
20:52:24 Nick [CGL]_ZZZzzzZZZ is now known as [CGL] (~CGL@
20:56:21funmanive found some code which reads IRQ_ENRD2 and set bit 0
20:59:46ranmaTo verify it's source 9, you could mask the irq, read the vic raw irq status, trigger it using an adc read, wait a little, read the vic raw irq status again.
21:00:24funmani see
21:01:03ranmaMaybe insert a 'read ENRD*' before the first 'read the vic raw irq status'
21:01:05funmanbut supposedly if we have continuous interrupts the condition is always present ?
21:01:32kugelfunman: hm, the scrollwheel works OK now, still not optimal though
21:01:42funmankugel: i think i borked repeat
21:01:43kugelvery fast scrolling is now impossible it seems
21:02:05funmankugel: try with 1 or 2 ms ?
21:03:02ranmaWell if you see some other masked irq in the raw status getting triggered by it you'd see it's most likely that one instead of 9
21:03:13kugelI think it needs more thinking. the e200v1 irq works well
21:03:42kugelthe difference is that you can scroll faster because the wheel is bigger size-wise, and that the fuze's wheel triggers twice as often
21:03:50 Join mt [0] (~mtee@
21:03:55 Quit mt (Changing host)
21:03:55 Join mt [0] (~mtee@rockbox/developer/mt)
21:03:56ranmaBut ideally you'd do the check somewhere early in the startup, so if something later triggers the stuck interrupt it's not stuck yet.
21:04:23kugelI'll have a look in a gsoc-free minute, not too early though :)
21:05:47kugelhm, since when does fm radio work? i missed something
21:06:39funmankugel: not sure if it works fine there were some bug reports when i added it (1 week ago or so)
21:07:44Luca_Slast time I tried, a couple of days ago, it worked great. even preset scanning :)
21:08:16funmani supposed it got fixed by a random commit
21:09:23funmanranma: it's the good one: before ascodec_init(), vic raw intr = 0, after, = 1<<9
21:09:49 Quit Luca_S (Quit: CGI:IRC)
21:11:02ranmaSo the question is why acking the IRQ doesn't work. :)
21:11:17funmanwe're only 'acking' the vic
21:11:24funmanah no we're not
21:11:55funmanwell it should be automatic with the vic
21:12:00funmanwith vector mdoe at least
21:12:21ranmaNo, the ENRD[0-2]-read should effectively be the ACK
21:12:38funmanon the as3517 of as3525
21:12:43funmanabout as3543, ..
21:13:14funman*you are now aware that we have some linux source code for as3543*
21:14:06funman#define AFE_AS3543_ID (0x27)
21:14:14funmanhu ho
21:14:21pamauryis there a define using as3543 ?
21:14:38funmanpamaury: in audio code only
21:15:15ranmaHmm, what about the AFE_PMU_ENABLE register muxing?
21:15:22funman0x27 shouldn't be an ID, so perhaps we're reading the wrong registers
21:17:15funmanit's used with registers 0x17 - 0x1B
21:17:55funman1B = backlight, the others = ?
21:19:13funman1A = AFE_PLL
21:19:38 Join b0hoon [0] (
21:20:32ranmaThe ascodec_read to ack the IRQ should make sure to switch to bank0 first (write 0 to AFE_PMU_ENABLE)
21:20:51 Join B4gder [0] (~daniel@rockbox/developer/bagder)
21:21:24funmanirq are disabled when using AFE_PMU_ENABLE but i didn't think a reset was required
21:21:34ranmae.g. the linux as3543_write/read is a wrapper to write/read_mux with bank=0
21:22:41funmanadding 'ascodec_write(AS3543_PMU_ENABLE, 0);' at the end of ascodec_write_pmu() -> same
21:24:29CIA-5New commit by 03b0hoon (r26061): Promote Packard Bell Vibe 500 to stable.
21:25:43funmanAMS website show as3518 & as3542 but no available datasheets
21:26:06funmanb0hoon: congrats for this port
21:26:22b0hoonfunman: thank you :)
21:27:54 Join CFP [0] (
21:27:59 Join TFGBD [0] (
21:28:04b0hoonfunman: but it's nothing comparing to your work on sandisk's...
