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 2010-05-12

00:03:02 Quit amiconn (Disconnected by services)
00:03:04 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
00:03:08 Quit pixelma (Disconnected by services)
00:03:08 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
00:03:26 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
00:03:27 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
00:04:30smumeysorry new to this client−−meant to search
00:06:22 Quit efyx (Quit: Quitte)
00:15:38 Join mikroflops [0] (
00:16:32 Quit jgarvey (Quit: Leaving)
00:20:09 Quit mikroflops_ (Ping timeout: 268 seconds)
00:23:45CIA-5New commit by 03nls (r25959): Faster assembler strlen for coldfire using the load-a-whole-word-and-test-i-for-nullbytes-at-one trick, benched 28% faster than the old version
00:24:10n1syay for typos
00:26:31 Quit bluebro ()
00:26:33 Quit DerPapst (Quit: Leaving.)
00:27:11 Quit joeyg (Quit: msg: ?)
00:29:47 Join hebz0rl [0] (
00:38:35 Quit n1s (Quit: Lämnar)
00:50:16 Part smumey ("Konversation terminated!")
00:52:40 Join apoelstra [0] (
00:52:43 Nick apoelstra is now known as joeyg (
00:52:50 Join phanboy4 [0] (~benji@
00:52:57 Join aberet [0] (
01:03:35 Quit ender` (Quit: In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.-- Douglas Adams)
01:07:43 Quit solexx (Ping timeout: 246 seconds)
01:09:25 Join solexx [0] (
01:09:31 Quit DataGhost (Ping timeout: 240 seconds)
01:11:29 Quit Stephen__ (Quit: Leaving)
01:12:16 Quit pixelma (Ping timeout: 246 seconds)
01:12:16 Quit amiconn (Ping timeout: 246 seconds)
01:21:51 Join Darkknight512 [0] (
01:28:53 Quit phanboy4 (Quit: Leaving)
01:28:56 Join amiconn [0] (quassel@rockbox/developer/amiconn)
01:29:22 Join pixelma [0] (quassel@rockbox/staff/pixelma)
01:37:03 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
01:44:56 Join Soap-N810 [0] (
01:47:47 Part toffe82
01:48:57***Saving seen data "./dancer.seen"
01:54:47 Nick aberet is now known as kcalder (
01:56:43 Quit M3DLG (Ping timeout: 240 seconds)
01:56:45 Nick kcalder is now known as aberet (
01:56:57 Join M3DLG [0] (
02:07:42 Quit amiconn (Remote host closed the connection)
02:07:43 Quit pixelma (Remote host closed the connection)
02:08:20 Quit Battousai (Remote host closed the connection)
02:08:37 Quit kramer3d (Ping timeout: 246 seconds)
02:09:25 Join pixelma [0] (quassel@rockbox/staff/pixelma)
02:09:30 Join amiconn [0] (quassel@rockbox/developer/amiconn)
02:10:50 Quit M3DLG (Quit: RAGE QUIT)
02:15:49 Join Stephen__ [0] (~S@
02:19:01 Quit Stephen__ (Client Quit)
02:20:20 Join Battousai [0] (~bryan@gentoo/developer/battousai)
02:22:49 Join Rob2223 [0] (
02:26:14 Quit Rob2222 (Ping timeout: 245 seconds)
02:28:13 Quit Soap-N810 (Ping timeout: 252 seconds)
02:38:49 Join BHSPitMonkey [0] (
02:38:51 Quit BHSPitMonkey (Changing host)
02:38:51 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
02:44:17 Quit jae (Quit: leaving)
02:45:20 Join jae [0] (
02:52:47 Join Forsaken [0] (~chatzilla@
02:58:07kisakwe should have all the elements needed to include the Fuze v2 into the manual install section of
02:59:56kisakany problem with me adding it?
03:03:09 Join hebz0rl_ [0] (
03:04:19 Quit hebz0rl (Read error: Operation timed out)
03:06:49 Join ischeria1 [0] (
03:10:27 Quit ischeriad (Ping timeout: 264 seconds)
03:49:01***Saving seen data "./dancer.seen"
04:00:19 Join Strife89|Tablet [0] (
04:03:14 Quit Strife89 (Ping timeout: 240 seconds)
04:08:25 Quit pixelma (Disconnected by services)
04:08:25 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:08:30 Quit amiconn (Disconnected by services)
04:08:31 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:08:45 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:08:53 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:12:40JdGordonbieber: hey
04:13:19JdGordonI created a branch last night for the changes and got almost all the tags converted
04:13:43bieberYou changed the existing theme engine to parse them?
04:14:03 Join Strife89 [0] (
04:14:08 Quit shaggy-h (Read error: Connection reset by peer)
04:14:24bieberSo I can pretty well assume that the changes will get merged into the main build?
04:14:48JdGordonthe timing is the only unknown
04:15:50bieberSo if you're just going with the existing infrastructure, I'll go ahead and keep on with my original plan and build the Theme Editor from the ground up to use C++ and Qt
04:15:51 Join intrados [0] (
04:16:21JdGordonno, I was hoping to replace the current parser
04:16:28bieberOh, okay
04:16:31JdGordonfixing the existing one to work was simple
04:16:52JdGordonI wanted to do that so your parser has something to compare against?
04:17:02bieberIn that case, I was wondering if there's a wiki doc out there somewhere with code formatting standards for the project?
04:17:23JdGordondocs/CONTRIBUTING in the sources
04:18:26bieberI guess if you're going to use my parser, I'll have to build the parse tree in a static buffer? I was just reading about the no-malloc() rule
04:19:46JdGordonno, malloc is fine if we can replace it in the core
04:20:00JdGordondont rely on free and it can work
04:21:36JdGordondont worry too much about integrating it.. your project is for the editor.. once you have the parser I can help integrate it
04:21:55bieberGot it
04:22:01JdGordonjust keep the parsing and handling seperate :)
04:25:36bieberI've got to run for now, I'm going to read through that formatting doc and get going on the parser tomorrow. The syntax parsing at least should be done pretty quickly
04:26:49bieberVerification will take a while, but that's also something that a lot of other people can help out on pretty easily if anyone's up for it, since it's just a matter of putting together the parameter lists
04:37:28 Join Buschel [0] (
04:49:13 Quit Buschel (Ping timeout: 240 seconds)
04:55:40 Quit adnyxo (Ping timeout: 246 seconds)
04:56:28 Join Barahir [0] (
04:57:00 Quit Barahir_ (Read error: Operation timed out)
04:58:06 Quit joeyg (Quit: WeeChat 0.3.0)
05:02:26 Join apoelstra [0] (
05:02:26 Nick apoelstra is now known as joeyg (
05:04:06 Quit Darkknight512 (Quit: ChatZilla 0.9.86 [Firefox 3.5.9/20100315083431])
05:07:33 Part jd ("Ω")
05:21:12 Join Boldfilter [0] (
05:31:47 Quit intrados (Quit: WeeChat 0.3.2)
05:32:44 Quit joeyg (Quit: WeeChat 0.3.0)
05:37:00 Quit Forsaken (Ping timeout: 246 seconds)
05:37:52 Quit Strife89|Tablet (Quit: Leaving)
05:38:19 Quit Horscht (Quit: Verlassend)
05:43:50 Join CGL [0] (~CGL@
05:49:02***Saving seen data "./dancer.seen"
05:58:19 Join grndslm [0] (
06:14:57 Quit shai (Quit: Leaving)
06:25:17 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
06:25:44 Join cdb [0] (
06:31:34jhMikeSwhy the heck is sound_val2phys only called if theres a number of decimals specified? that makes no sense. it should always be called for physical units. :\
06:39:34 Quit Boldfilter (Quit: Boldfilter)
06:47:07 Join shai [0] (
06:47:25 Join Forsaken [0] (~chatzilla@
06:52:10jhMikeSafter looking at every driver, it looks safe to change it around anyway
06:59:13 Quit Battousai (Read error: Operation timed out)
06:59:15 Join hd [0] (
06:59:15 Quit hd (Changing host)
06:59:15 Join hd [0] (~jd@Wikipedia/HellDragon)
06:59:15 Join Battousai [0] (~bryan@gentoo/developer/battousai)
07:08:37S_a_i_n_tJdGordon: I had a look at what you have so-far in the wps from 25954, readability is definitely improved. I must say, it will take me some time to get used to thinking of syntaxes this new way, but at least it won't be a huge effort converting a theme to use the newer syntax. ;)
07:08:58S_a_i_n_tI *think*, well...from what I've seen to date.
07:09:17JdGordonit does add a bit to line lengths though
07:10:17S_a_i_n_tHehehe, I'm somewhat used to that though...I have some *really* long lines in a few of my .sbs/.wps'es
07:10:29 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
07:11:10S_a_i_n_tit doesn't look as weird as I initially thought it would.
07:22:59 Join LinusN [0] (~linus@rockbox/developer/LinusN)
07:23:04CIA-5New commit by 03jethead71 (r25960): Have option_get_valuestring always call sound_val2phys before formatting rather than only if fractional places are specified.
07:23:28 Join Gump [0] (
07:23:55Gumpanyone here familiar with the sansa fuze?
07:23:56 Quit GodEater (Read error: Operation timed out)
07:23:59 Join GodEater [0] (
07:23:59 Quit GodEater (Changing host)
07:23:59 Join GodEater [0] (~bibble@rockbox/staff/GodEater)
07:24:11Gumpi installed rockbox and used it successfully for about 2 weeks
07:24:25Gumpand suddenly my device became "read only" so i can't add or remove any files
07:25:08linuxstbGump: Are you using Linux?
07:26:05 Quit Strife89 (Quit: Bed.)
07:27:38ForsakenI installed r25863 on a fuze v2 and no prob so far
07:40:55 Join stoffel [0] (
07:41:42 Quit stoffel (Remote host closed the connection)
07:43:48 Join stoffel [0] (
07:49:03***Saving seen data "./dancer.seen"
07:53:13 Join B4gder [0] (~daniel@rockbox/developer/bagder)
07:59:21 Join n1s [0] (~n1s@rockbox/developer/n1s)
08:04:03 Join Buschel [0] (
08:09:35 Join kugel [0] (~kugel@rockbox/developer/kugel)
08:16:00Gumplinuxstb: yes, ubuntu
08:16:09Gumpi've read about similar issues, but no solutions
08:18:34n1sGump: linux will mount a FS as read only if it detects errors, try running fsck.vfat on it
08:20:29 Join animalstyle [0] (
08:20:38 Quit BHSPitMonkey (Remote host closed the connection)
08:20:39 Quit animalstyle (Client Quit)
08:22:14 Join animalstyle [0] (
08:24:11 Quit animalstyle (Client Quit)
08:24:59 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
08:25:19 Join ender` [0] (
08:25:20 Quit Buschel ()
08:35:02 Quit liar (Ping timeout: 258 seconds)
08:37:54 Quit jhMikeS (Ping timeout: 252 seconds)
08:38:28 Join jhMikeS [0] (
08:38:29 Quit jhMikeS (Changing host)
08:38:29 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
08:41:35n1sI get connection refused when trying to svn up
08:42:36 Quit jhMikeS (Read error: Connection reset by peer)
08:43:01 Join jhMikeS [0] (
08:43:02 Quit jhMikeS (Changing host)
08:43:02 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
08:44:00*B4gder does the magic poke
08:44:18B4gderabra kadabra is the swedish term ;-)
08:44:31CIA-5New commit by 03jethead71 (r25961): Fix F_T_SOUND voicing for decimals and physical values.
