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

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

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

#rockbox log for 2011-09-08

00:01:44 Quit bertrik (Quit: :tiuQ)
00:12:50 Quit lebellium (Quit: ChatZilla 0.9.87 [Firefox 7.0/20110902161802])
00:18:31kugelshouldn't r30468 be only applied for mingw builds?
00:21:37rinakuOK, I found the same question as mine in the forum... and a link to mp4muxer, which works like a charm under Wine, and everything is perfect: I can read my music on my player
00:21:51rinakun1s: Llorean: thanks a lot for your help
00:22:15rinaku(and for doing rockbox if you're both developpers :) )
00:23:07n1sno problem, it's a limitation of our aac codec
00:24:25 Join saratoga [0] (9803ec71@gateway/web/freenode/ip.152.3.236.113)
00:25:11rinakuAnother question: would it be possible to make my player boot the original firmware by default but keep the possibility to boot rockbox?
00:25:37rinaku(this is just out of curiosity, I *will* use rockbox :-) )
00:25:37n1spossible, but not very easy
00:25:45rinakuok
00:26:06 Quit liar (Quit: hallowed are the ori!)
00:26:17n1syou'd need to make a (probably) small change to the bootloader but you'd need a build environmet to build it
00:26:41rinakuok
00:27:50 Join froggyman [0] (~seth@50.105.133.91)
00:27:50 Quit froggyman (Changing host)
00:27:50 Join froggyman [0] (~seth@unaffiliated/froggyman)
00:28:46 Quit ender` (Quit: C++ is a modern language where your parent can't touch your privates but your friends can!)
00:34:27 Join Scromple [0] (~Simon@115-64-195-104.static.tpgi.com.au)
00:34:41 Join TheLemonMan [0] (~LemonBoy@ppp-219-56.26-151.libero.it)
00:45:53 Quit JesusFreak316 ()
00:46:17 Join JesusFreak316 [0] (~JesusFrea@pool-173-65-77-65.tampfl.fios.verizon.net)
00:47:18***Saving seen data "./dancer.seen"
01:00
01:00:40 Join akaWolf [0] (~akaWolf@unaffiliated/akawolf)
01:08:39 Quit FoH_Phobos (Quit: leaving the market—and net access)
01:16:32CIA-14New commit by kugel (r30469): Buflib: Clarification about invalid handles ...
01:19:09CIA-14r30469 build result: 9 errors, 0 warnings (kugel committed)
01:20:13CIA-14New commit by kugel (r30470): Remove unecessary casting
01:23:15CIA-14r30470 build result: 8 errors, 0 warnings (kugel committed)
01:32:58 Join ReimuHakurei_ [0] (~kudo@wireless.sit-co.net)
01:33:03 Quit ReimuHakurei (Read error: Connection reset by peer)
01:42:23kugeljhMikeS: the short audio stop is heavily less noticable with Q_AUDIO_REMAKE_AUDIO_BUFFER
01:56:20kugeljhMikeS: can you please have alook at http://pastie.org/2500075 ? I'm not 100% certain if I implemented it correctly (although I think I did)
02:00
02:02:16 Join soap [0] (~soap@94.75.232.194)
02:02:16 Quit soap (Changing host)
02:02:16 Join soap [0] (~soap@rockbox/staff/soap)
02:05:27 Quit robin0800 (Ping timeout: 258 seconds)
02:05:45 Join Lalufu [0] (~sun@unaffiliated/lalufu)
02:06:08 Join soap__ [0] (~soap@66.178.206.246)
02:08:09LalufuHi.
02:08:49LalufuI could need a hand by someone familiar with the PP5020 chip.
02:08:56 Quit soap (Ping timeout: 276 seconds)
02:09:05 Join robin0800 [0] (~robin0800@149.254.61.29)
02:09:18LalufuI'm trying to get IAP support running on a iPod mini 2g, and it basically works.
02:09:33 Join soap [0] (~soap@80.90.43.9)
02:09:43 Quit soap (Changing host)
02:09:43 Join soap [0] (~soap@rockbox/staff/soap)
02:10:02Lalufuproblem is that all the code assumes that the serial pins on the apple connector are routed to the SER0 on the PP5020, while on the mini2g it seems to be connected to SER1.
02:10:27 Quit soap__ (Ping timeout: 245 seconds)
02:11:06LalufuI have no idea if this is something that can be changed by twiddling the right registers on the PP502
02:11:35Lalufubut if it can all that's missing for this to work OOTB would be a line in firmware/drivers/serial.c
02:14:01JdGordonkugel: "(0 is also not a valid handle)" I tihnk adds confusion. There is no need to specify it if you already know everything returns negative numbers for error values
02:14:20kugelcore_free() returns 0
02:14:35JdGordonthats different though
02:15:03JdGordonyou wouldnt expect people to do handle = core_free()
02:15:22kugelwell, saying a negative number is an error might lead to thinking 0 is ok
02:15:38kugelJdGordon: this is the intention of returning 0
02:15:53JdGordonI dont think that is a good idea
02:16:18JdGordoncore_free should return void to stop that
02:16:34JdGordondoes close(fd) return an int?>
02:17:08JdGordonhmm, interesting
02:17:17JdGordonit actually returns 0 on success
02:17:23 Join JesusFreak316_ [0] (~JesusFrea@pool-173-65-77-65.tampfl.fios.verizon.net)
02:17:31kugelas the comment says
02:17:31 Quit JesusFreak316 (Read error: Connection reset by peer)
02:17:31kugelwhy?
02:17:31kugelno
02:18:23kugelheh, right
02:18:23kugelI'm repeatedly annoyed that free() doesn't return NULL
02:19:04 Quit robin0800 (Ping timeout: 240 seconds)
02:20:03JdGordonhow does buf_get_data cope with handle == 0?
02:20:57kugelnot at all, and I don't really want to add useless error checking there
02:21:29JdGordonso anyone doing handle = core_free() and checks for handle < 0 before opening it will crash?
02:22:08JdGordonindeed your scrobbler commit is incorrect
02:22:18JdGordoncore_free(scrobbler_cache);
02:22:20JdGordonscrobbler_cache = 0;
02:22:46kugelyes
02:22:46kugelwell, no, it probably wont even crash
02:22:59kugelwhich one?
02:23:13JdGordonsorry, maybe not, viewing on webvcs is a bit hard
02:23:38JdGordoneven #define CORE_UNDEFINED_HANDLE 0 and using that would be ok
02:23:50JdGordonactually no
02:24:36JdGordonto keep it simple everywhere core_free() should return void, and 0 shuold be technically legal but never actually used
02:24:59kugelhow's that keeping it simple?
02:25:20JdGordonyou *always* check against <0
02:26:26kugelmost of the time I check ">0" actually. the return of alloc is "<0"
02:29:02JdGordon" (0 is also not a valid handle)" is still confusing
02:29:27kugelit sounds very clear to me
02:29:41JdGordonof course it does, you wrote it
02:29:49JdGordonthere is a reason authors use editors
02:30:44JdGordon"#define CORE_HANDLE_UNALLOCATED 0" to at least make it explicit then
02:31:04JdGordonerror < CORE_HANDLE_UNALLOCATED < legal handle
02:31:44 Join robin0800 [0] (~robin0800@genld-218-237.t-mobile.co.uk)
02:32:11JdGordonAlso, the first comment block in buflib.h talks about callbacks for when the data moves, but doesnt talk about how moving happens, or what causes it
02:32:41JdGordonIf I didnt know a yield could cause a move I'd almost certainly cause a crash on my first attempt to use buflib
02:34:32 Join kugel_ [0] (~kugel@rockbox/developer/kugel)
02:35:13kugel_because that's implementation details
02:35:13kugelbecause that's implementation details
02:35:13 Quit Zarggg (Read error: Connection reset by peer)
02:35:14 Quit kugel (Read error: Connection reset by peer)
02:35:14 Join Zarggg [0] (~zarggg@24.229.139.169.res-cmts.sm.ptd.net)
02:35:25kugel_it's mentioned that you need to re-get the pointer after yield()
02:35:40JdGordonwhere is the user documentation for it?
02:35:53jhMikeSkugel: the param shouldn't be necessary for audio_fill_file_buffer. the only place it can be trashed is if playback was stopped or a remake message was received, which make it trash it itself
02:36:06kugel_JdGordon: at buflib_get_data()
02:37:14JdGordonok, so once again I ask, How does one go about loading a file into a handle pointer without worrying about it moving data on me?
02:38:01 Quit akaWolf (Ping timeout: 250 seconds)
02:38:03kugel_lock the buffer (return BUFLIB_CANNOT_MOVE in the move callback while within read()/write())
02:38:10kugel_or use a temp buffer
02:38:37jhMikeSthe voice pcm out buffer ought to use buflib instead of being handled in playback. is there any sort of "unmovable allocation" that won't hurt anything?
02:38:59kugel_jhMikeS: no
02:39:13kugel_unmoveable does hurt
02:39:29kugel_more than one of those even more so
02:40:02JdGordonthe comments in that file are really badly layed out. Don't take that as a personal offense or disregard it because its me saying it...
02:40:03jhMikeSnot that it's required, voice_thread would just have to deal with it properly in that case
02:40:43*JdGordon still fails to see why there is no buflib_un/lock_handle() helper function
02:41:26*jhMikeS would have gone with "if obtaining a pointer from a handle, use it, then release it to make it movable"
02:42:23kugel_jhMikeS: you can do that..
02:42:26JdGordonactually yeah, that is probably better, always lock the handle in buf_get_data() and require and ungetter to pretty much guarentee it cant crash it
02:44:05kugel_you can implement a get&lock-use-release scheme just fine, it's done in many places
02:44:11jhMikeSI hacked that into buffer.c, sort of, with buf_pin_handle so that atomic codecs don't have to be scared of shrink ops
02:44:26kugel_it's not the default because it's desired to design the code so that locking isn't needed at all
02:44:35JdGordonkugel_: sure, but why implement it half a dozen times when it can be done by the lib?
02:44:50jhMikeSthen we'll never escape compulsive cooperative threading :\
02:45:13kugel_we wont anyway
02:45:26jhMikeSbunk
02:46:36jhMikeSI've been pondering quite some time. some hotspots would need some reworking but it should be possible and definity better for an app to run directly on the os
02:46:46 Quit milk (Read error: Connection reset by peer)
02:47:01kugel_we don't need that
02:47:17kugel_we're not going to have apps that we don't have complete control over
02:47:23***Saving seen data "./dancer.seen"
02:47:45JdGordonwe never know 100% what code we are calling will yield or not
02:47:46jhMikeSI'm talking about running hosted
02:48:39kugel_JdGordon: I did it very pessimistic, I only did not lock for mem* and str* functions basically
02:48:39jhMikeSJdGordon: sometimes it changes, sometimes it varies with the target.