21:28:13 Quit CFP (Client Quit)
21:28:13funmanah nevermind the 0x27 is a content, not an index
21:28:58funmanb0hoon: i think i spend wayyyyyyyyyyyy too much time on this
21:29:08TFGBDIs there an exe based rockbox bootstrap for Portable Media Center Windows CE based Platforms like the Zune?
21:29:54TFGBDI'm sure you are aware the Zunes can now run any unsigned exe, right
21:30:19TFGBDgood stuff
21:30:37funmanno one showed up with a patch for making rockbox run on zune though, we're all waiting ;)
21:30:55kugeljhMikeS: "[0 ... AUDIOHW_EQ_BAND_NUM-1] = -1" is that a gcc extension?
21:31:10TFGBDI'd imagine the Gigabeat PMC version would run without modification on the first models
21:31:30saratogabuttons are probably different
21:31:35funmanthe 0x38 'UID0' register reads as 0xA8 on fuzev2, not 0x27 like linux code suggests
21:31:37ranmaHuh, as353x has mp3 hw decoding?
21:31:45saratoganot as far as we know
21:31:48TFGBDI guess someone just needs to write that magic .exe
21:31:56CIA-5New commit by 03b0hoon (r26062): front page: move the PB Vibe 500 to the stable category.
21:32:10funmanranma: some have h264 decoding too, but we're not using the as353x soc, just parts of it
21:32:41funmanranma: i have the as3531 datasheet (the one without h264) but there's nothing to be seen there
21:33:41funmanooooh perhaps not
21:33:52funmanirqenrd_1 = 0x24
21:35:00funmanirqenrd_1 = 0x25, according to another page of the datasheet :/
21:37:05ranmaWell, you could write a 0 to it in ascodec_init(), just to be safe :)
21:37:18*b0hoon supposes that the front page won't change automatically, hm.
21:37:34b0hoonBagder: ping
21:38:13funmanranma: writing 0 to 0x23 & 0x24 + reading -> same :/
21:38:15b0hoonor... B4gder: ping
21:38:41funmanb0hoon: you should change status in tools/ too
21:38:43*B4gder pretends he's not here
21:39:05b0hoonB4gder: hi. why? :D
21:39:23B4gderbecause I'm lazy and scared of having to do actual work!
21:40:59funmanone can not argue with that :)
21:41:19b0hoonB4gder: ok, so may i ask you in your free time to update the front page, please?
21:41:26B4gderjust did it
21:41:57b0hoonfunman: have done it already
21:42:07*ranma needs some sleep now, oyasumi...
21:42:16funmanranma: btw, if we make usb work, we'll need new bootloaders i think
21:42:26funmanaligato ranma
21:42:41b0hoonB4gder: thanks alot
21:43:08funmanwriting 0xff to 0x23 / 0x24 has an effect: interrupts stopped quickly, next step: try each bit
21:43:50 Quit saratoga (Quit: Page closed)
21:44:13ranmafunman: Well, it's semi-working already :)
21:44:56CIA-5New commit by 03funman (r26063): adc-as3514.c: cosmetics ...
21:47:02mtthe ffmpeg decoder stored the decoded samples in floating point format ( line 1403), the sim can't work with that, can it ? (I'm getting too much noise with the output, sort of like the noise you get from applying a ridiculous gain on the signal)
21:48:33kugelmt: I don't know if sdl can, but my alsa sim work should be able to handle it
21:50:54***Saving seen data "./dancer.seen"
21:51:09funmanmt: look in uisimulator/sdl/sound.c:315 if you can change the format
21:52:18mtkugel: Is the alsa sim stuff still in your git repo or in svn ?
21:52:23funmanit's signed 16 bits system order -> little endian on x86 ?
21:52:28mtfunman: ok thanks, will look there
21:52:51kugelmt: in the git repo, and not up to date but I can have a look at updating
21:52:59funmanin SDL_audio.h you can only select byte order, signed/unsigned and 8/16 bits
21:54:07kugelalso the most recent commit removed that ability because I thought it's useless :p
21:54:25mtwell it sort of is
21:54:53mtI could commit the code as is right now, not caring about that noise, since I should get rid of the floating point stuff anyway
21:58:05 Quit bluebrother (Disconnected by services)
21:58:08 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
21:58:21 Quit bieber (Ping timeout: 246 seconds)
22:02:07funmanmt: try to dump samples in a file and ffmpeg -i foo -acodec pcm_s16le out.wav?