08:44:51jhMikeSB4gder: magic poke?
08:45:04jhMikeSis that why my commit was hanging up?
08:45:04kugeljhMikeS: hey!
08:45:20B4gderor rather, my poke made it unhang
08:45:33*jhMikeS was going a bit batty but is happy it worked now :) *thumbs up*
08:45:44kugeljhMikeS: do you know the reason why in the sim an extra sdl thread is created for the rockbox' main thread?
08:45:52jhMikeSkugel: I saw you rang earlier?
08:46:34jhMikeSkugel: it immitates the interrupt routine
08:46:38 Quit mc2739 (Ping timeout: 246 seconds)
08:46:39 Quit hebz0rl_ (Quit: Ex-Chat)
08:47:11kugeljhMikeS: the implicit main thread does
08:47:41jhMikeSlemme look again. Is it a *rockbox* thread or just a plain sdl one?
08:48:07kugelI'm about to change that for RaaA, i.e. use the implicit main thread for the rockbox' main thread. I currently went back to button polling but I think the interrupt-like fashion can stay if needed
08:48:20 Join Doleo [0] (
08:48:39 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
08:48:58Unhelpfulkugel: why change it? just no need for interrupt simulation in RaaA?
08:49:25kugelUnhelpful: there is interrupt simulation, I just changed the button reading
08:49:29jhMikeSwhich file?
08:49:47kugeluisdl.c & thread-sdl.c
08:50:33jhMikeSthere is a thread is also is the "main" thread and it stay independent from the sdl main thread
08:50:33kugelUnhelpful: but yes, I think there's no real need for that ;)
08:50:58kugelyes, I'm refering to that one with rockbox' main thread
08:51:07kugeland the sdl main thread is the implicit main thread
08:51:44jhMikeSit must be in the pool of rockbox threads. putting the main one there for int simulation would prevent it from working as an async button source
08:52:41kugelI think it must *not* be a rockbox thread?
08:53:11kugelI'm not sure I understand it :)
08:53:44 Join wincent_balin [0] (
08:53:56kugelbtw, when I say I changed something I mean I did it in my git branch kugel-rb.git/shortlog/refs/heads/sim-target-tree">
08:55:32jhMikeSthe one created by the threading during *is* a rockbox thread, "Main". the sdl thread after creating it, jumps into the button loop and serves much like a hardware source of events
08:55:57jhMikeSs/during/during init
08:56:01 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
08:56:09kugelI need that changed
08:57:23kugelI need to sdl main thread as the rockbox main thread. I think the old main thread can survive but for simplicity I changed to button polling
08:57:30 Quit antil33t ()
08:57:47kugelthe old main thread = the one that enters the button loop
08:58:34kugelis there anything preventing that?
08:59:10kugelmy changes work so far, but I'm maybe not aware of all the implications the current mechanism has
09:01:58jhMikeSthe main thread is the sdl thread application thread. the "Main" thread is just another rockbox thread like any other created in the scheduler. how can "Main" block and wait for enqueued events as expected then?
09:03:10jhMikeS*sdl application thread* not "sdl thread application thread"
09:03:50kugelI integerate the sdl main thread into the rockbox' threads array
09:04:08kugelwith the only difference that thread->context.t is NULL
09:04:20 Join esperegu [0] (~quassel@
09:04:23kugelbut I create a semaphore for it so it can be blocked
09:04:27jhMikeScan it service sdl tasks as well without requiring special casing
09:05:06kugelI've not experienced any problems. what do you mean with sdl tasks?
09:07:00 Join petur [0] (~petur@rockbox/developer/petur)
09:07:31jhMikeSI guess it's basically the button thing. who wakes up the thread when its blocked? I suppose I'll have to look at your repository to really comment further.
09:08:13kugelnothing wakes it up right now, as I said I went back to polling
09:08:39 Join flydutch [0] (
09:08:53jhMikeSkugel: ah right, duh
09:08:59kugelthe button thing relies on preemption, which does not work if pth-threads are used (either if sdl is configured for pth threads or if pth-threads are used directly)
09:10:09kugelI can probably create an explicit preemptive SDL (not part of the rockbox threads) to restore the interrupt-like behavior, so I would have basically swapped the main and the button-int thread, but it seems not really needed
09:10:11jhMikeSwhere's the code now in your tree?
09:12:47 Quit Doleo (Quit: ChatZilla 0.9.86 [Firefox 3.5.6/20091201220228])
09:13:49*jhMikeS still trying to find the "different" stuff
09:14:12 Join shaggy-h [0] (
09:14:49 Join efyx [0] (
09:15:19kugeljhMikeS: the sdl main thread is removed, gui_message_loop() is called from button_read_device
09:15:20Unhelpfuluse diff?
09:15:35kugeland it does SDL_PollEvent() instead of SDL_WaitEvent()
09:15:40 Quit sevard (Ping timeout: 240 seconds)
09:17:04*jhMikeS fails to find "SDL_PollEvent" in the four files in that tree
09:17:59kugellaptop battery empty :\ back in a few minutes
09:22:02 Quit kugel (Ping timeout: 240 seconds)
09:22:15*jhMikeS found it
09:23:43 Join Evilnick [0] (~Evilnick@rockbox/staff/evilnick)
09:23:59 Nick Evilnick is now known as evilnick (~Evilnick@rockbox/staff/evilnick)
09:25:07jhMikeSwhere and how is button_read_device now being accessed?
09:26:52 Join kugel [0] (~kugel@rockbox/developer/kugel)
09:27:08kugeljhMikeS: sorry, the button reading is still in uisimulator/sdl/button.c
09:27:22jhMikeSI found it eventually :)
09:27:33jhMikeSwho's calling button_read_device?
09:28:38 Join yukongt [0] (~chris@
09:30:17kugellike on target
09:31:24*jhMikeS only saw a tick_add_task for #ifdef HAVE_TOUCHSCREEN
09:32:01 Join DerPapst [0] (
09:33:28jhMikeSanyway, that doesn't matter. I see no problem offhand doing it from the faux tick instead
09:33:59kugelbutton_tick is in firmware/drivers/button.c
09:35:02*jhMikeS is just getting acquainted with all this being moved about. do have patience :)
09:40:33jhMikeSyou still have the tick on the outside to manage the various bits and drive the other threads ahead.
09:41:12kugel"drive the other threads ahead"?
09:42:11jhMikeSif they're all blocked, waiting for something else, they need something to wake them. they should genuinely sleep on the host.
09:43:34kugelthe tick cannot really be removed :)
09:47:22 Quit esperegu (Remote host closed the connection)
09:49:06***Saving seen data "./dancer.seen"
09:49:39 Join lpereira [0] (
09:50:20 Quit CGL (Quit: Soy spammero ---- \m/ d(>.<)b \m/)
09:51:09 Join esperegu [0] (~quassel@
09:53:45 Quit BHSPitMonkey (Remote host closed the connection)
10:07:20 Join funman [0] (~fun@rockbox/developer/funman)
10:10:05 Quit yukongt (Quit: Leaving)
10:11:43funmanrecording on as3525 using DMA gives saturated audio: , is 'saturated' the proper term?
10:18:13 Join komputes [0] (~komputes@ubuntu/member/komputes)
10:20:03 Quit arbingordon (Quit: `)
10:25:11 Quit Kitar|st (Ping timeout: 240 seconds)
10:25:24kugeljhMikeS: the polling method works with cooperative threads, like pth. I think pth has some advantages over the sdl threads
10:27:07 Join sevard [0] (~sev@
10:29:15 Join pamaury [0] (~pamaury@rockbox/developer/pamaury)
10:31:08 Join Kitar|st [0] (
10:38:28funmanseems to work now: fs#11257 'Sansa AMS: DMA recording'
10:39:39 Join anewuser [0] (anewuser@unaffiliated/anewuser)
10:49:35 Join mathw [0] (
10:50:16funmanoops the recording chunk size is 8192 bytes but we're limited to 8188 bytes in DMA transfers :o
10:51:25kugel8188 is a strange number
10:51:59funmanit's 32bits * 0x7ff
10:53:33mathwHello. I'm trying to write a theme that uses the new multifont code. The font file in question works perfectly if I select it as the user font from the menu. However, if I try to load it using a %Fl tag in the WPS, the WPS fails to load. The simulator says "Unable to load font 2: '12-Adobe-Helvetica-Bold.fnt'". Could anybody speculate on what might be amiss?
10:53:56jhMikeSfunman: make it 4192 then
10:54:00S_a_i_n_tthe font isn;t present in the sim?
10:54:13jhMikeSerr, 4096
10:54:24funmanjhMikeS: i'll just change the as3525 driver i guess
10:54:45jhMikeSI thought it was limited ??
10:55:09funmanwell i can make two requests of 4096 instead of one of 8188 and one of 4
10:55:10jhMikeSmulti-transfer per request?
10:56:15funmanpcm playback use fixed size too?
10:56:30funman#define PCMBUF_MIN_CHUNK 4096
10:56:43jhMikeSthat's for pcmbuf, not playback
10:56:53jhMikeS*pcm playback
10:58:19pamauryJdGordon: I just tried to get rid of your polling in usb.c and it works like a charm with a new message !
10:59:11 Quit ps-auxw (Read error: Operation timed out)
10:59:41jhMikeSthe beast has a limit of 65534, the smallest that i've seen, BUT the DMA controller can chain descriptors together itself. I just never set it up to do that with pcm. 64K-2 seems to be ok.
10:59:59AlexPmathw: Is the font named exactly that, and do you declare it with the extension?
11:00:31funmanchained descriptors are possible on as3525 too but i prefer simplicity
11:01:00 Quit slck (Read error: Operation timed out)
11:01:47 Join dfkt [0] (dfkt@unaffiliated/dfkt)
11:02:42jhMikeSfunman: on the beast, since the DMA is another CPU running a program, it's just part of it. btw, I think other hardware allow 2^24-1
11:03:35 Join ps-auxw [0] (~arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
11:04:29jhMikeSthere, it's just an array of structs with a buffer pointer and size for each, the controller reads that, generating an int on whichever one you want it to, optionally rewinding the pointer. it's rather a nice thing for the ATA driver
11:05:01 Join slck [0] (Venci@Slackware.SlackPix.Com)
11:06:32jhMikeSyeah, it's built into the scripts in the blob loaded into it at init
11:07:42funmanhm i have a parasite signal mixed with the recording, and I'm pretty sure it's not aliens trying to contact me :/
11:11:10S_a_i_n_tpretty sure, or positive? ;)
11:11:12jhMikeSperhaps it's voyager phoning home?
11:11:22S_a_i_n_tdo you have you aluminium hat on?
11:11:47jhMikeSv-ger's coming
11:12:04 Join pyro_maniac [0] (
11:12:24 Quit pyro_maniac (Remote host closed the connection)
11:12:55 Join pyro_maniac__ [0] (
11:13:22funman : FM recording, you can hear the woman pretty well
11:13:56funman(the woman's voice is normal, the other sound is the noise)
11:14:26jhMikeSthat's after fixing the transfers?