02:48:56JdGordonkugel_: ?
02:49:23kugel_for all otehr functions I assumed "may yield now or later"
02:49:43JdGordonimplictly locking will also do absolutly nothing if there is no yielding
02:49:51JdGordonso it is a simple clever safety mechanism
02:50:27kugel_it adds noise for releasing when locking wasn't needed in the first place
02:51:35*kugel_ also wants to note that he didn't design buflib that way, I just added other bits to it
02:51:56JdGordonadding 1 line in the dozen places it is used is not unmanagele noise
02:52:17JdGordonok good, so you wont be offended if someone just does this change anyway then?
02:52:18kugel_I rather have a pair of lock()/unlock() rather than a get_data() and a forgotten release()
02:52:49JdGordonit is exactly the same
02:52:52kugel_the former is the better pattern for this
02:53:18jhMikeSgarbage collection! :)
02:53:23kugel_I object if you want to change this
02:53:26JdGordonif you know for sure the code around your usage doesnt yield then the function is going to be small enough to not have buf_release() get in the way
02:53:37kugel_the implicit lock just helps your laziness
02:53:40JdGordonif you *dont know for sure* you add 2 calls
02:54:14kugel_which is not even sure since locking isn't needed in many many cases
02:54:20jhMikeSwell, you don't know for sure what will happen deep inside if you give up control to some other api but only when it's fairly simple and regulated
02:54:28JdGordonexactly
02:54:37JdGordonyou really should be locking on every access
02:54:45JdGordons/you/callers/
02:55:03kugel_what did I just say? "I only did not lock for mem* and str*"
02:55:30kugel_but actually half of the buffer uses are for just these
02:56:13jhMikeSkugel: you did catch my comment about the playback patch?
02:56:42kugel_jhMikeS: and did you catch that you can implement locking just fine, it's just that I object to the implicit lock?
02:56:45kugel_:=
02:56:46kugel_:)*
02:56:56*jhMikeS slaps kugel
02:57:08jhMikeSyes, sir!
02:57:25kugel_jhMikeS: the one that the param isn't needed?
02:57:31jhMikeSyeah
02:57:53jhMikeSif it's trashed from anywhere else, it also crashing :)
02:58:15 Quit kadoban (Read error: Connection reset by peer)
02:58:34 Join kadoban [0] (~kadoban@ip98-165-177-158.ph.ph.cox.net)
02:58:40kugel_well, i was a) to lazy to analyze the other 3-4 callers, and b) the non-noalloc version makes a buflib_alloc call which shouldn't be done if it's alraedy allocated
02:59:33kugel_and the other way around it shouldn't be ommitted if it's not allocated yet
03:00
03:00:04kugel_since the other 3-4 callers worked with doing the alloc() call now, I assumed it's not alloced when they're called
03:00:16jhMikeSaudio_play_start will alloc it, other callers call that only when already in operation
03:00:57jhMikeSspurious messages are properly rejected, since playback doesn't know what order they'll come in advance
03:01:10JdGordonkugel_: Unhelpful originoally did buflib right?
03:01:29*Unhelpful runs the hell away ;)
03:01:53JdGordonchicken! :)
03:02:03Unhelpfuli did, what's the question?
03:02:19kugel_jhMikeS: not sure I understand that
03:02:39JdGordonwhat your take on implictly locking the buffer in buf_get_data() for safety
03:03:20*kugel_ disagrees that the implicit lock adds safety
03:03:37JdGordonyou are just wrong
03:03:58jhMikeSkugel_: not sure it's imporant so much as, to go to anything but initialized, audio must be stopped. to start audio, it must be initialized, which audio takes care of just before needing the buffer
03:04:16JdGordonnothing stops someone unknowingly causing a yield and trashing the pointer
03:04:21kugel_lock()/unlock() pair is far more readable, and with the implicit lock you can forget the release (even mroe since the the aquire isn't explicit)
03:04:57jhMikeSif it's trashed at any time during playback, well, that's a bug, one I'm pretty sure doesn't exist
03:05:01kugel_JdGordon: alright. enough discussed
03:05:05Unhelpfulkeeping in mind that when i designed this there was basically one thread that could be manipulating buffers? :P
03:05:09kugel_tired of your attitude really
03:05:48JdGordonUnhelpful: indeed, but now its a free for all
03:05:49kugel_jhMikeS: trashed isn't the same as unallocated, is it?
03:07:03jhMikeSas far as playback is concerned, yes, someone else has been using that area for any purpose but playback
03:07:48kugel_what if someone did audio_get_buffer()? trashed but still allocated then
03:07:54jhMikeSwhat the heck it's still doing in audio_fill_file buffer, I should kick myself
03:08:20Unhelpfuli think i have no idea how we ensure that somebody else doesn't fail a buffer because a buf_get_data locked things and compact can't run. :/
03:08:51jhMikeSthe other calls are purely dynamic rebuffering of tracks while playing
03:10:08JdGordonUnhelpful: so the point is that if you dont need the lock doesnt affect anything, if you *do* yield then your pointer will be invalid when you come back
03:10:16 Quit robin0800 (Ping timeout: 264 seconds)
03:10:24JdGordonallocations dont necesarrily fail when locked
03:11:04JdGordonarg, "if you dont yield the lock doesnt ..."
03:11:26Unhelpfulright. but if you don't yield you are needlessly blocking compaction...
03:11:57JdGordononly if you get stuck in a loop
03:12:18JdGordonbut usages of the handle pointer should be short lived
03:13:08kugel_jhMikeS: so it should always call noalloc()?
03:14:01 Nick JesusFreak316_ is now known as JesusFreak316 (~JesusFrea@pool-173-65-77-65.tampfl.fios.verizon.net)
03:14:02kugel_it cannot call audio_reset_buffer() unconditionally, because it would then free/re-alloc the buffer within the shrink_callback
03:14:15kugel_(potentially)
03:14:59 Join robin0800 [0] (~robin0800@genld-217-237.t-mobile.co.uk)
03:14:59Unhelpfuli guess i don't really see a big pressure in either direction - if locking is implicit, you *may* lock when you don't need to, and you have to worry about making sure you balance get/release. if locking is explicit, you never have a needless lock, but must remember to lock before yield (or re-get after), and to unlock anything that you lock. yes?
03:15:24JdGordonyes
03:15:40JdGordonwith the point that locking when unnecesarily is less bad than not locking when it is required
03:15:54jhMikeSkugel_: just get the memory managment out of audio_fill_file_buffer altogether, move it to audio_start_playback just before that call
03:16:12*JdGordon would just add buf_get_and_lock(handle)
03:16:26JdGordonthen its not magically locking
03:17:01Unhelpfuli guess i just don't see that as being as big a concern as having to be careful about your lock balancing, whether the locks are implicit or explicit. that will itself require more care than minding what you might call that could yield.
03:18:21jhMikeSI guess the idea is that you can control your own code much better than another's
03:19:44kugel_jhMikeS: you're saying the other callers don't need it?
03:20:04 Quit robin0800 (Ping timeout: 240 seconds)
03:20:30Unhelpfulyou're concerned about not being able to reasonably know what could yield?
03:20:38kugel_playback.c isn't my area of expertise
03:20:54JdGordonUnhelpful: I think the possibility of using an invalid pointer outwieghs the "difficulty" of making sure you unlock correctly
03:21:22JdGordonand I'm more concerned that buflib users will forget that reqwuirement
03:23:11kugel_jhMikeS: I'd be happy if you play with this more :)
03:23:33jhMikeSkugel_: the param? no; the buffer, by the design of the engine must exist already for everyone but audio_start_playback. logically, one can know that by the fact that stop must be done to lower it's status from initialized to something else
03:24:53jhMikeSany other caller must be called after playback is started, ergo, any other caller must have the buffer in hand
03:25:06kugel_I meant the memory management part in fill_file_buffer
03:25:33jhMikeSyeah, it should just get out of there
03:25:56kugel_if you moved it I'd feel a lot more easy
03:26:22 Join robin0800 [0] (~robin0800@149.254.60.168)
03:29:08 Quit TheLemonMan (Quit: WeeChat 0.3.5)
03:30:45jhMikeSwhat I had in mind: http://pastie.org/private/mcghwnpcmuwr4d9c3bnx5a
03:37:39kugel_hm, that's it?
03:37:51kugel_well, I trust you on that :)
03:40:31jhMikeSthat should be it
03:41:53 Quit MethoS- (Read error: Connection reset by peer)
03:43:49*jhMikeS goes back to recall just why it was ever there, if the older revision could even sanely hint at such a thing
03:44:22 Quit robin0800 (Quit: Leaving)
03:45:05kugel_jhMikeS: as for when to realloc. it's pretty safe to realloc always, except from within that shrink_callback.
03:45:11kugel_when is it unsafe to not realloc?