22:04:35 Quit funman (Quit: free(random());)
22:05:10 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
22:25:25 Quit B4gder (Quit: It is time to say moo)
22:25:58CIA-5New commit by 03jethead71 (r26064): YUV Dither: r12 saving was removed but stacked parameter load offset wasn't changed to compensate, resulting in an improperly aligned dither kernel.
22:27:38 Quit kugel (Remote host closed the connection)
22:30:57 Join digitalz [0] (
22:35:37*digitalz Discounts!! Our Special Limited Time Offers Up To May,22!!!New BranD!! Notebooks,Plasma and LCD TV's.Buy your electronic needs at our unique prices. Laptop Sony VAIO VGN-FW590FFD-575,57$!!!Apple MacBook Air MC234LL/A-695,27$!!!
22:35:58 Quit digitalz (K-Lined)
22:37:02 Join kugel [0] (~kugel@rockbox/developer/kugel)
22:42:18*kugel is going the sim-target-tree in a few seconds
22:42:35 Join liar [0] (
22:52:35 Quit Jaykay (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
22:54:22 Part b0hoon ("GTG. Bye.")
22:54:24 Join robin0800 [0] (
22:57:00 Quit TheSeven (Ping timeout: 276 seconds)
22:59:08 Quit DataGhost (Ping timeout: 240 seconds)
23:00:21 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
23:00:27 Quit liar (Quit: Verlassend)
23:00:52 Join saratoga [0] (~9803c6dd@gateway/web/freenode/x-hlzjjzgdsdosrljp)
23:02:50CIA-5New commit by 03kugel (r26065): - Move uisimulator/sdl/*.[ch] into the target tree, under firmware/target/hosted/sdl, uisdl.c is split up across button-sdl.c and system-sdl.c. ...
23:04:51amiconnjhMikeS: r260055 breaks archos
23:05:37amiconnDevices which don't have HAVE_POWEROFF_WHILE_CHARGING defined are *physically* incabale of being powered off while connected to the charger
23:06:13amiconnThat means a recorder will _hard freeze_ when it's done charging now (and no actively used)
23:07:10 Quit benji13 (Quit: Leaving)
23:11:21CIA-5New commit by 03kugel (r26066): Get checkwps going again, don't mess with sdl in it.
23:12:16 Quit intrados (Ping timeout: 264 seconds)
23:12:22kugelamiconn: not only archos I think
23:13:38jhMikeSamiconn: I see.
23:15:13 Join intrados [0] (
23:15:17amiconnI think HAVE_POWEROFF_WHILE_CHARGING should be defined for the beast (because it actually can be powered off), and software controlled charging needs special handling in another way (another #define or perhaps a combined one with multiple possible values)
23:15:58jhMikeSyes, it's far more trouble than it's worth and quite annoying to see the splash just because it's plugged (which it still shows)
23:16:10amiconnThe Player and Recorder do have a charging screen
23:16:22jhMikeSif you try to power down/
23:16:34amiconnRight now it's only used when rockbox is flashed, and the box is plugged in
23:16:48Torneipods will just reboot also
23:17:08Tornei just came on to point that out but amiconn beat me to it ik see
23:17:09amiconnOne idea to extend it was to make "poweroff" go to the charging screen if the user tries to power off on targets which can't
23:17:33amiconnIt just hasn't been done yet...
23:17:49jhMikeSI wanted to do something like that, then once it's finished, power off
23:17:54 Quit robin0800 (Quit: No Ping reply in 180 seconds.)
23:18:20 Join robin0800 [0] (
23:22:16Torneare we really goinng to allow ape tags on mp3?
23:22:49saratogawe should probably discuss it
23:23:11AlexPTorne: The majority of people today have said no
23:23:19gevaertsI'd say no
23:23:21saratogaas currently implemented it looks pretty harmless on SWCODEC at least
23:23:28AlexPLast time it was discussed it came out as no I think
23:23:32gevaerts(unless we also support vorbis comments on SID)
23:23:33 Join fml [0] (
23:23:46Torneit means the files are corrupt, no?