11:14:53funmanif(size > 4096) size = 4096; //instead of 8188
11:15:32 Quit linuxstb (Ping timeout: 260 seconds)
11:17:38jhMikeScore recording still expects 8192
11:17:51JdGordonpamaury: cool :)
11:17:58mathwAlexP: Yes, the font name's actually copied and pasted out of a directory listing from sheer paranoia. The line is "%Fl|2|12-Adobe-Helvetica-Bold.fnt|"
11:18:44S_a_i_n_tmathw: Are you exceeding the skin buffer?
11:18:45 Quit Forsaken (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
11:18:49AlexPmathw: How old is the version of Rockbox? It used to not require the .fnt and it might be that you are using an old version
11:18:54S_a_i_n_tthe font will probably fail to load then...
11:19:04JdGordonwhich target?
11:19:19funmanjhMikeS: hm perhaps there is a problem with pcm_rec_dma_get_peak_buffer() because pcm_callback_more_ready() isn't called before transfers are finished
11:19:36LloreanAlexP: I imagine if it was giving that error message then the simulator he's using at least is new enough, right?
11:19:54jhMikeSfunman: you have to implement the to return physical addresses if using DMA with a cache
11:20:00LloreanAlexP: I don't know anything about how multifont works, but would the size of the font be an issue (not enough WPS buffer left or something?)
11:20:35funmanthe what?
11:21:22funmanif recording isn't using iram the addresses are flat mapped and dcache entries are dumped
11:21:38S_a_i_n_tLlorean: not enough skin buffer can be a problem
11:21:45S_a_i_n_tbut font size isn;t relative
11:22:18mathwS_a_i_n_t: the line before the font error says "Skin buffer usage: 2824/10240"
11:22:30JdGordonmathw: which target?
11:22:31jhMikeSfunman: it uses the audio buffer. entries may be dumped but looking at cached addresses will "undump" them. the buffer is off limits until the transfer is done.
11:22:38JdGordonand yes, its not enough room to load the font then
11:22:44S_a_i_n_tHmmm...ok, well, that's that theory out the window then...
11:23:02mathwJdGordon: Sansa Clip. Not much memory, I now remember.
11:23:23JdGordonif you can build your own copy I can tell you how to fix it
11:23:39AlexPLlorean: I seem to remember it giving that error if it can't find it too
11:23:50LloreanS_a_i_n_t: Why wouldn't font size be an issue? How much space it takes is directly proportional to the size and number of characters.
11:23:57jhMikeSfunman: the beast uses DMA for recording and dma_get_peak_buffer returns the uncached aliases (physical)
11:24:01AlexPBut yes, it seems that size could well be it
11:24:08JdGordonLlorean: no, it is always 10KB for skin fonts
11:24:25mathwSo multifont on the clip is just a no-go then
11:24:30AlexPFor now
11:24:32funmanjhMikeS: i think i don't understand what the function must point to
11:24:40mathwOkay, thanks everyone
11:24:50AlexPYou can increase the buffer (which reduces the amount of memory available for music)
11:24:52S_a_i_n_twait, no...I confused myself, don't listen to me. No, seems there isn't enough room for the font.
11:24:53JdGordonmathw: wait for the next build... ill fix it for you
11:25:14S_a_i_n_tit needs what? 10K for each font?
11:25:15LloreanJdGordon: Ah, is it another LRU cache, or?
11:25:23TorneLlorean: yes
11:25:26funmanin svn it points to the address which is currently being written by DMA, but on PP it points to the buffer given to pcm_record_more() or pcm_rec_dma_start()
11:25:46mathwJdGordon: I wouldn't want to reduce the music buffer too much...
11:25:56LloreanJdGordon: So the more fonts you use, does it divide space up evenly between them, or do the most used characters stay independently of which font they come from?
11:26:02pamauryJdGordon: does your patch really do something about the different HID modes ?
11:26:21JdGordonLlorean: each font takes a seperate 10KB buffer for its cache
11:26:24JdGordonpamaury: nope
11:26:24jhMikeSfunman: the address the DMA uses, not the virtual address given to the recording calls. I'm just making sure. I've been through the tribulations already. :)
11:26:50pamauryJdGordon: do you think it has to be done in this menu ? The ubs screen already allow the mode selection...
11:27:15JdGordonI tihnk it should, just so the default mode can be set
11:27:21funmanjhMikeS: ok about using physical uncached address, but address of what exactly?
11:27:28JdGordonthat menu doesnt change anything, but it shuold be simple to hook it up
11:27:43pamauryJdGordon: but it only make sens if Mass Storage is used
11:27:46 Join S_a_i_n_t_ [0] (S_a_i_n_t@
11:27:57jhMikeSfunman: of the currect position of the DMA
11:28:02JdGordonyeah, ok
11:28:52pamauryJdGordon: I think it will avoid confusion but that's my opinion, we can always change that later anyway
11:29:34funmanjhMikeS: so it should point to the last sample recorded+1 ?
11:29:35 Quit S_a_i_n_t (Ping timeout: 240 seconds)
11:31:20jhMikeSfunman: PP passes the uncached phyical address back. it's converted before writing it to dma_play_data.addr
11:31:20mathwThanks for the help everyone
11:31:20 Part LinusN
11:31:20 Part mathw
11:32:44CIA-5New commit by 03jdgordon (r25962): Borrow another 10K to make sure multifont will have room on even the smallerst screens
11:32:44jhMikeSfunman: it's in the other drivers for coldfire and the like. it should record the previous position requested, up to the last sample in the current buffer. reset previous position to new buffer when it changes.
11:32:44funmanoh pcm_rec_peak_addr is a global variable
11:32:44funmani must have missed something
11:32:57jhMikeSyeah, that the previous position, where it starts for the next peak calc, DMA's current pos is where the calc ends
11:33:17funmanit's not used in svn but recording seems to work anyway
11:33:18jhMikeSit's reset to the start of the buffer on every new buffer
11:33:25jhMikeSyeah it is
11:33:28funmanin the as3525 driver i mean
11:33:52 Join antil33t [0] (
11:34:27LloreanJdGordon: Was it really necessary to just throw away 10k on all targets to fix the Clip?
11:34:35jhMikeSfunman: the driver updates it
11:34:52JdGordonits all targets except colour to be more precise... and 10K really is nothing
11:35:03JdGordona cleaner system is coming...
11:35:39Llorean10k is nothing unless you just keep adding 10k bits here and there every time you feel like it.
11:36:28kugelwhy not a MAX(X, SKIN_BUFFER_SIZE)?
11:36:31 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
11:36:56LloreanJdGordon: And it shows color targets too in the build log.
11:37:13kugelnot a fan of wasting 10k just for a tiny subset of targets
11:37:29gevaertskugel: you could, but I'd say that the targets that "suffer" from this are the ones with large screens, which are usually the ones with lots of RAM anyway
11:37:46Lloreangevaerts: Ones with large screens allocate a *lot* more memory to WPS, so don't actually suffer.
11:37:50jhMikeSfunman: it updates in pcm_record_data, pcm_record_more, and uses it in pcm_rec_dma_get_peak_buffer as the current address. pcm_calculate_rec_peaks increments it itself for multiple peak calculations within a particular buffer.
11:37:57JdGordon10K is what? 0.001% of abilable ram?
11:38:05kugelgevaerts: I can't follow that
11:38:07Lloreangevaerts: The problem is actually that the Clip's screen is so tiny (and the WPS buffer is calculated from screen size and bit depth) that it doesn't even have room for one additional font
11:38:24gevaerts"suffer" as in "get 10K that they don't actually need"
11:38:47kugelall 'suffer' equally, not only the ones with the large screens
11:38:54JdGordonno they dont
11:39:00JdGordon10KB from 64MB is SFA
11:39:09JdGordon10KB from 2MB is slighty more than nothing
11:39:23*jhMikeS thinks maybe the global variable arrangement should be rearranged, to make it less confusing.
11:39:32gevaertsBut the 2MB targets are the ones that are likely to need the 10KB more
11:39:45Lloreangevaerts: Except they generally have smaller fonts anyway.
11:39:53LloreanIt might make sense to scale the font buffer by screen size as well.
11:40:09*S_a_i_n_t_ doesn;t really think so
11:40:09kugelWe shouldn't stop considering ram usage just because there's plenty available
11:40:28LloreanGive the large screen targets more space for larger fonts, and the small screen targets can save RAM since they're not going to be using 18-20 pixel fonts anyway
11:40:57JdGordonsmaller fonts means all the more reasont o want multifont
11:41:01S_a_i_n_t_a small screen target still may want to use numerous fonts, *and* have a nice .sbs/.wps
11:41:11LloreanJdGordon: What do you mean?
11:41:13gevaertsLlorean: maybe
11:41:24gevaertsThat could be a good idea I guess
11:41:26kugelalso, with the ams targets we have targets with large screens but not huge ram (8MB, 4MB of which is available for buffering)
11:41:49LloreanS_a_i_n_t_: Yes, but since memory is allocated *per font* then if you allocate 5kb per font because the actual font size is average, the *number* of fonts isn't important (in fact you can use more fonts, more efficiently)
11:42:03LloreanS_a_i_n_t_: If you want 3 fonts at 10kb per font, it takes twice as much memory even if you aren't using it all because the fonts are tiny
11:42:35S_a_i_n_t_but, you may not want to use a tiny font?
11:43:06gevaertsS_a_i_n_t_: larger fonts will still work
11:43:34LloreanS_a_i_n_t_: Do you really think a significant percentage of users will want a screen that is only able to display 2 or 3 lines of text? While larger fonts can be used, it will be optimized to be more efficient with fonts appropriate to the screen size, that's all.
11:44:08LloreanIt doesn't rule anything out, it just makes certain things more or less efficient based on what are probably more likely scenarios given the physical limitations of the player
11:44:35 Join arbingordon [0] (~w@unaffiliated/arbingordon)
11:45:08 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
11:45:21funmansomeone can try on AMSv1 ? all my other sansas are busy benchmarking a clock change
11:45:24Tornerelevant question: for the typical fonts used on the players with tiny screens, say, the one cabbie uses for a start..
11:45:25*jhMikeS thinks even more functionality can be moved up into pcm.c
11:45:31Tornehow many glyphs fit into 10kB?
11:45:32*Llorean just doesn't like when the first reaction to a problem is "Throw more RAM at it, and it'll be fine"
11:45:36Torneor 5kB?
11:45:42LloreanTorne: Depends on the font size
11:45:50Torneso guess what a typical font size is
11:45:56LloreanIf they're 1bpp, it should be a lot.
11:45:58Tornelike i said, the one cabbie uses might be a good start
11:46:32S_a_i_n_t_12 helvetica appears "standard"
11:46:34 Join LinusN [0] (~linus@rockbox/developer/LinusN)
11:46:48JdGordon12*12/8 per font?
11:46:52LloreanTorne: A good idea might be to say "we should try to fit X number of glyphs from the cabbie font" and go from there?
11:47:07Torneit is indeed 1bpp
11:47:10JdGordon18 bytes per glyph assuming no extra ovherhead... so lots
11:47:20gevaertsLlorean: we're discussing extra fonts here, so the one from cabbie might be irrelevant
11:47:23Tornethere's a bit more overhead, iirc about 4-8 bytes
11:47:29Tornebecause the font cache is crazy
11:47:41gevaertsLet's assume 32 bytes per glyph to get round numbers
11:47:41Lloreangevaerts: Yes, but for deciding what an "average size" font might be, it's a decent start.