03:45:54kugel_perhaps just ask if (audio_handle <= 0) realloc... else noalloc()
03:47:22jhMikeSnow it's my responsibility? :)
03:48:31 Quit pamaury (Remote host closed the connection)
04:00
04:02:10 Quit amiconn (Disconnected by services)
04:02:11 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:02:23 Quit pixelma (Disconnected by services)
04:02:25 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:02:27 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:02:31 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:05:19jhMikeSarighty - http://pastie.org/private/bu4htdg5ti20zqt50pmqaw
04:06:52 Join robin0800 [0] (~robin0800@genld-224-239.t-mobile.co.uk)
04:11:34*jhMikeS is ever twitchy about voice knowledge still existing inside playback.c when it's really now in effect two playback engines
04:12:43jhMikeStwo - between voice and audio that is
04:16:45jhMikeSbtw, maybe radio or any other source should get a volume cut during voice playback. it's easier now since voice just pushes the hint, which wasn't the case before
04:22:34jhMikeSthe old, old way of swapping codecs for voice just seems so surreal now ... :)
04:24:52 Quit robin0800 (Quit: Leaving)
04:28:22 Join robin0800 [0] (~robin0800@genld-219-248.t-mobile.co.uk)
04:32:41 Join Horscht [0] (~Horscht@p57B57130.dip.t-dialin.net)
04:32:41 Quit Horscht (Changing host)
04:32:41 Join Horscht [0] (~Horscht@xbmc/user/horscht)
04:33:08 Quit Horschti (Ping timeout: 258 seconds)
04:35:03 Quit saratoga (Quit: Page closed)
04:36:23 Quit TheSeven (Disconnected by services)
04:37:16 Join [7] [0] (~TheSeven@rockbox/developer/TheSeven)
04:47:27***Saving seen data "./dancer.seen"
04:55:52 Join tjb0607_ [0] (~quassel@208-100-128-206.bendbroadband.com)
04:56:34 Quit tjb0607 (Read error: Connection reset by peer)
04:59:37 Quit pixelma (Disconnected by services)
04:59:38 Quit amiconn (Disconnected by services)
04:59:39 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:59:40 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:59:42 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:59:59 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
05:00
05:04:34 Quit JesusFreak316 (Ping timeout: 252 seconds)
05:26:38 Quit Horscht (Quit: Verlassend)
05:39:45 Join Rob2222 [0] (~Miranda@p4FFF2A71.dip.t-dialin.net)
05:43:03 Quit Rob2223 (Ping timeout: 245 seconds)
05:48:53 Quit factor (Ping timeout: 245 seconds)
05:56:48 Quit jhMikeS (Ping timeout: 245 seconds)
06:00
06:02:01 Join factor [0] (~factor@74.197.205.204)
06:10:58 Join mr_boring [0] (4471cf4d@gateway/web/freenode/ip.68.113.207.77)
06:12:31mr_boringmy clip+'s database seems to need re-initializing much too often. can anyone help?
06:18:11JdGordonrockbox's or the sansa one?
06:19:18mr_boringrockbox :)
06:25:10JdGordonok, I dont really know the db well, it asks you to init each time you go into the db menu or something?
06:27:13mr_boringalmost. not quiet that predictable. often once it's already initialized properly, been played with, shut down and turned back on.
06:29:22mr_boringsometime it hangs at "[total # of] files found," sometimes at "0 files found"... just tonight it worked correctly after much struggling, then once it was done i went browsing, and it still said "database is not ready."
06:29:24 Join kevin128 [0] (~kevin@69-12-206-218.dsl.dynamic.sonic.net)
06:29:57kevin128Hey, is it possible to have rockbox read a txt file to me via speech synth?
06:30:05JdGordonnot currently
06:30:29kevin128that would be an awesome feature. i wonder if someone could hack the Espeak synthesizer to work as a plugin
06:31:12kevin128i know Espeak is supposed to be very lax on the system requirements so it should be possible
06:31:16JdGordonTTS in the core is something we want, we cant do it because the TTS we want to use is gplv3 and we dont want to do that
06:31:27JdGordonit might even be espeak that was looked at
06:41:45LloreanIt was espeak
06:42:20 Quit robin0800 (Quit: Leaving)
06:47:28***Saving seen data "./dancer.seen"
06:48:46 Quit Keripo1 (Ping timeout: 258 seconds)
06:51:15 Quit mr_boring (Quit: Page closed)
07:00
07:01:45 Join wodz [0] (~wodz@87-206-240-131.dynamic.chello.pl)
07:04:39wodzRegarding espeak - I am interested in porting it. I have done some preliminary steps. There is port of an ancient espeak version as a plugin (still maintained outside of FS) but I would strongly prefer to port recent version with minimal changes to make our life easier later with sync.
07:05:21JdGordondo i remember correctly that someone emailed the author asking to relicense to gplv2+?
07:05:22wodzThe problem with endianess seems to be matter of compile option + proper generation of dictionaries.
07:06:03wodzJdGordon: yes and the answer was that he doesn't see the point of changing license
07:07:20wodzThe BIG drawback of straight port is the memory requirement. Espeak loads all helper data to ram which takes ~600kB.
07:07:53JdGordonthats fine on most targets :)
07:08:14wodzthis will exclude the most popular sansas
07:08:25JdGordonis it all actually required?
07:08:52JdGordoncan we do the same trick hwcodec does with voice and just load it from disk as needed?
07:09:07JdGordonthose targets are flash so disk access is fast enough
07:09:23wodzFrom the first glance it's hard to tell. I was thinking of just what you said
07:09:41wodzsomething like memmap()
07:10:32wodzThere are a few places which uses floating point calc (but this is a minor issue)
07:10:33JdGordonbut if the license is an issue why bother?
07:11:44wodzJdGordon: We agreed at DevCon that we can accept v3 code BUT if there will be possibility to conditionally compile pure v2 binary. That means some HAVE_GPLV3 protect define or something
07:12:12wodzwith plugin that's fairly easy
07:12:22JdGordonthats going to be fun to manage!
07:13:20wodzI guess that enough v3 code which make us thinking about license change
07:13:35wodzbut this is far future if ever
07:17:00 Quit wodz (Quit: Leaving)
07:36:31LloreanJdGordon: Just for the record, I think mandatory code review is a good idea in an environment where people are really dedicated to making it work.
07:37:03LloreanMy main concern (I wouldn't call it an objection) is that I don't feel our environment has enough people to do the review, and qualified to review certain parts of the code, and it'll just create a larger bottleneck.
07:37:15LloreanI don't think I communicated that as well as I'd have liked to.
07:39:57 Quit God_Eater (Ping timeout: 252 seconds)
07:53:52 Join God_Eater [0] (93722cc8@rockbox/staff/GodEater)
08:00
08:03:52 Join liar [0] (~liar@clnet-p09-185.ikbnet.co.at)
08:07:07 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
08:07:32 Join wodz [0] (~wodz@87-206-240-131.dynamic.chello.pl)
08:08:33wodzkugel_: would you mind updating http://www.rockbox.org/wiki/AllocatingRAM (preferably with examples)
08:09:59 Quit powell14ski (Quit: powell14ski)
08:10:25 Join powell14ski [0] (~powell14s@c-174-51-194-6.hsd1.co.comcast.net)
08:10:40wodzeghm, *could you update
08:17:58 Join B4gder [0] (~daniel@www.haxx.se)
08:17:58 Quit B4gder (Changing host)
08:17:58 Join B4gder [241] (~daniel@rockbox/developer/bagder)
08:19:09 Join LinusN [0] (~linus@giant.haxx.se)
08:19:54 Join Scr0mple [0] (~Simon@115-64-195-104.static.tpgi.com.au)
08:23:39CIA-14New commit by buschel (r30471): MIPS uses GCC 4.1.2 which was wrongly detected as supporting '-fgnu89-inline'. Now only set this parameter if GCCNUM > 401.
08:23:55 Join Buschel [0] (~chatzilla@p54A3A9A3.dip.t-dialin.net)
08:24:02 Quit Scromple (Ping timeout: 260 seconds)
08:27:07CIA-14r30471 build result: 92 errors, 0 warnings (buschel committed)
08:27:27ZagorBuschel: sorry, that's my fault
08:27:32ZagorI'll rerun the build
08:28:21Buschelyesterday CIA said my commit was fine even though it was not, I hope now it doing the opposite :)
08:28:50wodzwhen speaking about mips - there is patch rotting on tracker about impossible asm constraint with newer gas/gcc
08:28:53Zagoryesterday was a database bug. today I added a bad build client.
08:30:24CIA-14r30471 build result: All green
08:31:53 Join nick-p [0] (~nick@82-69-105-120.dsl.in-addr.zen.co.uk)
08:32:07Buschelgotta go to work now
08:32:10 Quit Buschel (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110902133214])
08:32:59 Join ender` [0] (~ender@foo.eternallybored.org)
08:36:02 Quit powell14ski (Quit: powell14ski)
08:47:32***Saving seen data "./dancer.seen"
08:59:43CIA-14r30471 build result: All green
09:00
09:07:49 Join Scromple_ [0] (~Simon@115-64-195-104.static.tpgi.com.au)
09:08:22 Quit Scromple_ (Client Quit)
09:10:35 Quit Scr0mple (Ping timeout: 252 seconds)
09:34:51 Quit wodz (Quit: Leaving)
09:54:51 Quit God_Eater (Quit: Page closed)
09:56:21 Join God_Eater [0] (93722cd1@rockbox/staff/GodEater)
10:00
10:08:04 Join einhirn [0] (~Miranda@bsod.rz.tu-clausthal.de)
10:09:09 Join evilnick_ [0] (~evilnick@5ad33e3b.bb.sky.com)
10:09:09 Quit evilnick (Read error: Connection reset by peer)
10:11:54 Quit evilnick_ (Read error: Connection reset by peer)
10:32:20 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
10:36:22JdGordonusing buflib for background images in the skins is going to be *interesting*
10:36:54JdGordonevery time the buffer moves it will require a lcd_set_backdrop() :/
10:37:16gevaertsSounds like something that wants to lock the buffer then
10:38:22gevaertsOf course the buffer may move seldom enough for this not to be a major issue
10:39:24JdGordonthis is something which absolutly shldnt lock the buffer :)
10:39:39JdGordonthe allocated size is pretty large and could cause big fragmentation
10:40:03JdGordonactually, after typing that second line I dont tihnk it wil be so problematic
10:40:36JdGordoneach handle is indivdually "locked" right?
10:41:04amiconnkugel_, bluebroth3r: When building mingw32-sdl as per wiki instructions, it isn't prefixed. You don't need to change path either; 'configure' figures out which 'sdl-config' is the correct one from 'sdl-config −−libs' output
10:41:57 Quit mudd1 (Read error: Operation timed out)
10:42:21amiconnThe mingw32 one has '-mwindows' while the native one does not.
10:43:40JdGordonerr... how the heck do I go about adding a new block to a handle?
10:43:53*JdGordon is very scared by union buflib_data
10:44:38gevaerts"adding a new block"?
10:45:50gevaertsActually, I suspect we want one or two more callbacks. The current move callback is called *before* moving, which means you can't do things like call lcd_set_backdrop() from it
10:46:24JdGordonI'm going to do regular images first which is easy
10:46:26amiconnZagor: ping
10:46:33Zagoramiconn: pong
10:46:52gevaertsAnd moving that call to *after* the move will make it impossible to deny moving from it, which also isn't too good
10:47:10*gevaerts decides that kugel's input is needed here
10:47:12JdGordonbut I think this locking mechanism is way overkill for most users, I want to add a simple bool to the handle for locked and then buf_get_and_lock(handle) and buf_unlock(handle) to make that work and ignore the callbacks
10:47:35***Saving seen data "./dancer.seen"
10:47:43amiconn[15:52:06] <amiconn> B4gder: Where in svn can I find the build colouring script? And if it's not in svn atm, could you add it there?