23:23:56Torneape tags aren't properly desynched
23:24:13Torneso they can be interpreted as broken mpeg frames
23:24:16amiconnAPE tags aren't native to mp2/mp3
23:24:26AlexPMy main beef is just that it wasn't discussed at all, and done for hwcodec too
23:24:34Torneyah, i mean from the decoder's pov
23:24:53amiconnThe consensus in the past was that rockbox should support the native tagging format(s) of every audio format it supports, but not more
23:25:29saratogamp3 is a special case though since it has no native format, 2 widely used ones, and a third less common one
23:25:59Torneid3 is designed to be ignorable by decoders though, ape is not
23:26:13Torneso its presence means the file is observably corrupt
23:26:25saratogabecause it takes care not to put a sync word in the tag somewhere?
23:26:38fmlHello. There has been some discussion about FS #11271 (font with digits only). We came to the conclusion that it should not be part of the font pack but rather be part of a theme bundle. But if someone would install such theme, the font would be copied to the font folder, and we'd have the same situation as we had if the font was in the font pack.
23:26:48Torneyes, great care in fact
23:26:49amiconnI'd consider id3v1 and id3v2 the native tagging formats for mp2/mp3
23:26:56AlexPfml: That is a good point
23:27:12AlexPfml: The codepage idea sounded a good one to me
23:27:24CIA-5New commit by 03jethead71 (r26067): Revert r26055 since it breaks certain Archos targets.
23:27:32Lloreanamiconn: Maybe not native, but at least a standard in practice.
23:27:40saratogais avoiding a sync word really a problem for a tag that can only go at the end of the file?
23:28:06saratogai though id3v2 needed them since it has to be at the start of the file for streaming, while id3v1 didn't since its always the end
23:28:50 Quit robin0800 (Quit: No Ping reply in 180 seconds.)
23:29:03fmlI mean there's no difference in how the font gets into the font directory. And hence there's no point in not letting the new font into the pack. The idea with the lang files telling what char areas they need and fonts telling what char areas they cover is of course very good.
23:29:15 Join robin0800 [0] (
23:29:38gevaertsAnyway, I think the current position is "native tags only", and if people want to change that they need to get a consensus *first*
23:29:44Tornesaratoga: files with ape tags can make bad noises at the end of playback on some dumb decoders. i have heard. :)
23:30:00Lloreanfml: The difference is that we don't provide fonts that can't be used to render the default language. If someone downloads a theme it's more or less their own responsibility - themes can also break the UI in plenty of other ways as well.
23:30:01 Quit TheSeven (Remote host closed the connection)
23:30:07AlexPgevaerts: The consensus earlier seemed to be revert, but he hasn't replied to mails
23:30:15amiconnLlorean: Well, if you define native as "being designed for that audio format", id3 *is* native - just implemented after the format itself, not together with it
23:30:23saratogaTorne: is our decoder dumb like that?
23:30:34 Join Buschel [0] (
23:30:47 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
23:30:58Tornesaratoga: no idea, and it could be fixed not to be if it was, on swcodec
23:31:09Lloreanamiconn: Ah. I tend to think of native as a two-way street but that's a good way of looking at it too.
23:31:12gevaertsWe should probably show a splash if we detect a corrupt file :)
23:31:18fml Llorean: wouldn't it then be sensible to keep the fonts that have been installed with a theme in a special ("private") folder?
23:31:22Lloreangevaerts: I think currently we just skip them.
23:31:22Tornea player that knows what ape is can obviously ignore it
23:31:30AlexPgevaerts: nah, silently crash :)
23:31:47Tornewe should do something for unplayable files, yeah
23:31:53gevaertsfml: That could be a good idea
23:32:03saratogaIIRC you walk an mp3 file one frame at a time looking at the sync words, so if the tag is at the end like id3v1 or apev2, the risk of detecting it is very small, since the sync word must appear, and it must appear at precisely the right location
23:32:06CIA-5New commit by 03jethead71 (r26068): Gigabeat S: Not defining HAVE_POWEROFF_WHILE_CHARGING is far too disruptive given that it's not obligatory for the hardware to avoid poweroff while ...
23:32:22Lloreanfml: It might actually be ideal to allow themes to keep their personal fonts in the same folder as their bitmaps (as themes with fonts using the same name can overwrite each others' which can be a problem if two people have converted the same font in different ways, or similar)
23:32:26Tornewe skip unplayably large aac files atm
23:32:26 Quit amr1 (Quit: CGI:IRC (EOF))
23:32:26Torneand this is a source of confusion to some peopple i've takled to
23:32:40saratogai'm sure it can be made to happen, but i'm skeptical a correctly parsing decoder will ever have it happen on a real file
23:33:11saratogaless that we skip them and more that the parser fails on them gracefully :)
23:33:21amiconnFeeding random data to the MAS (on hwcodec) often causes funny noises
23:33:29Tornesaratoga: it's only an *additional* argument against it
23:33:38LloreanTorne: Is there a reason we still don't support arbitrary length MP4? Is it a "can't" with the way we're set up or something?