11:47:58Tornegevaerts: it's 10000 bytes, though, so it's not a round size ;)
11:48:00Lloreangevaerts: So that's ~312 glyphs?
11:48:01Torneit's not really 10kB
11:48:29*Torne looksin fontcache to check the overhead also
11:49:10***Saving seen data "./dancer.seen"
11:49:11*gevaerts thinks that the standard latin alphabet in upper- and lowercase, digits, and some punctuation is probably plenty for extra fonts
11:49:26gevaertsso about 75 glyphs would be enough I think
11:49:41Torneit's 7 bytes overhead per glyph
11:49:49Torne(not 8)
11:50:16gevaertsok, so assuming 18 bytes per glyph, 7 bytes overhead, 75 glyphs, we get 1875 bytes
11:50:32Tornei think we can live with <10kB, then
11:51:06Tornethe targets with tiny screens/ram are all flash, right? (apart from archos)
11:51:25Tornefontcache only really sucks hard performance-wise on disks (and not even on all disks then)
11:51:25JdGordondepends on your definition of tiny
11:51:27S_a_i_n_t_safe bet
11:51:28gevaerts5kB gives us 22-sized fonts
11:51:30LloreanTorne: Pretty much
11:51:55TorneJdGordon: also no, you asked the other day if i've done anything about font_load and i haven't ;)
11:52:01gevaertsand 22-sized fonts on the clip are insane if you actually want to display strings that use those 75 glyphs
11:52:01Tornei still hasven't decided what a good thing to do would be
11:52:22*gevaerts votes for 3kB per font on small-mem
11:52:52TorneThis reminds me; do we want to, in fact, increase the main UI font buffer size on targets with huge screens and loads of ram?
11:52:59Torneas a temporary workaround for font_load() being slow
11:53:00S_a_i_n_t_wouldn;t actually finishing dynamic skin-buffer fix this "properly"?
11:53:12TorneS_a_i_n_t_: not really
11:53:20funmanranma: does have an effect on power usage?
11:53:20Torneyou still nede to allocate *some* fixed size per font
11:53:27Torneunless you are going to allocate sizeof(entire font)
11:53:31Torneand avoid the cache entirely
11:53:44LloreanTorne: But you could allocate is at sizeof(glyph)*65 or something
11:53:56Llorean*it as
11:53:57TorneLlorean: That's true..
11:54:04Tornebut equally, you could do that *now*
11:54:21LloreanThat might improve the situation somewhat, then?
11:54:28Tornehow much you allocate per font is independant from how much you allocate for the skin buffer overall
11:54:34Torneexcept the former has to fit in the latter :)
11:54:55LloreanI mean, according to that guy, the clip had like 8kb of skin buffer free anyway, right?
11:55:07gevaertsTorne: decreasing the font buffer size would also speed up loading I suspect
11:55:11Llorean2824/10240, so 7ish
11:55:28LloreanSo even without the additional 10kb, if we used (glyph)*75, a lot of fonts would load just fine on it
11:55:28Tornegevaerts: you mean the main buffer and font_load problem, or skins?
11:55:39Torneah. well, technically yes, it will
11:55:59Tornepossibly contentuous question: couldn't we just dynamically allocate the main UI font buffer at boot? :)
11:56:23pamauryJdGordon: why did you add a list of USB_* in settings.h but never use it ?
11:56:27LloreanTorne: And then use fontcache only if they switch fonts?
11:56:34TorneLlorean: yah, or for skin fonts
11:56:42Torneor possibly for targets with small ram
11:56:49Tornewe could *only* do this on players with 16mb or more anyway
11:56:52LloreanSo allocate it dynamically, but cap it?
11:56:58LloreanI mean, unifont is *huge*
11:57:19JdGordonpamaury: for the setting
11:57:28*funman thinks 8mb ought to be enough for everyone
11:57:36TorneLlorean: True, it is..
11:57:40pamauryJdGordon: but it's currently unused ?
11:58:51JdGordoni ddint get around to finishing it
11:58:51pamauryAnd what is your plan about it ?
11:58:51 Part LinusN
11:58:51Torneit would still be nice to actually fix font cache loading to not suck, of course :)
11:59:46funmanjhMikeS: thanks for the help
11:59:46 Quit funman (Quit: free(random());)
11:59:46S_a_i_n_t_Torne: Looking into allocating the the font buffer at boot doesn't sound like a bad road to go down...
11:59:46TorneLlorean: the cap would have to be at least 256kB to actually fix the observable problem on ipod video
12:00:01Tornesince 14-Adobe-Helvetica is 231,946 bytes :(
12:00:14 Join LinusN [0] (~linus@rockbox/developer/LinusN)
12:00:28S_a_i_n_t_and, I have a "small screen" player, and that is my most used font for multifont.
12:00:39TorneS_a_i_n_t_: haha ;)
12:00:45LloreanTorne: 256kb isn't too bad.
12:00:46TorneS_a_i_n_t_: i strongly recommend you find a different one ;)
12:00:52LloreanS_a_i_n_t_: None of the nanos are small screen.
12:00:59LloreanThey're physically small, but they've got a decent number of pixels
12:01:00*JdGordon wants 32bpp fonts!
12:01:09S_a_i_n_t_that and unifont, but I may be an exception..
12:01:29TorneS_a_i_n_t_: by small we mean, 128x64 as on clip
12:01:30Torneor similar
12:01:37LloreanThe AJBR, the Clip.
12:01:49LloreanThe H100 is at the top of small / bottom of medium
12:01:58S_a_i_n_t_176x132 is hardly *huge* ;)
12:02:13LloreanNo, but it's twice the vertical resolution of the screens we're talking about.
12:02:15S_a_i_n_t_but, I see your point
12:02:22TorneS_a_i_n_t_: that's three times as many pixels!
12:03:07Torneanyway, a lot of people would probably see a *benefit* from sizing the buffer dynamically, since a lot of the fonts people use are less than 80k :)
12:03:15S_a_i_n_t_three times as many *tiny* pixels ;P
12:03:24*S_a_i_n_t_ waits to be chastised.
12:03:48JdGordonTorne: thing is, its not the size of the cache that is the problem.. its the brainddead loading system.. going by the above numbers it is very unlikely many people would ever need the current 60KB for the main font
12:04:00AlexPS_a_i_n_t_: s/chastised/ignored/ :)
12:04:10pamauryJdGordon: see for the modified patch:
12:04:37LloreanTorne: Would it be possible to write the current cache in nvram on shutdown, and just load that on boot?
12:04:44JdGordonqueue_broadcase probably isnt the best wya to do it
12:08:54 Quit stoffel (Remote host closed the connection)
12:08:54TorneLlorean: yes, you could persist the whole cache, it's one of the optoins we discussed before
12:08:54LloreanTorne: Was it objected to?
12:08:54LloreanIt seems like a really obvious solution for the short term
12:08:54Tornethere's a few minor issues (the cache has a couple of pointers in it at the beginning that would need fixing up)
12:08:54JdGordonyou could NOT save it to nvram
12:08:54Torneand you'd need to know whether the font had changed and so on
12:08:54 Part LinusN
12:08:54DBUGSent KICK Torne to server
12:08:54Torneand it breaks multifont ;)
12:08:54Torneat the moment multifont behaves pretty weirdly with the cache anyway
12:08:54***Alert Mode level 1
12:08:54DBUGsent MODE #rockbox +b *!*torne@rockbox/developer/Torne
12:08:54Tornebut i *think* the result is right, it just wastes effort getting there
12:08:54***Alert Mode level 2, maybe not, i've not really tested
12:08:54***Alert Mode level 3
12:08:54Torneall the fonts kinda share the same glyphcache, i'm not sure who actually gets to write it
12:08:54JdGordononly the main font
12:08:54***Alert Mode level 4
12:08:54TorneJdGordon: so what if th emain font doesn't use the cache?
12:08:54DBUGSent KICK JdGordon to server
12:08:54JdGordonthen it is ignored?
12:08:54***Alert Mode level 5
12:08:54Tornethe skin fonts just load the default ascii glyphs?
12:08:54JdGordonyes, iirc
12:08:54***Alert Mode level 6
12:08:54Torneah, that's not terrible
12:08:54 Join Forsaken [0] (~chatzilla@
12:08:54***Alert Mode level 7
12:08:54Torneit does try to load more than will fit
12:08:54***Alert Mode level 8
12:08:54TorneLlorean: there's no major objections, anyway, i don't think
12:08:54***Alert Mode level 9
12:08:54Torneit's just effort :)
12:08:54 Part Zagor
12:08:54 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
12:08:54Kick(#rockbox Torne :No flooding!) by logbot!
12:08:54***Alert Mode level 10
12:08:54Mode"#rockbox +b *!*torne@rockbox/developer/Torne" by logbot (
12:08:54Kick(#rockbox JdGordon :No flooding!) by logbot!
12:09:43 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
12:09:52JdGordondid i accidently pastedump?
12:10:11*S_a_i_n_t_ didn;t see anyhthing... :/
12:10:17S_a_i_n_t_*didn't even
12:10:33JdGordonlogbot kicked me
12:10:45Mode"#rockbox +o gevaerts" by ChanServ (ChanServ@services.)
12:10:46Mode"#rockbox +o GodEater" by ChanServ (ChanServ@services.)
12:10:48Mode"#rockbox -o GodEater" by ChanServ (ChanServ@services.)
12:10:52Mode"#rockbox -b *!*torne@rockbox/developer/Torne" by gevaerts (~fg@rockbox/developer/gevaerts)
12:11:03Mode"#rockbox -o gevaerts" by ChanServ (ChanServ@services.)
12:11:12 Join Torne [0] (torne@rockbox/developer/Torne)
12:11:46Torneokay apparently i talk too fast, or at least hit enter too often
12:12:00JdGordonI think logbot got bored of the convo :p
12:12:04S_a_i_n_t_nah...Jd got kicked too.
12:12:50LloreanLogbot may have lagged and seen all of their conversation in a much shorter time frame?
12:13:14gevaertsBut JdGordon didn't get banned
12:13:29Lloreangevaerts: He had a lot less lines?
12:13:44Torneyeah, that seems likely
12:15:17S_a_i_n_t_Perhaps Jd and logbot ar in cahoots?
12:15:36S_a_i_n_t_conspiring against Torne
12:15:44JdGordonok, so I'm happy shrinking the font buffer size on some targets.. any suggestions how to do it?
12:15:56JdGordonheight < 176?
12:16:11JdGordon|| width < 176?
12:16:21LloreanMaybe even < 160 for height?
12:17:25JdGordonor go by ram amount even
12:17:33JdGordonseen as whatever it is its entirely arbitrary
12:18:55***Alert Mode OFF
12:22:02CIA-5New commit by 03jdgordon (r25963): ok, use a smaller font size on shorter display targets (3K instead of 10K)
12:23:56jhMikeSfunman: think I'm going to fix that stuff up and give the low level even less to worry about (shouldn't be too much trouble at all).