10:47:43amiconn[15:52:32] <amiconn> It needs fixing...
10:47:54amiconnZagor^
10:51:28 Nick kugel_ is now known as kugel (~kugel@rockbox/developer/kugel)
10:51:52kugelyou can do lcd_set_backdrop() from the callback
10:52:10kugelit only sets the pointer, and doesnt draw from it
10:52:17JdGordonwhy oh why are the block numbers not coming from an enum? :'(
10:52:20kugelJdGordon: go add this thing, I'll still object
10:52:27JdGordongood for you
10:52:43kugelI suggest you go reading logs where I considered adding lock mechanism into buflib and discussed it with gevaerts
10:53:01kugelit's not like you're the first one with this idea
10:53:13JdGordondid I say i thought that?
10:53:21JdGordonI think you'll find I was in that conversation also
10:53:39Zagoramiconn: http://svn.rockbox.org/viewvc.cgi/www/buildserver/showbuilds.pl?view=log
10:53:47kugelthere was another one where you were absent, IIRC
10:54:44*JdGordon waiats a link
10:54:46JdGordonawaits*
10:55:38JdGordonunless someone else wants to go to the efffort of mpving the 400K skin buffer acrosss?
10:56:07kugelyou want to make it unmoveable entirely?
10:56:26JdGordonno
10:57:03amiconnZagor: That one does the build table; I'm looking for the script that's colouring the individual build logs
10:57:10Zagoroh
10:58:03gevaertskugel: right, and there shouldn't be a yield between the callback and the actual move I guess, so things are indeed fine, at least for this case
10:58:36kugelJdGordon: I explicitely decided against making having locking in buflib (it's not meant to provide that since it wants things moveable). I would prefer if you respect this decision
10:58:48kugelor rather, gevaerts and I did
10:59:57JdGordonits alot easier to argue against a wrong descision than a right one....
11:00
11:00:22kugelit's right in my book
11:01:27CIA-14New commit by zagor (r30472): Added build log viewer to svn.
11:01:57Zagoramiconn: http://svn.rockbox.org/viewvc.cgi/www/buildserver/shownewlog.cgi?view=log
11:02:20JdGordonI'm not convincesd, and i cant find the logs
11:03:24kugelI also don't want multiple ways of locking, but a single one
11:03:51JdGordonbuflib isnt your baby
11:04:30 Quit kevin128 (Quit: Leaving.)
11:04:54gevaertsCan we evaluate the current system the way it stands, and discuss changes if and when it *actually* appears that the current API has problems?
11:06:28JdGordonI have issues wth it right now. Instead of a simple function call I have to implement the whole callback system and then keep track of wheather its ok to move or not now
11:06:53kugelyou always need to keep track
11:07:04JdGordonnot to mention the overhead of the callbacks being called and the binsize that would incur
11:07:24kugeland yes, you need to implement callbacks (that's even a wanted hurdle) because not requiring locks at all is the default
11:07:50JdGordonpsudeo code for what i want: { alloc_buffer(); lock_buffer(); load_bmp(get_buf_data()); unlock;}
11:07:59JdGordon5 lines which you're turning into 3 functions
11:08:13JdGordonnot requireing locks means someone is not going to lock it when it needs to be
11:08:33kugelyou can implement it that way in your .c file
11:08:44kugeljust don't make it into buflib
11:09:34kugelimplicit locking means someone is not going to release the lock, making the whole buflib effort useless
11:09:34JdGordondinner time
11:09:50JdGordonhow is get_data_and_lock() implicit?
11:13:16kugelhave you written any working code by now? or are you complaining without having even used the api?
11:14:27 Join thomasjfox [0] (~tomj@rockbox/developer/thomasjfox)
11:15:27 Join petur [0] (~petur@rockbox/developer/petur)
11:18:05 Quit God_Eater (Quit: Page closed)
11:23:00 Quit thomasjfox (Remote host closed the connection)
11:25:29 Join casainho [0] (~chatzilla@pal-213-228-181-14.netvisao.pt)
11:25:55CIA-14New commit by zagor (r30473): Sanitize parameters.
11:38:43kugelZagor: can you look at r30153? that commit was reverted later but the issue is still there
11:39:50 Quit petur (Read error: Connection reset by peer)
11:57:28 Quit casainho (Remote host closed the connection)
11:58:13 Join casainho [0] (~chatzilla@pal-213-228-181-14.netvisao.pt)
12:00
12:00:13JdGordongevaerts: out of curiosity, have you had a play with your skin list idea yet?
12:01:06gevaertsJdGordon: not yet. I've been a lot busier than expected these last two days
12:04:49JdGordonok
12:06:02 Join God_Eater [0] (93722cc8@rockbox/staff/GodEater)
12:06:43n1sdoes this new sking thingy make it possible to replace all lists?
12:07:05JdGordonis it guarenteed nothing will yield between getting the move callback and the move completing?
12:07:15JdGordonn1s: all except multi-line-selector lists
12:07:39n1sJdGordon: aha
12:07:55gevaertsJdGordon: it's like that in the current implementation, and I'm pretty sure breaking that will cause a *lot* of pain, so I'd say yes
12:08:00JdGordonactually, I'm not even sure if I left that check in the code or not, so it might work
12:09:10JdGordongevaerts: it would be nice if that guarentee was in the existing code so it doesnt change by someone later
12:09:35gevaertstrue
12:09:58n1sJdGordon: is the selector bar customizable as well?
12:11:20JdGordonn1s: with a bit of skin trickery you can do pretty much whatever you want. each item draws all the viewports with a specified label
12:11:25JdGordonso images, colouring, whatever
12:11:39JdGordonthe skin engine needs some drawing primatives to make it better though
12:11:48JdGordoni.e draw rect, line, etc
12:12:17n1scool, so i could draw a custom image as a selector?
12:12:43JdGordonsure
12:15:07 Join TheLemonMan [0] (~LemonBoy@ppp-153-16.26-151.libero.it)
12:20:14 Join chu [0] (~ma@115.74.72.185)
12:20:17chuHi
12:20:24chuI have a gigabeatS
12:20:36chuwhen I run drictly from AC adapter
12:20:52chuthe rockbox report the volage 2.4V
12:21:50chuI measure the voltage output from the adpater is 5.35V
12:22:16chuso I wonder if that is a problem with my player
12:22:43chuor do rockbox report inaccurate ?
12:23:07God_Eaterrockbox reports the *battery* voltage - not the voltage being supplied by the power adaptor.
12:23:15chuah
12:23:27kugelJdGordon: sure (although not written explicitely)
12:23:45JdGordonkugel: is there a helper macro to move a cached pointer?
12:24:56kugelno, not yet, but tagtree.c has one
12:25:37JdGordonseems filetypes.c also has one... not yet means its coming?
12:25:46*JdGordon realised thugh he can use new directly here though
12:25:53kugelsomeone could ad it
12:35:49 Join user890104 [0] (~Venci@89.252.237.109)
12:38:26*JdGordon wonders what the val in the buflib debug menu means
12:39:56kugelthe same the val member of union buflib_data means. number of union buflib_data to skip past the end of the allocation
12:40:09kugelso it relates to the size
12:41:04JdGordonit would be "nice" if selecting the handle would give a full description of the handle
12:41:35 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:41:51JdGordondoes http://pastebin.com/ExZV9ibh look correct?
12:42:48 Quit God_Eater (Ping timeout: 252 seconds)
12:42:52JdGordonoh poo... [saint] how would you like a tag to change the background image dynamically in skins?
12:43:28JdGordonhmm, core_* really shold just use malloc for hosted targets
12:43:43JdGordonseen as using the 8mb audio buffer is a waste for them
12:45:12JdGordoncool! adding in the core_free() call in that patch killed it :)
12:46:28 Quit chu ()
12:47:38***Saving seen data "./dancer.seen"
12:52:40 Join soap__ [0] (~soap@66.178.206.246)
12:52:46 Quit soap__ (Client Quit)
12:53:13 Join Jerom1 [0] (~jerome@2a02:8420:216:a800:f66d:4ff:fe45:790f)
12:55:39 Join God_Eater [0] (93722cd0@rockbox/staff/GodEater)
12:55:54*JdGordon just loves memory leaks :)
12:56:38 Quit soap (Ping timeout: 264 seconds)
12:59:27*JdGordon thinks there may be a serious mem leak in skin backdrops in svn
13:00
13:00:43JdGordonhmm, actually no, its safe, only because of the terrible buffer managemnt there already
13:00:45 Join wodz [0] (~wodz@iwl138.internetdsl.tpnet.pl)
13:00:50JdGordoncausing mem leaks moving to buflib though :/
13:02:39 Quit Jerom1 (Quit: Leaving.)
13:02:53Zagorkugel: updated now. web site commits needs to be manually pulled by a swede.
13:03:29CIA-14New commit by zagor (r30474): Rename $gecko to $push and increase block to later firefox versions.
13:07:41 Quit nick-p (Quit: Leaving)
13:08:30JdGordonkugel: I'm getting a crash in buflib_get_data() with handle==13
13:09:27kugelwell, that's not nice
13:13:56JdGordonkugel: http://pastebin.com/GbVQuFUG see anything obvious?
13:14:23JdGordonit all loads fine the first time, reloading a theme crashes it
13:16:37LalufuIf I have a patch regarding logf.{c,h}, what category would I file that under?
13:16:53kugelJdGordon: I don't see anything obviously wrong
13:16:58 Quit wodz (Read error: Operation timed out)
13:17:04JdGordonhaha oops
13:18:22 Join wodz [0] (~wodz@iwl138.internetdsl.tpnet.pl)
13:18:41kugelyou don't need to init handles with -1, just leave them 0 and check for >0 everywhere else
13:20:53n1sLalufu: i'd say operating system
13:22:25Lalufun1s: thanks
13:24:32CIA-14New commit by jdgordon (r30475): Use buflib for the skin backdrop images potentially allowing multiple images per skin in the future
13:26:36kugelJdGordon: do you want to make each skin object buflib aware?
13:26:37 Join evilnick_ [0] (~evilnick@92.40.254.205.threembb.co.uk)
13:27:07CIA-14r30475 build result: 225 errors, 1 warnings (jdgordon committed)
13:27:37JdGordonkugel: no, no need, only the big items, i.e images
13:27:54JdGordonway too many pointers to make it worth the effort
13:28:10JdGordonoh FFS
13:28:21 Quit evilnick (Read error: Operation timed out)
13:28:43JdGordondeltas are pretty nice :p
13:29:00kugelthe skin ram usage info doesn't include backdrops now does it?