23:33:49amiconn*Imo* we should detect and skip ape tags on mp3, but nothing more
23:33:57saratogawe don't know how to do it since none of us can understand the mp4 spec
23:34:05TorneLlorean: cant without changing the way wed use the seek index
23:34:11Tornei think
23:34:23LloreanTorne: Does that mean they could be playable without seeking (or without fast or accurate seeking)?
23:34:32Tornenot sure
23:34:32fmlgevaerts: Llorean: the problem with this would be (a) that some fonts would be stored many times and (b) that fonts are loaded automatically only if they are stored in the fonts folder, and we don't explicitly load themes on start. If we did it then we could store fonts anywhere and they would load correctly.
23:34:38saratogayeah i had a patch to do that a while ago, but we weren't sure it would work on all files
23:34:51Tornethe main use case is apple audio books though..
23:34:54saratogait basically stepped an mp4 file like an mp3, going from frame to frame without using the header to find them
23:35:03Lloreanfml: If they're WPS fonts (which is the only location it would make sense for) does the automatic loading actually come into effect?
23:35:21saratogai think the concern was that it might have been possible to put stuff in between frames in an mp4 file
23:35:22Tornefor which seeking is useful :)
23:35:43AlexPLlorean: So it would still have to put a custom ui font defined in the theme in fonts?
23:35:54LloreanAlexP: Huh?
23:35:57fmlLlorean: do you mean that the WPS specifies where the font is stored anyway?
23:36:16LloreanAlexP: Can we currently have more than one font for the list? I'm not too familiar with multifont.
23:36:22AlexPLlorean: A theme can define a custom UI font
23:36:23fmlLlorean: so that these fonts can be stored anywhere?
23:36:30AlexPA font that isn't in the fontpack
23:36:31saratogaif the main argument against apev2 support isn't how much we like the standard, nor how much it costs to implement, what exactly is the argument?
23:36:39gevaertsLlorean: theme!=wps
23:36:45AlexPSo wouldn't that need to be in fonts to be autoloaded?
23:37:18saratogayears ago i thought we concluded it wasn't worthwhile because it would be hard to work into our metadata parsing efficiently, but it seems that is not the case
23:38:21pixelmawasn't it also about potentially useless seeking?
23:38:25Lloreangevaerts, AlexP: Aren't fonts loaded via viewport definition?
23:38:35LloreanWhich means they're in a wps/sbs/whatever, which is manually loaded at boot.
23:38:43LloreanSo the only font that is "autoloaded" is the one in the .cfg?
23:38:47AlexPEven if there isn't a ui viewport used?
23:38:54gevaertsLlorean: the one in the .cfg comes with the theme...
23:39:04saratogapixelma: maybe, but that only happens if you have the apev2 tag and no id3 tag, so it hardly seems like a problem
23:39:44pixelmawhat if I have no tags at all?
23:40:09Lloreangevaerts: If it's a UI font, it should be okay to go in the main fonts folder. I didn't mean that *all* their fonts should go in a personal folder, just ones that aren't necessarily usable to switch to manually
23:40:29Lloreangevaerts: Basically, give them the option of using a personal folder when necessary.
23:40:44saratogait'll seek to the end of the file anyway if you have no tags (since the id3v1 tag could be there), i'm not sure if theres another seek or not
23:41:03gevaertsLlorean: ok, that could work
23:41:06Buschelhmm, I experience hard crashes with my iPod Video builds at an address that points to settings_list.
23:42:29Lloreangevaerts: It's not ideal, since sometimes they'll make all the fonts "private" except the main UI one when they don't need to be (and sometimes they'll stuff extra ones in the fonts folder that shouldn't be there) but it gives conscientious authors the tools they need, which is close to the best you can do unless you get "picky" loaders that reject fonts that are incompatible with the language (and vice versa)
23:43:13gevaertsLlorean: yes, it looks like a good compromise that's easy to implement
23:43:26*gevaerts only guesses at that last bit
23:43:30saratogapixelma: looking at the code, i think it'll seek 96 bytes forward from the current file position
23:44:35saratogai'm not sure what that does physically, since sectors are hundreds or thousands of bytes, probably it doesn't actually do anything to the disk, but i don't understand well enough to say for sure
23:44:44*kugel doesn't the ape problem, i.e. what's the reason against reading them?