12:25:31JdGordonI'm about to commit fm skin patch (as in really commit it now...) which means... someone needs to update cabbiev2 for it :)
12:25:38JdGordonand do screenshots, and update docs :/
12:26:07JdGordon(no added tags will be affected by the |->() breaking change)
12:28:59 Join mikroflops_ [0] (
12:33:38 Quit mikroflops (Ping timeout: 276 seconds)
12:33:53n1sthat remids me, someone should make RDS work on the targets that have such HW (the beast at least)
12:36:09 Join LinusN [0] (
12:36:10 Quit LinusN (Changing host)
12:36:10 Join LinusN [0] (~linus@rockbox/developer/LinusN)
12:38:04CIA-5New commit by 03jdgordon (r25964): FS #10853 - Skin support in the radio screen! Check CustomWPS for the new tags
12:42:37 Part LinusN
12:42:39pamauryJdGordon: why do you say broadcast is not the good way ? You don't have access to the usb queue
12:43:25JdGordonwell only the usb thread should actualy see that message. The way I did it was in that usb_allow_connection() function to queue_post() to the usb queue which it does have access to
12:43:40 Join LinusN [0] (~linus@rockbox/developer/LinusN)
12:44:22JdGordonThose massive deltas are the addition of a backdrop buffer.. nothing to worry about...
12:44:39pamauryhum, yeah that works too
12:44:50pamauryI'll change that you're right
12:45:03JdGordon1k for that change :) not a bad effort if I do say so for myself
12:46:45pamaurywhat is the difference between queue_post and queue_send?
12:47:43JdGordonsend blocks the thread, post doesnt
12:47:52pamauryblocks until what ?
12:50:24JdGordonuntill queue_reply() or another message is taken off the queue apparently
12:50:36JdGordonwhy the hell does ipod have HAVE_RDS_CAPS defined?
12:51:15pamauryJdGordon: new version
12:53:06TorneJdGordon: presumably because the radio supports RDS?
12:53:23JdGordonit isnt set up right apparently :/
12:55:18JdGordonI'm #including tuner.h which includes the ipod specific file... no dice
12:59:09pamauryJdGordon: if you have nothing more to comment on my patch, I'll integrate it to my mtp branch
12:59:22CIA-5New commit by 03jdgordon (r25965): fix red and yellow (most of it anyway)
13:00:05JdGordonpamaury: busy atm, the apps/ side isnt ready but if you want to integrate it then cool
13:00:24pamauryJdGordon: I know but at least it's a base to work on
13:03:37 Join moos [0] (moos@rockbox/staff/moos)
13:05:45JdGordonsnafu on the e200 build?
13:05:58 Quit komputes (Ping timeout: 240 seconds)
13:06:24JdGordon has the new tags
13:07:58 Quit kugel (Ping timeout: 240 seconds)
13:10:42*JdGordon is stumped by the ipod red
13:18:54JdGordonoh bloody hell.. its sim only :/
13:22:14S_a_i_n_t_wah wah...
13:22:47S_a_i_n_t_what's the problem in the sim?
13:23:02JdGordonthe RDS #ifdefs havnt been done correctly
13:25:00n1syay, i fixed my asm bug
13:25:31n1sno i didn't
13:30:43 Join watto [0] (~watto@
13:44:12CIA-5New commit by 03jdgordon (r25966): fix the last of the reds, and make RDS tags actually work (and add them to the hardcoded radio skin)
13:46:21 Quit Forsaken (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
13:49:13***Saving seen data "./dancer.seen"
13:49:13 Join Forsaken [0] (~chatzilla@
13:52:01 Join pondlife [0] (~Steve@rockbox/developer/pondlife)
13:52:03CIA-5New commit by 03jdgordon (r25967): merge the fm skin and other changes from HEAD to the breakingskin branch
13:53:39 Join Tuplis_ [0] (
13:53:44 Quit Tuplis (*.net *.split)
13:53:50 Nick Tuplis_ is now known as Tuplis (
13:56:38pondlifeJdGordon: Where is this big backdrop buffer that ate 40k of my RAM ;)
13:57:03JdGordon1 backdrop buffer per skinnable screen
13:57:47pondlife40k? That's the price for having radio + remote + colour, I guess..
13:58:07JdGordoneasy enough to change if you do your own build anyway
13:58:20pondlifeWill do, but which .c file is it in?
13:58:54JdGordonmake sure your skins have %Xd then or are all using the same bmp
13:59:06JdGordonwhich target you on now? or you got the h300 fixed?
13:59:28pondlifeOnly got the h300, still in surgery...
13:59:41pondlifeSo, I run a sim!
13:59:47amiconnjhMikeS: Iirc sound_val2phys() is meant to be phased out.
14:00:13pondlifeOw, +98848 for Onda fans :/
14:00:22JdGordonBIG screen :)
14:00:49*pondlife thinks a config to allow skins or not might be brewing
14:00:56JdGordonone of the next things on my plate is putting in an option so that buffer isnt so huge
14:01:12pondlifeOK, sounds good
14:01:58gevaertspondlife: I don't think disabling skin support is workable
14:02:23JdGordonwe can disable loading skins
14:02:28JdGordonbut not outright disabling it
14:02:28pondlifeAnything happen to the idea of pre-parsing a theme/fonts to size buffers? (With, natch a reboot needed to free RAM)
14:03:50JdGordonkugel put it in the too hard basket
14:08:57amiconnn1s: The head byte handling can be improved so that the unconditional branches can go away
14:09:53mc2739JdGordon: is there a way to do station logos on the FM radio skin?
14:10:28JdGordonpamaury: what do you tinhk about adding a "usb screen" menu item when it is connect to the main menu? mostly for HID
14:10:31JdGordonmc2739: not yet
14:10:43JdGordonpatches welcome for that though :)
14:11:54TorneJdGordon: I like that idea..
14:12:37pamauryJdGordon: that would be a good idea indeed
14:13:00JdGordonshould be a simple addition
14:13:12JdGordonthen we put back the default HID mode in the setting menu
14:14:19n1samiconn: ah, right
14:14:49*amiconn wonders whether the dozen little glitches in the ui caused by viewportification will ever be fixed :\\
14:15:22*JdGordon bites
14:15:26JdGordonare then in flyspray?
14:15:52amiconnLike scrolling lines running across the usb screen, usb screen sometimes only using half the lcd...
14:21:10JdGordonare they in flyspray?
14:25:52amiconnIdk... I would expect those glitches to be obvious without reporting though
14:27:50jhMikeSamiconn: why should it be phased out?
14:28:56n1shmm, whe i start my h300 and it autoresumes and i quickly press stop, the boost counter is stuck at 1 sometimes
14:30:19S_a_i_n_t_wasn't the wps text scrolling in the USB screen fixed?
14:30:28S_a_i_n_t_I haven't seen it in ages nway...
14:30:37S_a_i_n_t_*anyway too
14:31:18 Join komputes [0] (~komputes@ubuntu/member/komputes)
14:31:48amiconnThe sound code should pass around either centibels, or percentages for linear settings. Converting to hardware bitpatterns should be handled by the respective driver
14:31:59JdGordonscorche: pingaroony?
14:32:16JdGordonnone of the theme site people are online :(
14:32:27amiconnS_a_i_n_t_: May be it was fixed for wps specifically... I often have menu strings scrolling across it
14:32:36CIA-5New commit by 03nls (r25968): Coldfire strlen, slight optimization pointed out by amiconn, eliminate some unconditionla branches by rearranging the code handling the head bytes
14:32:44 Join Schmogel [0] (
14:33:21S_a_i_n_t_amiconn: huh...weird. I've never seen that personally. But WPS txt I used to see on the USB screen regularly.
14:33:42S_a_i_n_t_can't remember the last time though, so perhaps it fixed itself magically
14:36:42jhMikeSamiconn: I was thinking the complete opposite, pass things around as a range from 0 to n-1, n=number of levels. this way any pattern of values can be supported and it's simpler. no sound_min needed and thus fewer fields in the sound structures.
14:37:29 Join apoelstra [0] (
14:39:07 Quit apoelstra (Client Quit)
14:39:51amiconnYou'll save one field - and introduce a whole code path in sound_val2phys()
14:39:54*n1s smacks forehead
14:40:34amiconnAlso, using hardware values makes dependency handling (like volume prescaling for positive treble/bass gain) a nightmare compared to using centibels
14:41:24amiconnBtw, this is nothing I'm making up - we used to have what you describe - with way less targets, and hence audio dacs, than now
14:41:33CIA-5New commit by 03uchida (r25969): fix: the problem to which new filetypes cannot be registered. ...
14:41:59JdGordonIf I wanted to use the audiobuffer in the radio screen, does anyone know if I would be told when playback steals it back?
14:42:43n1shmm, on the other hand, not as bad as i though
14:43:10amiconnJdGordon: On hwcodec you can't do that anyway - the audio buffer is set up for recording then
14:43:21 Quit Schmogel (Read error: Connection reset by peer)
14:43:24jhMikeSamiconn: you'd save calls to sound_min and only code that needs to care (like formatting for display) has to do anything. volume prescaling only needs conversion to centibels.
14:43:36JdGordonamiconn: sure, but im talking swcodec only for now
14:44:35jhMikeSJdGordon: It steals it back just before restarting playback, I think the playlist code might grab it for a bit at that point at well
14:44:49JdGordonis there any way of knowing it was stolen?
14:45:08JdGordonI want to use it as an image buffer, nothing fancy, but I need to know if it was trashed
14:45:45*n1s wonders why we need *so* many filetypes
14:46:20JdGordonor do I just use the buffering mechanism and not worry?
14:47:19jhMikeSit has a trashed flag set when it's stolen
14:48:07jhMikeSit can be stolen with voice in tact as well, with the right parameters
14:48:26amiconnn1s: Your current code doesn't align properly
14:48:57amiconn(well, it does for 0 or 2 bytes, but not for 1 or 3 bytes)
14:49:40JdGordonjhMikeS: if playback is stopped, do you know if bufalloc() will work?
14:50:50n1samiconn: could please elaborate? (it benched the same or slightly faster than the old code)
14:51:11 Part LinusN
14:51:31JdGordonmaybe I claim the codec buffer instead to keep it simpler?
14:51:32jhMikeSJdGordon: I believe it has to be non-talkable (trashed) to use bufalloc safely.
14:51:40amiconnNvm, you're right
14:52:02n1sok, np
14:53:15jhMikeSJdGordon: just use audio_get_buffer with talk_buf=false
14:53:19 Join kugel [0] (~kugel@rockbox/developer/kugel)
14:53:43kugelJdGordon: is there any reason the current AA buffering mechanism shouldn't work?
14:53:58JdGordonI don't know how it works...
14:54:18kugelI imagine you'd just need to call it directly because it doesn't happen automatically with track buffering
14:54:33jhMikeSif you don't know how something works, homebrew another method? :??
14:55:06jhMikeSJdGordon: btw, AUDIOBUF_STATE_INITIALIZED means playback trashed your buffer (because it reinitialized)
14:56:07 Join Jaykay [0] (
14:57:21kugelJdGordon: in audio_finish_load_track() the actual loading happens
14:57:54kugelcoping that part and making sure find_albumart() actually finds the station logos should get you somewhere
14:58:13JdGordonwhat state is the buffer in when the radio is started? how do i know when to release those handles?
14:59:00kugelit's empty, I don't know which state exactly
15:00:03*JdGordon attempts to open a handle in the fm screen
15:00:34kugelnot sure if there's any magic logic to detect it, but it should be pretty safe to assume that you need to release them when the wps is entered
15:01:06jhMikeSwith playback stopped, the whole engine should be idle and all threads inactive
15:01:21kugelAFAIK there's no way to play audio files without entering the wps at the same time
15:01:47jhMikeSkugel: how does the playback control do it then?