13:29:28JdGordonnot anymore it doesnt
13:29:34JdGordonit did before that commit
13:31:16 Quit evilnick_ (Ping timeout: 252 seconds)
13:34:09CIA-14New commit by jdgordon (r30476): fix checkwps
13:35:45kugelJdGordon: why don't you add the missing bits to checkwps?
13:35:46 Join jordan` [0] (~gromit@2001:660:3302:2826:225:90ff:fe20:d9a8)
13:35:58kugelwith each #ifdef checkwps gets more useless
13:36:10kugeland the code uglier
13:36:24 Quit kadoban (Ping timeout: 252 seconds)
13:36:39CIA-14r30476 build result: All green
13:37:43JdGordonthe first statement is nonesense
13:38:29kugelit's not
13:39:08JdGordoncheckwps is only to check the language parses
13:39:10kugelin r30463 checkwps doesn't parse some stuff anymore
13:39:12JdGordonso yes, it is incorrect
13:40:54kugelthere are other ifdef like this in parsing
13:41:50JdGordongo ahead and fix them if they annoy you
13:42:01JdGordonyou'll quickly realise there is a danm good reason for them
13:42:16JdGordonI *wont* have a hissy fit if you do that
13:42:42kugelI just asked you about the reason for that
13:43:57JdGordonmaybe its different for german, but in english questions usuually end with a ?
13:45:07kugel"13:36 < kugel> JdGordon: why don't you add the missing bits to checkwps?"
13:46:04JdGordonsorry, I didnt see that line :/
13:46:19JdGordonbecause they are unecessary, the parserlib does it
13:47:12kugelbut the other parser can reject tags outside the parserlib?
13:47:24JdGordonyes
13:47:42JdGordonthat particular !checkwps block doesnt rejct anything
13:47:54JdGordonand the others are for when it is impossible for checkwps to know
13:48:54kugelokay
13:49:57kugelbut I still think r30476 could have added buflib to checkwps to avoid #ifdef
13:50:40JdGordonif you want to do it go for it, I think that is completly pointless though
13:51:54JdGordononce I move image loading into buflib I wonder what a sensible size for the remaining required buffer is
13:52:02JdGordonit doesnt really make sense to base it on the screen size
13:55:20kugelthe screen size might still make sense (as in, a larger screen can fit more info, and this more LOCs in the .wps), but rather loosely
13:58:29JdGordoneach tag is something like 12 bytes+param data... I think 2K is probably more than enough
13:58:39JdGordonmaybe 4K
14:00
14:03:46 Quit user890104 ()
14:05:40 Join nick-p [0] (~nick@82-69-105-120.dsl.in-addr.zen.co.uk)
14:10:28 Join user890104-freen [0] (~user89010@2a01:4f8:101:2a4:0:bc28:b2e1:9)
14:10:47 Nick user890104-freen is now known as user890104 (~user89010@2a01:4f8:101:2a4:0:bc28:b2e1:9)
14:11:04 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
14:18:58 Join WalkGood [0] (~WalkGood@unaffiliated/walkgood)
14:19:56JdGordone00-sim with cabbie loaded is now 28KB/241KB :)
14:20:00JdGordone200-sim*
14:20:07JdGordonstill seems too high
14:21:07JdGordonah, the font
14:21:17JdGordonhmm. thats going to be tricky to move :/
14:22:22kugelcabbie loads an additional font?
14:23:17 Quit user890104 (Quit: .)
14:23:38 Join user890104 [0] (~Venci@2a01:4f8:101:2a4:0:bc28:b2e1:9)
14:24:15JdGordondont think so
14:24:23JdGordonbut still, nfi how to put the font in buflib
14:24:49JdGordongot the buffer down to 94K on e200 from the origional 400 or so which i tihnk is a good days work
14:28:08 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
14:37:12 Quit jhMikeS (Read error: Connection reset by peer)
14:38:15CIA-14New commit by jdgordon (r30477): make some stuff static
14:38:23CIA-14New commit by jdgordon (r30478): Use buflib for skin images. Allows much more images to be loaded
14:39:08JdGordon^ annoyingly simple... doesnt mean having to implement the lock mechanism myself isnt a bad thing
14:40:03 Join jhMikeS [0] (~jethead71@c-68-61-166-99.hsd1.mi.comcast.net)
14:40:03 Quit jhMikeS (Changing host)
14:40:03 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
14:41:02CIA-14r30477 build result: All green
14:41:41JdGordonthat commit is going to crew android which used to use malloc for all those images
14:43:25CIA-14r30478 build result: All green
14:43:39 Quit casainho (Ping timeout: 276 seconds)
14:44:35CIA-14New commit by nls (r30479): FS #12257 by Michael Goerner, fixing stone colours in goban for inversed displays (clips and mrobe100).
14:45:31 Quit B4gder (Quit: Konversation terminated!)
14:46:35JdGordonamiconn: any thoughts on if it would be possible to use buflib with fonts? or because of its current implementation its just Too Damn Hard with all the pointers?
14:47:01JdGordonthat would free up a heap f statically allocated ram which is often unused
14:47:19CIA-14r30479 build result: All green
14:47:40***Saving seen data "./dancer.seen"
14:47:46Torneare there actually many pointers when the whole font is loaded?
14:48:07Tornethe font cache is complicated, but if we used buflib we could, in many cases, just allocate a buffer big enough to load without the cache being used.
14:48:35JdGordonwhat about unifont?
14:48:50Tornedepends how much ram you have :)
14:48:54JdGordonactually im not sure if the cache even uses pointers
14:49:13TorneIt has some kind of header bit that has at least some addresses in it
14:49:18Tornebut most of the cache is just an array, yes
14:49:25 Quit wodz (Read error: Operation timed out)
14:49:27Torne(iirc)
14:49:52JdGordonif there are only a few pointers which need updating it wont be a problem to move
14:50:20JdGordonkugel: any plans for a plugin to visualise the buffers state with all the current allocs?
14:51:10kugelvisualize how?
14:51:47JdGordona literal map of the allocations... |<name>|<name>|empty|<name>|....|
14:52:21JdGordoni hope the allocation algorithm is correct :) I've just double the normal alocations
14:52:36kugelthe debug menu doesnt do that?
14:53:22JdGordonthe debug menu is truly useless.... maybe if val had a meaning it would be better but just an address and a name isnt very useful
14:53:38kugelval relates to the size, as I said
14:53:47kugelessentially, multiply it by 4
14:53:50 Join wodz [0] (~wodz@iwl138.internetdsl.tpnet.pl)
14:54:03kugelit is a map of the buflib buffer
14:59:54 Join zerly [0] (~91f8c301@www.haxx.se)
15:00
15:01:05 Quit wodz (Quit: Leaving)
15:04:04 Quit zerly (Client Quit)
15:07:48kugelJdGordon: did you try changing themes during playback also?
15:08:16JdGordondoes it work?
15:08:23kugeldont know
15:08:42kugelI mean it should, but I didn't try
15:08:48 Quit EvanCarroll (*.net *.split)
15:08:49 Quit scorche|sh (*.net *.split)
15:08:49 Quit crwl (*.net *.split)
15:08:50 Quit rasher (*.net *.split)
15:08:50 Quit preglow (*.net *.split)
15:08:50 Quit feisar-_ (*.net *.split)
15:08:50 Quit merbanan (*.net *.split)
15:08:50 Quit Xerion (*.net *.split)
15:08:50 Quit user829385 (*.net *.split)
15:08:50 Quit Zagor (*.net *.split)
15:08:51 Quit Torne (*.net *.split)
15:08:52 Quit ruskie (*.net *.split)
15:08:52 Quit aevin (*.net *.split)
15:08:52 Quit otih (*.net *.split)
15:08:52 Quit Guest33080 (*.net *.split)
15:08:53 Quit [fred] (*.net *.split)
15:08:54 Join preglow [0] (thomj@tvilling2.pvv.ntnu.no)
15:08:54 Join feisar- [0] (jljhook@ihq.in)
15:08:58 Join aevin [0] (eivindsy@unaffiliated/aevin)
15:08:58 Join scorche|sh [0] (~scorche@squisch.net)
15:09:07 Join Torne [0] (~torne@lil.wolfpuppy.org.uk)
15:09:08 Quit Torne (Changing host)
15:09:08 Join Torne [0] (~torne@rockbox/developer/Torne)
15:09:13 Join EvanCarroll [0] (~ecarroll@64.200.109.13)
15:09:15JdGordonif it doesnt its either because i messed up the locks or buflib doesnt work correctly
15:09:18 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
15:09:19 Join merbanan [0] (~banan@c-62-220-165-114.cust.bredband2.com)
15:09:22 Join Xerion [0] (~xerion@5419A766.cm-5-2c.dynamic.ziggo.nl)
15:09:33 Join crwl [0] (~crwlll@dsl-jklbrasgw1-ffb9c300-103.dhcp.inet.fi)
15:09:34 Join ack [0] (~ack@88.191.116.126)
15:09:42 Join user829385 [0] (~aoeu@112.166.15.141)
15:10:01 Nick ack is now known as Guest35783 (~ack@88.191.116.126)
15:10:04 Join otih [0] (~otih@0x01.f2tec.de)
15:10:04 Quit otih (Changing host)
15:10:04 Join otih [0] (~otih@CAcert/Assurer/otih)
15:10:55jhMikeSkugel: commit what all the hub-bub was about last night for playback.c?
15:11:57kugeljhMikeS: you think it's fine that way, yes?
15:12:39jhMikeSthe final version should work. I'm going to build it and actually run it through a few check though (which is happening now)
15:21:14 Join [fred] [0] (fred@ircop.efnet.at)
15:29:02 Join rasher [0] (~rasher@rockbox/developer/rasher)
15:30:46 Join JesusFreak316 [0] (~JesusFrea@131.247.39.75)
15:35:04jhMikeSit's being - interesting. it doesn't resume properly if restarted from shrink_callback but instead restarts from the beginning of the track. it does work for cuesheet and crossfade
15:39:40 Quit krazykit (Remote host closed the connection)
15:39:47 Join krazykit [0] (~quassel@li135-248.members.linode.com)
15:42:06JdGordonis core_shrink() used anywhere?
15:42:11 Join y4n [0] (y4n@unaffiliated/y4ndexx)
15:42:24JdGordonwould bad things happen if that always returns false?