23:45:17*gevaerts thinks that the main problem is that this was committed without any prior discussion
23:45:29saratogalike 10 years ago there was a big argument about id3v2 verses apev2 on the internet, ever since then people have had very strong feelings about file tags
23:45:46fmlLlorean: but then we don't need to change anything! It's just the matter of how authors organize their theme bundles.
23:46:02kugelwell, I can understand that someone doesn't see a discussion needed he/she didn't know about that general consensus to not read ape
23:46:09kugelwas it on the NoDo list?
23:46:11 Join ssorgatem [0] (~ssorgatem@
23:46:18gevaertskugel: yes, I just checked
23:46:26gevaertsWhat's the point of having a consensus if people are going to ignore it anyway?
23:47:02Lloreanfml: Well, the ideal would be to fix the problem in the language / font loaders so that nothing else matters.
23:47:32CIA-5New commit by 03alle (r26069): Format FM frequency depending on the regional settings (FS #11273)
23:47:33kugelgevaerts: ah, I didn't know it's on the NoDo list. then yes, he simply ignored it
23:47:46 Quit robin0800 (Quit: No Ping reply in 180 seconds.)
23:48:13 Join robin0800 [0] (
23:48:29saratogawell presumably he didn't realize it was on the list
23:48:47AlexPsurely reading the nodo list isn't too much to expect?
23:49:16fmlLlorean: that would be a safety check, but if all authors behave then we wouldn't actually need it
23:49:49Lloreanfml: There's no way to make authors behave though. Many of them won't even know they need to, probably.
23:49:56AlexPmind you, the one in the docs folder is not exactly up-to-date
23:50:09 Quit ssorgatem__ (Ping timeout: 276 seconds)
23:50:58***Saving seen data "./dancer.seen"
23:51:04AlexPIf he looked at the NODO in the source, then I can't blame him. That should be updated or removed
23:51:08fmlLlorean: sure! That's like everywhere in the real life! :-)
23:51:31saratogaanyone mind if I replace the one in the source with a link to the wiki?
23:51:43gevaertssaratoga: sounds like a good idea
23:52:05 Quit merbanan (Ping timeout: 246 seconds)
23:52:15gevaertsAlexP: maybe, but IMHO we *can* blame him for not discussing this at all
23:52:26AlexPAnd it isn't the first time
23:52:31Tornehe hasn't... discussed other things either
23:52:35Tornein the past ;)
23:53:44AlexPGiven that NODO discussion is on the devcon list, what about this now then?
23:54:15Torneyah, it can always go back in again if decided that way..
23:54:26CIA-5New commit by 03saratoga (r26070): Remove outdated NODO list from the docs folder. Replace it with a link to the wiki.
23:54:29AlexPI say revert for now (given it is on the NODO), then discuss at Devcon and re-add depending on decission
23:54:29Tornethe longer it sits there the more likely any mp3-with-ape-using users will expect it ;)
23:54:42saratogaat very least revert the HWCODEC changes, since those don't make much sense IMO
23:55:21pixelmanow there's some discussion about themes and fonts, can I plug my question about the "Now playing" in cabbiev2 again? My idea was to make it a text (and then language) string so it is more flexible for further use and translatable to some extent. The latter refers to the problem that it is only possible in a smallish font which likely won't support many non-latin characters. Would it be acceptable to still make a lang string out of it and put a "Use
23:55:21pixelma English if your translation can't be displayed with the smallish font"? That'll give us a bit more localisation than there is now and as I said flexibilty for e.g. the FM screen and less bitmaps
23:56:01fmlA question to all C experts: how can I ensure that an int value is in the range 0..N-1? val %= N; if (val <0) val += N;? It's from skin_tokens.c:417
23:56:13AlexPpixelma: Sure, localisation is always good
23:57:21CIA-5New commit by 03gevaerts (r26071): Revert r26048. APE tags in mp3 is explicitely on ...
23:58:05pixelmahave to recheck which font that could be - and until there is a way for "inverted" text, it'll probably stay a bitmap on monochrome screens
23:58:31pixelmaso it feels a bit hackish

Previous day | Next day