15:02:10kugelwhat playback control? you mean in the plugins?
15:02:23 Join dfkt_ [0] (~dfkt@unaffiliated/dfkt)
15:02:50jhMikeSyes, that and pictureflow?
15:03:23 Quit dfkt (Disconnected by services)
15:03:25 Nick dfkt_ is now known as dfkt (~dfkt@unaffiliated/dfkt)
15:04:01kugelI think for playback control there needs to be a current playlist which isn't given when the radio is on (doesn't playback control control the radio then anyway?) pictureflow could be problematic indeed
15:04:14JdGordon-4 file error?
15:04:23*jhMikeS balks that pictureflow can start playback, but not adjust the volume when you get a loud blast after playing something quiet
15:04:39kugelfix it! :)
15:04:57kugelpictureflow should be able to re-use volume in lists feature
15:05:27jhMikeSyou have to go to playback control first
15:05:53jhMikeSoh, pictureflow can jump right back to wps from its menu too. :)
15:06:19kugeljhMikeS: I mean it doesn't now, but it should be able to
15:06:29JdGordonbufopen() does work in the radio screen... ok, now to figure out when to close it
15:06:31kugelit uses the standard contexts and get_action() I believe
15:07:01jhMikeSkugel: would be nifty :D (or swell) <insert 50's lingo>
15:07:23kugelJdGordon: I'd think that if audio_play() is called ?
15:07:34 Quit antil33t (Read error: Connection reset by peer)
15:07:41 Join antil33t [0] (
15:08:06JdGordonI suppose I could wait for the "track buffered" event, but that does run the risk of there not being enough room
15:08:16JdGordonhighly unlikely for now, but if I make it mass load images...
15:08:36*jhMikeS wonders if pictureflow just absolutely *has* to run at 150fps on the beast
15:10:24JdGordonis the radio actually turned off it music is restarted?
15:10:45jhMikeSJdGordon: when it sets the audiomux back to playback it turns it off
15:10:59 Join adnyxo [0] (
15:11:05JdGordonthe apps/ layer doesnt know this though?
15:11:21kugelJdGordon: maybe when radio is stopped? currently music AA is cleared when music stopped too (no matter of whether the radio starts after or whatever)
15:11:33jhMikeSJdGordon: first call in audio_play_start
15:12:10 Quit kugel (Quit: exit(0);)
15:12:50JdGordonOK, I tihnk adding an event "Audio is restarting" might be useful here
15:15:03jhMikeSwhat if it stops again before you're done?
15:16:01JdGordonthe event would just close the handle which should be instant
15:16:41JdGordonif audio is starting the radio needs to go into shutdown mode anyway so it will have to rebuffer next time it is started anyway
15:19:23 Join JohannesSM64 [0] (
15:20:28 Quit Forsaken (Ping timeout: 252 seconds)
15:25:31pondlifeJdGordon: Does recording from the radio use the audio buffer?
15:25:49JdGordonprobably, only hwcodec can do that
15:25:55JdGordonand it doesnt have albumart so no big deal
15:27:06 Join Forsaken [0] (~chatzilla@
15:28:09 Join jgarvey [0] (
15:28:42 Quit Rob2223 (Quit: Rob2223)
15:29:20 Join evilnick_B [0] (~0c140464@rockbox/staff/evilnick)
15:32:41JdGordonUnhelpful: ping
15:33:22pondlifeIs there no radio recoding on SWCODEC? Or no chance of it...?
15:33:37*pondlife has forgotten lots
15:35:44n1sthere is radio recording on SWCODEC
15:35:57 Join Rob2222 [0] (
15:36:00pondlifeI though there was..
15:37:08*JdGordon has radio AA working!
15:37:15JdGordonquick and dirt.. but working
15:37:21pondlifeAirwave Art
15:37:31JdGordonn1s: not from inside the radio screen though
15:37:41JdGordonbut that will break this anyway
15:37:58JdGordonscrew the lot of you :D
15:38:02JdGordonradio Art!
15:38:19pamaurythen it's RA !
15:39:53JdGordonnow I need to make it actually be useful.. but that can wait for another day! midnight snack time
15:42:14 Join joeyg [0] (
15:49:16***Saving seen data "./dancer.seen"
15:53:45 Quit komputes (Quit: I haven't slept for ten days, because that would be too long.)
15:54:34 Join komputes [0] (~komputes@ubuntu/member/komputes)
15:58:26 Join MethoS- [0] (~clemens@
16:05:41CIA-5New commit by 03jethead71 (r25970): PCM bottom layer simplification. pcm_rec_peak_addr variable no longer has to be handled there. Driver can just return current pointer for recording ...
16:07:27dfktcould i please get the official v2 bootloader-clipplus.sansa from someone? it's not on the site yet.
16:07:55*B4gder gets reminded of something...
16:10:32 Join gerbil [0] (
16:10:53 Join CGL [0] (~CGL@
16:11:50amiconnpondlife: There is radio recording on swcodec, but not in the radio screen
16:12:21amiconnYou have to go to the recording screen for that... so no prerecording in the radio screen either
16:14:40gerbilHi all, I've created my first patch for Rockbox. The kaybox plugin now can import data from a text file. For me it's a big improvement as it's totally boring to key in a bigger data base using the few buttons of the device. I'd like to take this opportunity to ask you to roughly scan the patch and give me advice how to improve it. I'd really like to have it included into Rockbox.
16:15:49n1sgerbil: i looked at it breifly, but am a bit busy right now, but will try to take a closer look sometime in the future
16:15:52gerbilSorry, s/kaybox/keybox/
16:16:15gerbilThanx a lot n1s!
16:16:31n1sone thing i though of was to ask the user if they want to delete the unencrypted file and then doing it directly instead of just telling them to do it
16:17:02n1salso, i think your splashes have too long timeouts and also different timeouts
16:17:12gerbilok, I think about it how to do it best
16:17:43gevaertsJust deleting the file won't do much
16:18:47gerbilaccording the splashes, if a line is too long for the data base the user is informed. especially for this message, I had to increase the time to be able to read the full text.
16:19:33gerbil@gevaerts: you are right, I'll change it.
16:20:02*gevaerts will just look for passwords on the raw disk :)
16:21:17gerbilyou want to state I have to overwrite the raw data? That's surely harder to do. there I have no idea how to do it.
16:21:51gevaertsI think it needs to be thought about anyway
16:23:52 Quit pyro_maniac__ (Remote host closed the connection)
16:24:02 Join pyro_maniac [0] (
16:27:07jhMikeSwell, all this uncovered an interesting pp5002 bug in the pcm
16:30:38n1smaybe we need a schred() function in the fat driver...:)
16:30:54 Quit m0ar (Quit: Lost terminal)
16:31:13CIA-5New commit by 03jethead71 (r25971): Fix r25970 red and yellow *and* a bug uncovered for pp5002 where pcm_play_dma_start wasn't actually initializing the dma_play_data with the first ...
16:31:25n1sgerbil: also, you are using strcpy in some places to copy user provided strings, are the lengths chaecked before?
16:31:50gerbilanother question, according the virtual keyboard layout the manual states "If you want the new layout to be loaded on each boot, put it into your .rockbox folder. " This does not work for me. I can "play" my new layout but it gets not auto loaded when I copy the file into the .rockbox folder. Have I to name the file any special or is it a bug I should file?
16:32:11 Join hebz0rl [0] (
16:32:56gerbiln1s: I'll look for it.
16:33:28 Join raphi_ [0] (
16:33:36n1sgerbil: does it work if the file is inside .rockbox when you "play" it?
16:33:54n1si think that's the way it's supposed to work at least
16:34:52gerbiln1s:I've named it opti.kbd, I can play it fine but it gets not auto loaded?!
16:35:13n1sbut is it inside the .rockbox dir when you "play" it ?
16:35:48CIA-5New commit by 03jethead71 (r25972): Missed a yellow on Cowon D2. :)
16:37:11gerbiln1s:sorry for the noise, it seems to work by now. Don't know what I've made wrong yesterday.
16:37:58 Quit GodEater (Ping timeout: 248 seconds)
16:39:40 Join dys [0] (
16:42:53 Quit pamaury (Quit: Quitte)
16:43:20gerbiln1s:yes, before the strcpy I check that string length does not exceeds FIELD_LEN. Should be ok.
16:45:58 Join S_a_i_n_t [0] (S_a_i_n_t@
16:46:39 Quit S_a_i_n_t_ (Ping timeout: 240 seconds)
16:55:08 Quit gerbil (Quit: CGI:IRC (EOF))
16:57:32 Join toffe82 [0] (~chatzilla@
17:00:35JdGordonarg.. the albumart search code sucks for radio... does it really need to search for <name>.widthxheight.[jpg|bmp] ?
17:00:48JdGordonor is <name>.jpg good enough?
17:01:53AlexPPersonally with resizing I don't think it needs to look for specific sizes, but there are probably those that would like it to look for the specific size and save a few cycles on resizing
17:01:53 Part Zagor
17:03:29JdGordonLAME! I need a presets file which works with the sim now
17:08:34UnhelpfulJdGordon: was it about searching for the size-tagged AA? that's been debated a few times and there never seems to be any consensus
17:09:00JdGordonno, I wanted to ask how to actually load a jpg/bmp but I figured that out
17:09:14JdGordonwhat's with the dimensions struct in playback.c?
17:09:25JdGordonwhy not directly use width/height from the aa struct?
17:11:09JdGordonit works :D
17:11:20JdGordonanyone want to test this out on target?
17:12:28mc2739JdGordon: I can test on target
17:12:59JdGordonyou need an fmr file and a bmp image in the fmpresets folder named the same as the preset name.bmp
17:17:49 Quit antil33t ()
17:22:21 Quit tchan (Quit: WeeChat 0.3.3-dev)
17:27:17mc2739JdGordon: compile errors -
17:28:16JdGordonapps/recorder/radio.c line 526 change strncpy to strlcpy
17:29:20 Quit esperegu (Read error: Connection reset by peer)
17:43:30 Quit B4gder (Quit: It is time to say moo)
17:44:51JdGordonanyone got an OF.mi4 for a e200v1 handy?
17:45:27 Quit lpereira (Quit: Leaving.)
17:46:58 Join Xerion [0] (
17:49:13 Part pondlife
17:49:19***Saving seen data "./dancer.seen"
17:53:27 Join bluebrother [0] (
17:53:27 Quit bluebrother (Changing host)
17:53:27 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
17:53:35amiconnJdGordon: Why does fm skinning support increase binsize for targets without radio?
17:55:28amiconnBasically all of them (<100 bytes though), but for archos recorder it's 300 bytes
17:57:16JdGordonbecause a new tag was added for "does this target have a radio?"
17:57:17 Quit bluebroth3r (Ping timeout: 264 seconds)
17:57:51amiconnHmm, but why does this need 300 bytes on recorder, while it only needs 92 bytes on Ondio SP (also no radio)?
17:58:12JdGordonmc2739: nice! any noticable delay when changing presets?