15:43:35jhMikeSit's called in playback.c
15:46:56JdGordonhttp://pastebin.com/ZDWkWFAr is my patch to make APPLICATION builds use host malloc for everything except playback.. unfortunately it panifc()'s playback and backdrop images dont load...
15:46:58*jhMikeS isn't sure how core_shrink could work safely if playback is still going
15:47:03JdGordona second set of eyes would be nice
15:47:17kugeljhMikeS: it resumes from the start if it's called repeatedly
15:47:34kugelI tried with enabling scrobbler and timestretch
15:48:01kugelworks fine with scrobbler (1 alloc), but starts from the beginning for timestretch (4 subsequent allocs)
15:48:02jhMikeSthe codec probably isn't finished
15:48:24kugelJdGordon: no, nothing bad
15:48:44kugelcore_shrink is all or nothing. if it returns false then everythign is as before
15:48:56JdGordonok
15:49:06kugelit's most prominently used in the shrink_callback in playback.c, but also for tagtree
15:49:47kugelis the comment unclear about it?
15:49:58jhMikeSlooks like it depends on the target then, which means it needs to not lose the offset if the codec hasn't completed starting by the next restart
15:50:36 Join ruskie [0] (ruskie@sourcemage/mage/ruskie)
15:50:43kugeljhMikeS: this is also observable in svn with teh plain stop/start calls
15:51:35jhMikeSyou're saying resume is being dodgey?
15:52:33 Join MethoS- [0] (~clemens@134.102.106.250)
15:56:25kugeljhMikeS: yea
15:57:03kugeljhMikeS: if the stop/start pair is too frequently it'll "resume" from the beginning
15:58:29jhMikeSnot surprising, and I'm not sure what to do about it. the codec gets around to updating things whenever it does and it may never get around to it between calls
15:59:44kugelbecause the codec thread doesn't run?
16:00
16:00:01JdGordonhmm, ok, after playback does core_alloc_maximum() how do the following allocs work?
16:00:36kugelbuflib asks playback to get rid of some by calling teh shrink callback
16:00:45jhMikeSkugel: or not enough to get back into a decoding loop
16:01:01kugelif there's not enough free space gained by compaction, that is
16:01:40kugeljhMikeS: can stop block until resume info is ready?
16:03:14kugelanyway, it's only a minor annoyance if you ask me
16:03:15jhMikeSit might never be ready depending upon what transpires
16:03:31jhMikeSit's clunky though
16:03:56JdGordonmy debug output makes no sense... I can see the shrink happening untill the 16th alloc and because it is not happening it fails
16:03:57jhMikeSand not what's intended
16:04:01kugelperhaps re-use the previous resume info?
16:04:11JdGordonis 16 a magic number? or a fluke?
16:04:47kugelmagic numbers shouldn't be involved
16:05:03jhMikeSit's there, in playing_id3, at least initially, then gets clobbered after the first restart
16:06:26 Join stoffel [0] (~quassel@p57B4BB59.dip.t-dialin.net)
16:06:45JdGordonwhats the playbacks buffer callback function name?
16:06:51jhMikeSonce control is given to the codec, the codec master of elapsed and offset
16:06:57jhMikeS*is
16:08:15JdGordonwhat the heck? it just stops calling the shrink callback :/
16:09:04kugelperhaps because the alloc succeds without it?
16:09:12JdGordonno, it fails
16:09:41JdGordonand this is with size=4
16:09:54kugelsize=4?
16:10:13kugelwhat do you need such a tiny alloc for?
16:10:26JdGordonhttp://pastebin.com/G4JMvu8q the log of the run
16:10:47 Quit JesusFreak316 (Ping timeout: 260 seconds)
16:10:51JdGordonsee the pastebin before, use host malloc on app builds
16:12:24kugelcan we post-pone this work until things are settled with buflib?
16:12:36 Quit antil33t (Read error: Connection reset by peer)
16:12:55 Join antil33t [0] (~antil33t@203-100-223-143.callplus.net.nz)
16:13:08kugelI use RaaA exclusively these days so I basically only test through this, and don't want a separate code path now already
16:13:33JdGordonits not being commited yet
16:13:38kugeljust saying
16:14:07kugelbut, fwiw, I'd just implement the buflib api with a malloc backend instead of what that diff is doing
16:15:13*jhMikeS would like the id3 to stop being used directly for positioning and instead use a function call to ask for the position
16:17:02JdGordonkugel: well, doing it in core_* means it keeps buflib.c completly clean
16:17:43kugelI have a buflib-with-malloc.c in my mind
16:18:25kugeland since you can call buflib directly (e.g. in plugins) it makes more sense to back buflib with malloc instead of core_* IMO
16:18:30jhMikeSkugel: fine, if nothing's gotten worse, commit that stuff, or I will and I'll think about what makes sense in regard to hardening the resume information against rapid starts and stops
16:18:36JdGordoni might give up anyway... no way im going to dive into buflib.c to figure out why it isnt working
16:19:03JdGordonkugelhoq would you do alloc_maximum() though?
16:19:32kugelmalloc(MEMORYSIZE)
16:20:30kugelor the size the context was initialized with
16:21:25JdGordonwont that leave an 8MB buffer not being used at all?
16:21:44JdGordonerr, where buffer_alloc() ends
16:22:15kugelwhat do you mean?
16:22:35kugelthe static buffer[MEMORYSIZE] would be removed
16:22:42 Join JesusFreak316 [0] (~JesusFrea@2607:fe50:0:8100:4dd6:5391:f627:d523)
16:24:54 Join JesusFreak316_ [0] (~JesusFrea@2607:fe50:0:8100:4dd6:5391:f627:d523)
16:26:49 Quit JesusFreak316_ (Client Quit)
16:27:19 Join JesusFreak316_ [0] (~JesusFrea@2607:fe50:0:8100:4dd6:5391:f627:d523)
16:27:40 Quit JesusFreak316 (Ping timeout: 264 seconds)
16:27:42 Nick JesusFreak316_ is now known as JesusFreak316 (~JesusFrea@2607:fe50:0:8100:4dd6:5391:f627:d523)
16:29:17JdGordonI'm not completly sure buflib is actually working
16:30:00kugelwell, it shouldn't fail to shrink after the 15th alloc
16:30:06kugelI'll have a look
16:30:23JdGordonim looking into fs#12265
16:30:24fs-bluebothttp://www.rockbox.org/tracker/task/12265 Sansa Clip+ cannot display cabbiev2 wps at r30478 (bugs, unconfirmed)
16:31:28CIA-14New commit by kugel (r30480): Buflib/swcodec: Improve playback stopping behavior on allocation. ...
16:31:57JdGordonwhat does Buffer: in rockbox info point to now days?
16:32:45kugelthe audiobuf if allocated, the what's available in buflib otherwise
16:32:58kugelwhat audio_buffer_available() returns, specifically
16:34:05CIA-14r30480 build result: All green
16:36:32JdGordoncore_allo_ex() is only being called once as expetced for each of my handles, yet they are appearing quite a few times in the debug menu
16:36:57kugelin svn or with that diff?
16:37:05JdGordonsvn
16:37:49jhMikeShow does the buffer shrink work if playback is still using the buffer?
16:37:52CIA-14New commit by kugel (r30481): Make genlang faster by doing better regexes. ...
16:38:30kugelhm, that Torne's work, git-svn dropped the author of course :(
16:38:57kugeljhMikeS: ?
16:39:43kugelour patch was about exactly that
16:39:52TorneWait, I thought that got committed ages ago :0
16:40:16jhMikeSkugel: it calls core_shrink then restarts playback
16:40:23kugelyes
16:40:24CIA-14r30481 build result: All green
16:40:56kugelgevaerts: I think re-running the build scores is in order
16:41:12jhMikeSkugel: it can't change anything there as far, not safely
16:41:26jhMikeS*as far as overwriting playback's data
16:41:36kugelwhat?
16:41:41gevaertskugel: I'll try to remember doing that in the weekend. That's several hours of hot laptop :)
16:42:04Tornekugel: i had an idea for how to do that even faster but haven't gotten around to trying it yet. it may not actually be faster.
16:42:05jhMikeSkugel: is there any chance something would be able to overwrite the audio buffer there?
16:42:57Tornekugel: profiling genlang properly turned out to be annoying and i got bored of messing with perl
16:42:57kugeljhMikeS: you said it's safe :)
16:43:13kugeljhMikeS: it should use the new buffer when audio_start_playback() returns, no?
16:43:32kugelTorne: the profiling I did was quite convinient
16:43:42Tornekugel: The profiling I did *looked* convenient
16:43:50Tornebut i found it was extremely unstable
16:44:00Torneand with miniscule almost-noop changes to the source, it would radically change the profile
16:44:09Torneer, not "and" there.
16:44:45jhMikeSkugel: within the fill_file_buffer it is, that part
16:45:39jhMikeSthe restart will task switch unless on the audio thread
16:45:51kugelyes
16:46:12kugelnot sure what you mean
16:46:37JdGordonkugel: can you grab svn and run the clip+ sim in gdb and watch it crash? it crahses on core_free() on a valid handle
16:46:48kugelcan do
16:46:54jhMikeSthere are enough yield points inside the audio thread that, well, I don't know what it's doing at the time
16:47:03kugel5
16:47:10kugels/5//
16:47:35kugelthe shrink_callback() is called on the thread that does the alloc
16:47:42***Saving seen data "./dancer.seen"
16:48:11JdGordonkugel: skin_parser.c:1495 is where it gets interesting
16:48:26kugelplus, core_alloc() can't return before the shrink_callback returned
16:48:34JdGordonthe first few images allocate fine, then one doesnt allocate and the core_free() for the valid ones all crash
16:48:37jhMikeSis the alloc always at the end?
16:48:38kugelso the new alloc shouldn't be able to overwrite audio buffer
16:51:16jhMikeSwhat if reset_buffer is called while another thread is waiting inside shrink_callback?
16:52:01 Join JesusFreak316_ [0] (~JesusFrea@2607:fe50:0:8100:4dd6:5391:f627:d523)
16:53:29CIA-14New commit by kugel (r30482): Exit of SDL_WaitEvent() returns an error. ...