17:58:33JdGordonamiconn: just lucky I guess
17:58:44JdGordonrun bloat-o-meter if you really care about 100 bytes
17:58:56amiconnIt's 300, not 100
17:58:56 Join webguest99 [0] (
17:59:22amiconnYou explained the ~100 (which also apply to some other radioless targets, e.g. the Minis)
17:59:30JdGordonmc2739: change "now playing" to "fm radio" (maybe) and that looks like a good new default
18:00:21mc2739JdGordon: just a slight delay when changing presets
18:00:27 Join Strife89 [0] (
18:00:52 Quit petur (Quit: gone)
18:01:12JdGordonnext step is to store more than one in the buffer so there is no delay
18:01:26JdGordonbut anyway, bed time
18:14:21 Join esperegu [0] (~quassel@
18:15:00 Quit webguest99 (Quit: CGI:IRC)
18:15:42 Quit komputes (Ping timeout: 246 seconds)
18:18:10pixelmawasn't the "does this target have radio" tag there before for sbs?
18:19:35 Quit hebz0rl (Read error: Connection reset by peer)
18:19:35 Quit soap (Ping timeout: 246 seconds)
18:22:36 Join stoffel [0] (
18:28:16 Quit pyro_maniac (Quit: Konversation terminated!)
18:29:13 Join moos_ [0] (
18:31:05 Quit moos (Ping timeout: 258 seconds)
18:31:09 Nick moos_ is now known as moos (
18:33:33 Quit mikroflops_ (Ping timeout: 240 seconds)
18:34:19 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:34:36 Quit DerPapst (Ping timeout: 245 seconds)
18:35:00 Join DerPapst [0] (
18:35:23 Join mikroflops [0] (
18:40:17 Quit komputes (Ping timeout: 258 seconds)
18:41:39 Quit n1s (Quit: Lämnar)
18:44:09 Quit DerPapst (Ping timeout: 265 seconds)
18:47:09 Join MagusG [0] (
18:49:04 Nick fxb__ is now known as fxb (
18:53:37 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:54:52 Join n1s [0] (~n1s@rockbox/developer/n1s)
18:55:09 Quit komputes (Read error: Connection reset by peer)
18:55:58 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:56:35 Quit komputes (Client Quit)
18:57:48 Join komputes [0] (~komputes@ubuntu/member/komputes)
18:58:14 Join pamaury [0] (~c2c7a50a@rockbox/developer/pamaury)
18:59:06 Quit adnyxo (Ping timeout: 246 seconds)
19:00:43 Quit flydutch (Quit: /* empty */)
19:02:06 Quit wincent_balin (Ping timeout: 245 seconds)
19:02:21 Join wincent_balin [0] (
19:11:21 Quit jgarvey (Quit: Leaving)
19:11:44 Join adnyxo [0] (
19:13:01 Join DerPapst [0] (
19:15:52 Join soap [0] (~soap@rockbox/staff/soap)
19:21:02 Join liar [0] (
19:33:47 Quit moos (Read error: Connection reset by peer)
19:34:41 Quit Forsaken (Ping timeout: 260 seconds)
19:35:14 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
19:36:36 Quit bmbl (Quit: Bye!)
19:36:44 Join bmbl [0] (
19:36:45 Quit bmbl (Changing host)
19:36:45 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
19:37:50 Join Boldfilter [0] (
19:40:03 Nick fxb is now known as fxb__ (
19:41:05 Join archivator [0] (
19:41:05 Join Forsaken [0] (~chatzilla@
19:49:22***Saving seen data "./dancer.seen"
19:54:09 Quit CGL (Quit: Soy spammero ---- \m/ d(>.<)b \m/)
19:58:28 Quit Xerion (Ping timeout: 264 seconds)
20:04:41 Quit Forsaken (Ping timeout: 260 seconds)
20:06:59 Quit Jaykay (Ping timeout: 252 seconds)
20:21:17archivatorIs it me or does the build system not parallelize anymore? make -j2 errors out with "sysfont.h missing".
20:23:21 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
20:25:39 Quit JohannesSM64 (Ping timeout: 240 seconds)
20:28:36gevaertsarchivator: most build clients use -j<some-high-number>, so it would surprise me
20:29:11archivatorbizarre. Oh, well, I'm sure it'll pass :)
20:33:27archivatorRight, back to business: flite now compiles fine for target (well, if "fine" means "with ton of warnings") but doesn't link. My new favorite error message: "region PLUGIN_RAM is full". I've seen it so many times I can't help but love it :)
20:34:10archivatorNaturally, the first section that errors out is .rodata, followed by .bss and a few messages about overlaps.
20:34:42gevaertsYou could increase the plugin buffer for now
20:34:48 Join JohannesSM64 [0] (
20:34:52gevaertsOr use overlays
20:34:58archivatorIsn't it maxed out by default?
20:36:17Torneit doesn't consume all of ram, no :)
20:36:30 Quit stoffel (Remote host closed the connection)
20:37:39archivatorHm, I was left with the impression that the plugin buffer is as big as it can possibly be.
20:39:39Torneno. it could be almost the whole of ram, if you wanted
20:39:48Tornebut then there'd be very little for audio buffering :)
20:40:10TorneTher eare some targets where the plugin buffer is already a reasonably large proportion of ram
20:40:18Torneand so making it bigger on those targets would likely be unpopular
20:40:45Tornebut it's target-specific, and the easiest way while you are still developing your plugin is to make it bigger :)
20:42:54Tornewhat target are you actually building for?
20:43:19mtIs is possible to link the standard math library to a codec in a sim build ?
20:46:52archivatorTorne: ipodvideo
20:47:01archivator32MB version
20:47:46Tornewell, it's 512kb already on ipodvideo, but you can just make it bigger for now
20:48:04Tornethere's ~28MB of ram free or so, audio buffering won't mnind losing another meg :)
20:48:35Torneyou'll probably have to do something about the size *later*, i doubt anyone's going to eb too thrilled if it needs >512KB plugin buffer when it's *done*.. ;)
20:49:24archivatormake that another 3 megs :) and I will do something about this size, I'm already looking into compression options (not really easy to do right, though)
20:49:47Tornei hope that's data, not code.. ;)
20:50:27archivatorit is. and that's the smallest it gets!
20:51:14Tornewell, data is dealable-with, generally
20:51:22Tornething with giant code are more awkward
20:52:10linuxstbarchivator: This is the "lite" version of festival? ;)
20:52:56Torne(frotz has 55KB of actual code, not even counting const data :) )
20:53:06Torne(and i've not even ported half of it)
20:53:17Torne(always seems like a lot to play text adventures)
20:53:31archivatorlinuxstb: the code is far lighter than festival. The data is virtually the same, actually. Structured differently (i.e., no Scheme!) but more or less the same.
20:53:33Torneit's only fair that making the thing talk is bigger ;)
21:00:08archivatorre: compression - this is my top choice at the moment -
21:00:57archivatorSimple, gpl2, claims to be really fast (haven't done any testing), in-place (!)
21:01:03Tornewell, is compression the answer, though? How small will the data go and still be decompressable quick enough?
21:01:51 Part watto
21:02:08archivatorTorne: no idea, I'm just brainstorming at the moment. However, the voice data can't really be trimmed any more than it already is. I mean, it *is* practically every diphone in the English language..
21:04:32linuxstbWould it make sense to partially load it into a cache, in a similar way to font glyphs?
21:04:44Torneyah, i'm just saying.. it's worth at least trying to compress it to see how generally compressable it is at *all*
21:04:58Torneobviously the rate will vary depending on the technique/algorithm/etc
21:05:21Torneand yeah, what llorean said. does it all need to be in ram at once?
21:06:08Torneer, linuxstb
21:09:59markunarchivator: I think espeak would have been smaller :)
21:10:08markun(much smaller)
21:10:15markuntoo bad of the GPL version
21:11:37archivatorlinuxstb: I'm not familiar with the way fonts are handled. The word "partially" scares me, though. If I had access to some enormous corpus and the necessary linguistic skills, I could've in theory rearranged the voice data such that you could speak the most common words from a subset of the voice. But I don't (and I can't), so there's that :)
21:11:56archivatormarkun: trust me, I know. See the report on the wiki page. However, there are other problems with eSpeak.
21:12:27markunbut I really like espeak :)
21:12:38markun(the horrible dutch output was done by me ;)
21:13:00archivatormarkun: really? The default English voice is horrendous imho. C++ doesn't help, either.
21:13:20markunwell, I believe he removed the C++ stuff (at least the templates)
21:13:25markunuses structs now
21:13:42pixelmagets a lot better with mbrola
21:13:56archivatorDefine "now". 'Cause we can't use the latest versions due to GPLv3 :)
21:13:58markunpixelma: yes, but that doesn't run on rockbox :)
21:14:14archivatorpixelma: and festival is wonderful with multisyn. Your point being? :P
21:21:22 Join petur [0] (~petur@rockbox/developer/petur)
21:21:25archivatorDear $DEITY! 6MB are not enough for this monster. This is *really* strange, considering that the respective x86_64 version is 3.2MB. The rockbox version should be even smaller as I've stripped everything but the library and even removed the other tts engine.
21:22:47mtWhy doesn't the linker complain when I use pow() for example, yet the function behaves incorrectly (just outputs zeros) ? (been working on stuff for 11 hours straight now, can't think anymore tonight !)
21:23:41linuxstbmt: Compiling for a target or the sim? IIRC, SDL has a dependency on libm, so libm is linked with the sim.
21:24:28mtlinuxstb: sim
21:24:31linuxstbDon't ask me why it's returning wrong results though...
21:24:55mtah .. that was my next question
21:26:23linuxstbmt: It sounds like you should return to it tomorrow morning - I am sure it will be clearer then...
21:27:20archivatorHm, I'm rather stupid, really. The 7 MB thing is 'cause the code is really 'bout 3 MB and then another 3 MB are the memory I allocate for tlsf + some wiggle room. Gotta shrink that heap I guess.
21:27:53mtmust .. hear .. sound :(
21:28:20linuxstbmt: So this is the original floating-point code you're trying to get to work?
21:29:31mtlinuxstb: Everything seems to be working fine now, except for those maths functions which ruin everything
21:32:46linuxstbarchivator: What did you mean in -community by "flite has preliminary utf8 support" ? Rockbox uses utf8...
21:34:22archivatorlinuxstb: It's where I'm headed next, actually. UTF-8 support looks (from a brief investigation) incomplete, I really want to make sure it works before I do anything else (even fixed-point math)
21:34:54archivatorBecause, if it doesn't, that means the voice scripts are unlikely to support it, either, which leaves us in a very bad place.
21:35:29linuxstbWhat encodings does it support?
21:37:53archivatorlinuxstb: that's the thing, I have a strong feeling (still uncomfirmed) that it's just ASCII. Some of the low-level text manipulation functions mention utf, so there's something I'm missing. I really hope it turns out that the phoneme translation is ASCII and everything else is UTF-8 but there's a chance it's far worse.
21:42:01 Join S_a_i_n_t_ [0] (S_a_i_n_t@
21:44:33 Quit S_a_i_n_t (Ping timeout: 260 seconds)
21:49:06archivatorAh, wonderful. The scripts handle utf8 by well, "exploding it into multiple byte values" (whatever that means). Since the voice data would be in this format, flite would need to do the same. Only, it doesn't. The utf8explode function's there but it's never called. Given how there are no pre-built non-English flite voices out there, I'm kinda at loss.