16:53:39kugelthat annoyed me long enough
16:54:25 Quit JesusFreak316 (Ping timeout: 260 seconds)
16:55:40kugelread_bmp_fd() returns tiny sizes for me
16:55:45kugel50 bytes, etc
16:56:00JdGordonvery small images
16:56:00CIA-14r30482 build result: All green
16:56:18JdGordoncore_free() shouldnt be crashing though
16:56:19 Quit JesusFreak316_ (Read error: Connection reset by peer)
16:56:29JdGordonit looks like something is trashing buflib
16:57:39 Join JesusFreak316 [0] (~JesusFrea@2607:fe50:0:8100:4dd6:5391:f627:d523)
16:59:21 Quit nick-p (Quit: Leaving)
16:59:22kugelhmm, the shrink callback is called more than once for the allocs?
16:59:48JdGordonwho are oyu asking?!
17:00
17:00:26kugelnobody :)
17:00:44*JdGordon mutters something about missing unit tests
17:00:54kugelah
17:01:01kugelthere is indeed a magic number involved
17:01:30*JdGordon cries a little
17:02:02kugelbuflib_available() reserves space for 16 future handles
17:02:45kugelI that's what BUFLIB_SHRINK_POS_BACK is about, if there's not enough room to store the handles (the integer)
17:03:49JdGordon16 byutes for the name? thats a bit wrong
17:04:12 Quit otih (Excess Flood)
17:04:12 Quit amiconn (Excess Flood)
17:04:14 Quit Unhelpful (Excess Flood)
17:04:19 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:04:20 Quit niekie (Excess Flood)
17:04:24 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
17:04:39 Quit amiconn (Excess Flood)
17:04:51 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:05:07 Quit balintx (Excess Flood)
17:05:10 Quit amiconn (Excess Flood)
17:05:16 Join otih [0] (~otih@0x01.f2tec.de)
17:05:16 Quit otih (Changing host)
17:05:16 Join otih [0] (~otih@CAcert/Assurer/otih)
17:05:17 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:05:25kugelI need to go, will look later
17:05:26 Quit Unhelpful (Excess Flood)
17:05:26 Join balintx [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
17:05:28 Quit [7] (Excess Flood)
17:05:34 Quit otih (Excess Flood)
17:05:35 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
17:05:39 Join niekie [0] (~niek@CAcert/Assurer/niekie)
17:05:41 Quit amiconn (Excess Flood)
17:05:50 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
17:06:10 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
17:06:14 Quit amiconn (Excess Flood)
17:06:29*jhMikeS suggests not changing filebuflen off-thread
17:06:36 Quit Unhelpful (Excess Flood)
17:06:39 Join otih [0] (~otih@0x01.f2tec.de)
17:06:39 Quit otih (Changing host)
17:06:39 Join otih [0] (~otih@CAcert/Assurer/otih)
17:06:40 Quit JesusFreak316 (Ping timeout: 260 seconds)
17:06:44 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
17:06:51 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
17:06:51 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:06:56 Quit balintx (Excess Flood)
17:07:31 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
17:07:32 Quit benedikt93 (Read error: Connection reset by peer)
17:07:56 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
17:07:58 Join balintx_ [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
17:08:25JdGordonwell, its after 1am here, and its not my bug so im going to bed. If there is lots of complaints someone please revert all my commits tonight
17:09:08JdGordonits not crashing all targets. very annoying anyway
17:09:11 Quit niekie (Excess Flood)
17:09:12 Quit y4n (Quit: PANTS OFF!)
17:09:39kugelpjhMikeS: before buflib filebuflen wasn't a global at all, should be safe
17:09:49 Join simonlnu_ [0] (n1orJxlc1b@unaffiliated/simonrvn)
17:09:59 Join y4n [0] (y4n@unaffiliated/y4ndexx)
17:10:15 Nick balintx_ is now known as balintx (~quassel@szerver1.gulyasp-koll.sulinet.hu)
17:10:30 Join niekie [0] (~niek@CAcert/Assurer/niekie)
17:11:17 Join tjb0607 [0] (~quassel@208-100-128-206.bendbroadband.com)
17:11:17 Join Lalufu_ [0] (~sun@ryoko.camperquake.de)
17:11:18 Join desowin_ [0] (~desowin@ip-69-162-133-250.static.chi2.systeminplace.net)
17:11:35 Join rinaku_ [0] (johan@nibox.dgmil.net)
17:12:19 Join rasher_ [0] (~rasher@rockbox/developer/rasher)
17:13:19 Join ack [0] (~ack@mingbai.org)
17:13:35jhMikeSit was based upon some globals
17:13:46 Nick ack is now known as Guest64469 (~ack@mingbai.org)
17:16:17jhMikeSI suppose I shouldn't worry about this without removing some further dependencies with voice
17:16:41 Join maraz_ [0] (maraz@kapsi.fi)
17:16:46 Quit ps-auxw (Disconnected by services)
17:16:48 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
17:17:01 Join ps-auxw [0] (~arneb@p4FF7FE6B.dip.t-dialin.net)
17:18:10 Quit rasher (*.net *.split)
17:18:23 Quit Guest35783 (*.net *.split)
17:18:23 Quit user890104 (*.net *.split)
17:18:23 Quit tjb0607_ (*.net *.split)
17:18:23 Quit froggyman (*.net *.split)
17:18:23 Quit fs-bluebot (*.net *.split)
17:18:23 Quit Farthen (*.net *.split)
17:18:23 Quit simonlnu (*.net *.split)
17:18:23 Quit parafin (*.net *.split)
17:18:23 Quit ender| (*.net *.split)
17:18:23 Quit Lalufu (*.net *.split)
17:18:23 Quit desowin (*.net *.split)
17:18:23 Quit maraz (*.net *.split)
17:18:23 Quit Hadaka (*.net *.split)
17:18:23 Quit rinaku (*.net *.split)
17:18:23 Nick rasher_ is now known as rasher (~rasher@rockbox/developer/rasher)
17:18:23 Nick simonlnu_ is now known as simonlnu (n1orJxlc1b@unaffiliated/simonrvn)
17:18:47 Quit otih (Excess Flood)
17:18:47 Quit Unhelpful (Excess Flood)
17:18:48 Quit amiconn (Excess Flood)
17:18:48 Quit niekie (Excess Flood)
17:18:51 Quit TheSeven (Excess Flood)
17:18:52 Quit balintx (Excess Flood)
17:18:55 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:18:56 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
17:19:06 Join balintx [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
17:19:19 Quit amiconn (Excess Flood)
17:19:21 Quit balintx (Excess Flood)
17:19:36 Join balintx [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
17:19:38 Join user890104 [0] (~Venci@2a01:4f8:101:2a4:0:bc28:b2e1:9)
17:19:50 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
17:19:50 Join otih [0] (~otih@0x01.f2tec.de)
17:19:54 Quit otih (Changing host)
17:19:54 Join otih [0] (~otih@CAcert/Assurer/otih)
17:19:55 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
17:19:55 Quit balintx (Excess Flood)
17:19:56 Quit Unhelpful (Excess Flood)
17:19:59 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:20:04 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
17:20:06 Join niekie [0] (~niek@CAcert/Assurer/niekie)
17:20:09 Quit otih (Excess Flood)
17:20:21 Quit amiconn (Excess Flood)
17:20:28 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:21:02 Quit Unhelpful (Excess Flood)
17:21:13 Join otih [0] (~otih@0x01.f2tec.de)
17:21:14 Quit otih (Changing host)
17:21:14 Join otih [0] (~otih@CAcert/Assurer/otih)
17:21:38 Join Farthen [0] (~Farthen@2a01:4f8:101:2a4:0:bc28:b2e1:9)
17:21:45 Join balintx [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
17:22:07 Quit TheSeven (Excess Flood)
17:23:02 Join froggyman [0] (~seth@unaffiliated/froggyman)
17:23:09 Join parafin [0] (parafin@paraf.in)
17:23:25 Join ender| [0] (~ender1@foo.eternallybored.org)
17:23:25 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
17:23:39 Quit niekie (Excess Flood)
17:23:50 Join fs-bluebot [0] (~fs-bluebo@g226068205.adsl.alicedsl.de)
17:24:02 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
17:24:11 Join Naked [0] (~naked@naked.iki.fi)
17:24:24 Nick Naked is now known as Hadaka (~naked@naked.iki.fi)
17:24:59 Join niekie [0] (~niek@CAcert/Assurer/niekie)
17:26:15 Quit antil33t ()
17:26:41 Quit balintx (Read error: Connection reset by peer)
17:26:43 Quit tjb0607 (Read error: Connection reset by peer)
17:26:45 Quit otih (Read error: Connection reset by peer)
17:26:45 Quit pixelma (Remote host closed the connection)
17:26:45 Quit amiconn (Write error: Connection reset by peer)
17:26:55 Quit Unhelpful (Read error: Connection reset by peer)
17:28:16 Quit niekie (Read error: Connection reset by peer)
17:29:58 Join amiconn [0] (quassel@rockbox/developer/amiconn)
17:29:58 Join pixelma [0] (quassel@rockbox/staff/pixelma)
17:31:08 Quit TheSeven (Ping timeout: 240 seconds)
17:31:35jhMikeSthink I'll have to have codecs update their offset explicitely as well as elapsed instead of inferring it from ci.curpos automatically
17:34:26 Part Zagor
17:34:56 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
17:36:40 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org)
17:38:14 Join Keripo [0] (~Keripo@165.123.49.253)
17:42:12 Nick Lalufu_ is now known as Lalufu (~sun@ryoko.camperquake.de)
17:42:24 Quit Lalufu (Changing host)
17:42:24 Join Lalufu [0] (~sun@unaffiliated/lalufu)
17:43:05 Quit Keripo (Client Quit)
17:46:43 Quit user890104 (Quit: .)
17:46:43 Quit Farthen (Quit: ZNC - http://znc.in)
17:47:15 Join user890104 [0] (~Venci@2a01:4f8:101:2a4:0:bc28:b2e1:9)
17:47:45 Join Farthen [0] (~Farthen@2a01:4f8:101:2a4:0:bc28:b2e1:9)
17:48:24 Join lebellium [0] (~chatzilla@i02m-212-194-176-149.d4.club-internet.fr)
17:49:51 Part LinusN
17:51:53n1sjhMikeS: recently i've experienced some hangs on my h300 with voice enabled while browsing the filesystem
17:53:54pixelman1s: doesn't have to be a new thing. It sounds familiar on coldfire targets, something that hit irregularly and became more rare over time
17:54:24n1spixelma: ah
17:54:49 Join Keripo [0] (~Keripo@seas584.wireless-pennnet.upenn.edu)
17:54:50jhMikeSthere was a fix for that, but now, things work very differently, but is it in pcm or talk?