21:49:23***Saving seen data "./dancer.seen"
21:55:37linuxstbarchivator: Just use espeak ;)
21:57:28archivatorlinuxstb: it's becoming ever more appealing, actually.
22:00:08linuxstbarchivator: Hasn't the final GPLv2 espeak already been ported to Rockbox to some extent?
22:00:53archivatorlinuxstb: it was a GPLv3 version but yes, there's a patch that does most of the crude work
22:01:17linuxstbarchivator: I thought that a later patch on that task was a port of the gplv2 version?
22:02:47 Quit joeyg (Quit: WeeChat 0.3.0)
22:04:14linuxstbarchivator: No, seems I mis-remembered...
22:04:30archivatorlinuxstb: yeah, the last thing is just an update of the gplv3 version
22:05:00 Join apoelstra [0] (
22:05:03 Nick apoelstra is now known as joeyg (
22:06:45Gumphey, question about rockbox+sansa fuze+ubuntu
22:07:09Gumpafter about 2 weeks of using my rockbox'd sansa fuze, it became a "read only device" and i can't figure out why
22:07:29 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
22:07:29Gumpi've tried changing it with superuser privileges and it still says read only
22:07:56linuxstbGump: n1s answered you this morning - Linux mounts corrupted filesystems read-only.
22:07:56gevaertsThe filesystem is corrupted. Run fsck.vfat
22:10:06Gumpoh, thanks. yeah i was gone this morning, sorry.
22:10:11 Quit komputes (Ping timeout: 252 seconds)
22:10:19Gumpi ran fsck and no problems supposedly
22:10:35Gumpso i guess i'll just do a reformat. again.
22:10:53gevaertsDoes dmesg say anything?
22:11:23 Join komputes [0] (~komputes@ubuntu/member/komputes)
22:12:13archivatorGump: if you have access to a Windows machine, try chkdsk.
22:12:16 Quit shai (Quit: Leaving)
22:12:41Gumpsorry only been using ubuntu for a few months. what is dmesg?
22:13:00Gumpyeah, i'll try chkdsk later today. thanks for the suggestion
22:14:27linuxstbGump: "dmesg" is a command you run to display the Linux kernel error/status messages.
22:27:54archivatorEither on-target flite crashed silently/entered an infinite loop, or it's taking more than two minutes to synthesize 5 sec of audio (using floating-point math)...
22:28:48gevaertsFP is going to be horribly slow.
22:28:58gevaertsTwo minutes sounds entirely reasonable :)
22:29:22 Join aaron__ [0] (
22:30:13 Quit aaron__ (Client Quit)
22:30:27archivatorgevaerts: make that 4 now. No, something else is going on here.
22:31:01gevaertsI don't know
22:35:55 Quit JohannesSM64 (Quit: WeeChat 0.3.3-dev)
22:36:48 Quit bluebrother (Ping timeout: 260 seconds)
22:38:31 Join bluebrother [0] (
22:38:32 Quit bluebrother (Changing host)
22:38:32 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
22:44:08 Join CGL [0] (~CGL@
22:47:31mtsomething's is wrong with libm in codecs (sim)
22:49:37 Quit ender` (Quit: Programmer thinks that he needs to validate an e-mail address. "I know," he says, "I'll use a regexp!" Now the user has a problem.)
22:51:23 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
22:54:30 Quit esperegu (Remote host closed the connection)
22:55:20 Quit archivator (Quit: "no msg")
22:55:48 Join archivator [0] (
22:55:52 Part archivator
22:56:11 Join archivator_ [0] (
22:56:14 Part archivator_
22:56:55 Join Forsaken_ [0] (~chatzilla@
22:58:15mtbacktrace shows that the correct function is being called .. yet the output is _always_ -0.000 :/
22:59:10 Join Strife1989 [0] (~Strife89@
22:59:37 Nick Forsaken_ is now known as Forsaken_Boy (~chatzilla@
23:02:12 Nick Strife1989 is now known as Strife89_B (~Strife89@
23:07:55Strife89_BI intend to try and conjure a radio WPS later this evening.
23:08:27 Nick Strife89_B is now known as Strife1989 (~Strife89@
23:08:44bertriknow that there is a radio WPS, it makes more sense to work on RDS support now
23:09:48 Join anewuser [0] (anewuser@unaffiliated/anewuser)
23:13:20 Quit tchan (Quit: WeeChat 0.3.3-dev)
23:13:33 Quit Forsaken_Boy (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
23:14:52 Join Forsaken_Boy [0] (~chatzilla@
23:15:10mtWe're using glibc right ?
23:15:17 Quit xavieran (Ping timeout: 276 seconds)
23:15:26 Join robin0800 [0] (
23:15:33 Nick Strife1989 is now known as evi1nick_C (~Strife89@
23:16:46 Nick evi1nick_C is now known as Strife1989 (~Strife89@
23:18:52 Nick evilnick_B is now known as goodnick (~0c140464@rockbox/staff/evilnick)
23:19:14 Quit Strife1989 (Quit: Leaving work.)
23:20:04 Quit jfc (Ping timeout: 248 seconds)
23:20:57 Join xavieran [0] (
23:24:24 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
23:25:58 Join jfc [0] (
23:30:14 Join togetic [0] (~togetic@unaffiliated/ibuffy)
23:33:32Strife89bertrik: What do you mean, "RDS support"?
23:35:13bertrikStrife89, radio data system, some FM channels send their station name and some other info along with the FM signal, so I mean support for showing that in rockbox on the fm screen
23:35:20 Quit goodnick (Quit: Page closed)
23:35:34kenguestbertrik: do some players even support that?
23:36:19Strife89bertrik: Ah; I suppose I should try and do a design that will allow for this data?
23:36:24bertrikkenguest, there are indeed only a few players where the fm chip supports RDS, basically just the gigabeat s
23:37:00gevaertsbertrik: also the apple fm remote IIUC
23:37:25bertrikStrife89, it's no use really to make a WPS or anything before it's supported in the rockbox core
23:37:32bertrikgevaerts, ah yes, true
23:37:59kenguestbertrik: the cunning plan there would be for those players to save the data, upload it and thus have the info available for the likes of others that don't. that way users of players that don't have RDS could possibly have a list of likely stations to tune into.
23:38:08Strife89bertrik: But it would still be handy to have base WPSs for the next release.
23:38:19kenguestjust a thought!
23:39:07Strife89bertrik: Besides, I can't really contribute anything else to the project. :(
23:39:30 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
23:39:30bertrikStrife89, we could have a receiving strength indicator on quite short notice
23:39:44mtkugel : (logs) do you know anything about libm+codecs in the sim ? any libm funciton I use in a codec just returns zero (sin, cos, pow ... ). It seems to be something with the linking ?
23:40:09gevaertsStrife89: most players won't have RDS, so there's no need to wait anyway :)
23:40:47bertrikthere's a patch on FS that adds support for RSSI for most fm chips, so we basically just need some kind of tag for it and a WPS using it
23:40:59 Join bluebro [0] (~4d2c4024@gateway/web/freenode/x-eozkcwroppmdnsbu)
23:41:11 Quit bluebro (Changing host)
23:41:11 Join bluebro [0] (~4d2c4024@rockbox/developer/bluebrother)
23:42:37*Strife89 isn't sure whether he should start working on radio WPS concepts or not. :(
23:43:28bieberQuick question if any of the GSOC admins are around: do I claim copyright on my code, or assign it to Rockbox? And I'm assuming I should go ahead and copy the standard header files I see in the rest of the Rockbox source?
23:44:07gevaertsbieber: we don't have copyright assignments. The standard is useful, with your name in it
23:44:26CIA-5New commit by 03alle (r25973): Avoid duplicated code, create a formatting function for FM frequency
23:44:31bieberIs the $Id line auto-generated, or do I need to fill that out?
23:44:49bluebrobieber: that is generated. It"s a keyword substitution done by svn
23:44:54bertrikdon't fill it out
23:45:11bluebrobieber: see for details.
23:45:36bluebrobasically, you can just copy it (no need to adjust a wrong filename) and set the svn:keywords property correctly.
23:45:51bieberOkay, so I'll just write $Id$?
23:47:16 Quit adnyxo (Ping timeout: 248 seconds)
23:47:25bluebroyes, basically. I for one usually just copy an existing Rockbox file header and leave the line as-is.
23:47:45pixelmaI never saw anything else than MHz for FM frequncies (except in preset files but that's not something that will be presented usually)
23:47:55bluebroif it contains wrong information it doesn't matter at all, svn will replace that upon commit.
23:48:06pixelmafrequencies too.
23:49:09gevaertsStrife89: there's no technical reason to wait I think
23:49:24 Quit xavieran (Ping timeout: 248 seconds)
23:49:25***Saving seen data "./dancer.seen"
23:49:28bieberGot it
23:50:04Strife89Time to compile a bunch of sims, then.
23:51:28AlexPkenguest: The info is dynamic, pretty much (song/station info, that sort of thing).
23:51:55AlexPbertrik: And yes, RDS support makes lots of sense now :P (hint hint) :)
23:52:53kenguestAlexP: ah ok. i thought perhaps there'd be usable info there for configuring 'preset radio stations'
23:52:58pixelmaand how would you chose the format type for frequencies? The commit message doesn't tell :\
23:53:37AlexPkenguest: It is also used for things like alternative frequency lists, so it depends where you are
23:54:17gevaertsAlexP: there's the unique station identifier as well. That's not dynamic
23:54:18AlexPFor instance (in the UK at least), national radio stations have different frequencies in different parts of the country (so overlap etc. isn't a problem due to limited range)
23:54:44AlexPgevaerts: Yeah, but then you need to tie it to the frequency used for that station in a particular area
23:55:16AlexPFor instance Radio 2 doesn't have just one frequency, it depends if you are in London/Manchester/Cleethorpes/wherever
23:55:40gevaertsSure, but if you're in a location (and let's admit it, most people often are), you could automatically generate a usable preset file for that location
23:55:44AlexPAnyway, the point I'm making is that it is probably easier to do it by hand/get it from existing lists
23:55:46bieberbluebro: JdGordon has been talking about me writing a parser and a verifier for WPS files in C that they could integrate into the theme engine. Should I put my new code in the skin_engine directory of his skinenginebreakbranch, or just open a whole new theme editor directory somewhere and let others modify/integrate it with the rest of the engine?
23:55:46bertrikgevaerts, yeah the station name should be static (but only 8 characters IIRC), the free text that stations send is usually full of non-related info (advertisements etc.)
23:56:02*Strife89 sees no sign of (a) page(s) that list targets with radios.
23:56:06gevaertsbertrik: not the name, the ID. That's different again :)
23:56:30bieberI'd be fine with writing this part in Rockox style C with static buffers and such, if it would make integration easier
23:56:38gevaertsThe PI field
23:57:00bluebrobieber: integrate into the theme engine as in alternate parser for checkwps? (sorry, didn't got around reading the logs)
23:57:33gevaertsbluebro: JdGordon wants to replace the entire wps parser
23:57:44bieberThe impression I get is that he wants to replace the existing parser with mine once it's finished, I'm assuming so that things will stay in sync more easily
23:57:54bertrikStrife89, I made a list once:
23:57:58bluebrook. So that parser would become the Rockbox wps parser IIUC?
23:58:10gevaerts(IIUC as well :)

Previous day | Next day