17:55:19n1sno idea, it just hangs
17:55:34jhMikeSno backlight action?
17:56:45 Join powell14ski_ [0] (~powell14s@c-174-51-194-6.hsd1.co.comcast.net)
17:59:17jhMikeSheh, playback get 5 restart messages in rapid fire when enabling timestretch (??)
17:59:28n1sjhMikeS: nope, seems to be triggered by filename spelling
17:59:51n1scompletely frozen with bl on and disk spinning
18:00
18:00:18 Quit Keripo (Quit: Leaving.)
18:02:47 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
18:03:12 Quit pamaury (Remote host closed the connection)
18:03:17jhMikeStry upping the voice stack size? that seems to be cut rather close
18:04:18jhMikeSwait, wtf. I tried it on the beast and the entire framebuffer was splashed with random pixels
18:04:44jhMikeSthe backdrop buffer got blasted
18:04:53 Join balintx_ [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
18:05:05pixelmajhMikeS: that has never been fixed completely for me. It happened recently with an around r30000 build on M5 for me. I usually have voice enabled but it didn't happen very often anymore
18:05:52jhMikeSthere is something wrong and it's not just coldfire
18:06:20jhMikeSI was testing the filename spelling btw. voice menus are alright
18:06:45 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
18:08:22 Nick balintx_ is now known as balintx (~quassel@szerver1.gulyasp-koll.sulinet.hu)
18:08:51jhMikeSon the beast with filename spelling: jhmikes.cleansoap.org/dump">http://jhmikes.cleansoap.org/dump 110908-120426.bmp
18:08:54 Quit balintx (Read error: Connection reset by peer)
18:09:09jhMikeSbah jhmikes.cleansoap.org/dump%20110908-120426.bmp">http://jhmikes.cleansoap.org/dump%20110908-120426.bmp
18:09:47n1sfunky
18:10:06Stummi.bmp? :o
18:10:07n1si can't reproduce the hang with a larger voice stack though
18:10:37 Join Keripo [0] (~Keripo@seas584.wireless-pennnet.upenn.edu)
18:10:42jhMikeScould be undetectable stack overflow then
18:10:59jhMikeSStummi: it's USB screenshot, which is .bmp :)
18:12:07 Quit factor (Read error: Connection reset by peer)
18:12:14 Quit Stummi (Remote host closed the connection)
18:17:45 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
18:18:25jhMikeSok, I guess voice menus aren't fine afterall. they were before last time I did anything with voice. it's clobbering everything now
18:20:22n1syeah i can reproduce with larger stack, i just messed up with the settings before :/
18:21:26jhMikeSI'm pointing finger now. It was fine a few days ago for me.
18:28:47 Join factor [0] (~factor@74.197.205.204)
18:29:14 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
18:31:51jhMikeSI think the audio buffer is getting changed but not getting marked as no longer valid
18:32:29jhMikeSif voice tries to run, it's still using the old buffer values
18:43:32 Quit Stummi (Remote host closed the connection)
18:44:51 Join kadoban [0] (~kadoban@ip98-165-177-158.ph.ph.cox.net)
18:46:09 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
18:46:50 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
18:47:45***Saving seen data "./dancer.seen"
18:51:35jhMikeSIf I go back to the last commit I made, the voice is alright.
18:52:23 Join WalkGood_ [0] (~WalkGood@adsl-108-132-12-123.mia.bellsouth.net)
18:52:23 Quit WalkGood_ (Changing host)
18:52:23 Join WalkGood_ [0] (~WalkGood@unaffiliated/walkgood)
18:52:50 Quit WalkGood (Disconnected by services)
18:52:56 Nick WalkGood_ is now known as WalkGood (~WalkGood@unaffiliated/walkgood)
18:54:44bluebroth3rwe have a wiki page on how to cross compile the sim with MinGW?
18:57:13jhMikeSaha, voice only gets a renew on the audio buffer if it went to state trashed, but resetting as a restart takes it down to voiced only state, meaning voice never gets it updated
19:00
19:22:39 Join Horscht [0] (~Horscht@p5DD57871.dip.t-dialin.net)
19:22:39 Quit Horscht (Changing host)
19:22:39 Join Horscht [0] (~Horscht@xbmc/user/horscht)
19:29:25 Quit Keripo (Ping timeout: 260 seconds)
19:33:06 Join Keripo [0] (~Keripo@hou450.wireless-pennnet.upenn.edu)
19:37:03CIA-14New commit by bluebrother (r30483): Fix toolchain installation. ...
19:37:07 Quit Keripo (Ping timeout: 240 seconds)
19:39:38CIA-14r30483 build result: All green
19:48:03preglowbluebroth3r: are you sure that doesn't break windows installs?
19:48:03 Quit fs-bluebot (Read error: Connection reset by peer)
19:48:03 Quit bluebroth3r (Read error: Connection reset by peer)
19:48:19CIA-14New commit by bluebrother (r30484): Improve sim win32 cross compiling. ...
19:48:21 Join bluebrother [0] (~dom@g226068205.adsl.alicedsl.de)
19:48:22 Quit bluebrother (Changing host)
19:48:22 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
19:48:48preglowbluebrother: sure that android installer change doesn't break windows/linux installs?
19:50:22bluebrotherpreglow: it doesn't break on Linux. I've tried that :)
19:50:30CIA-14r30484 build result: All green
19:50:48bluebrotherand on Windows ... well, building for Android with Cygwin is broken anyway.
19:51:00preglowand who cares about windows! \o/ :P
19:51:17bluebrotherI had a look into that some time ago but wasn't motivated too much to get it fixed. I looks like it will be quite a bit of work.
19:51:41preglowpft, sod it
19:51:56 Join fs-bluebot [0] (~fs-bluebo@g226068205.adsl.alicedsl.de)
19:52:00bluebrothersometimes I have such kind of motivations ;-)
19:52:24bluebrother(likely to happen when I'm using Windows for some Rockbox Utility thing)
19:52:41 Join Buschel [0] (~chatzilla@p54A3A789.dip.t-dialin.net)
19:52:44 Join Strife89 [0] (~Strife89@207.144.201.128)
19:57:13 Quit stoffel (Remote host closed the connection)
20:00
20:04:03 Join Horschti [0] (~Horscht@p5DD56B71.dip.t-dialin.net)
20:04:03 Quit Horschti (Changing host)
20:04:03 Join Horschti [0] (~Horscht@xbmc/user/horscht)
20:05:03 Quit Horscht (Ping timeout: 252 seconds)
20:06:02 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
20:10:51 Join Keripo [0] (~Keripo@hou450.wireless-pennnet.upenn.edu)
20:11:09Buschelhmm, since r30478 my WPS does not load anymore. cabbiev2 works fine
20:12:11 Quit Stummi (Ping timeout: 258 seconds)
20:12:15 Join robin0800 [0] (~robin0800@149.254.60.171)
20:12:18Buschelsimulator give a warning: "Unable to load font 3: 'blalbla.ftn'"
20:13:27Buscheleverything works as expected at r30477 and below
20:15:18pixelmacabbiev2 doesn't use any extra fonts, so it looks like some font loading problem
20:15:38BuschelI will add a note to FS #12265
20:15:38fs-bluebothttp://www.rockbox.org/tracker/task/12265 Sansa Clip+ cannot display cabbiev2 wps at r30478 (bugs, unconfirmed)
20:15:39 Quit Keripo (Ping timeout: 260 seconds)
20:16:00*bertrik wonders why show_logo has a return argument
20:16:16bertrikreturn value
20:16:49 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
20:17:20 Join balintx_ [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
20:20:45 Join niekie [0] (~niek@CAcert/Assurer/niekie)
20:20:53 Nick balintx_ is now known as balintx (~quassel@szerver1.gulyasp-koll.sulinet.hu)
20:21:25bertrikAlso a bit peculiar that each bootloader declares show_logo it again instead of using a common header file
20:21:54Bagderblame copy n' paste programming I'd say =)
20:22:37 Quit robin0800 (Ping timeout: 240 seconds)
20:22:38 Quit balintx (Read error: Connection reset by peer)
20:23:51 Quit niekie (Read error: Connection reset by peer)
20:24:41 Join balintx_ [0] (~quassel@szerver1.gulyasp-koll.sulinet.hu)
20:27:51 Join Keripo [0] (~Keripo@165.123.49.226)
20:28:28 Join kevin128 [0] (~kevin@69-12-206-218.dsl.dynamic.sonic.net)
20:28:30 Quit pixelma (Excess Flood)
20:28:32Buschelit seems my WPS issue is caused by too small SKIN_BUFFER_SIZE, I have attached a patch to FS #12265 which fixes this for me
20:28:33fs-bluebothttp://www.rockbox.org/tracker/task/12265 Sansa Clip+ cannot display cabbiev2 wps at r30478 (bugs, unconfirmed)
20:28:39 Join pixelma [0] (quassel@rockbox/staff/pixelma)
20:29:10kevin128so it turns out there is an espeak plugin for Rockbox, but I have to compile it. can anyone tell me where I can learn to compile *plugins* please? I'm looking on the wiki and it just talks about compiling rockbox itsself.
20:29:45 Join niekie [0] (~niek@CAcert/Assurer/niekie)
20:30:20bluebrotherkevin128: plugins are a part of Rockbox. You need to build Rockbox to be able building a plugin
20:31:16CIA-14New commit by bertrik (r30485): Remove return value from show_logo() in bootloader/
20:32:27 Quit pixelma (Excess Flood)
20:32:38 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
20:32:40 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
20:33:25CIA-14r30485 build result: All green
20:36:31kevin128so I would download and extract the plugin into a rockbox source tree and then build the whole thing for my particular device?
20:40:13 Quit pixelma (Excess Flood)
20:40:22 Join pixelma [0] (quassel@rockbox/staff/pixelma)
20:41:18 Quit niekie (Read error: Connection reset by peer)
20:41:52 Nick balintx_ is now known as balintx (~quassel@szerver1.gulyasp-koll.sulinet.hu)
20:46:21 Quit benedikt93 (Read error: Connection reset by peer)
20:47:43 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
20:47:47***Saving seen data "./dancer.seen"
20:53:20 Quit Keripo (Quit: Leaving.)
20:53:33 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
20:57:19bluebrotherkevin128: basically, yes
20:59:27 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
21:00
21:14:41 Quit Stummi (Ping timeout: 260 seconds)
21:15:09bertrikLalufu, you have some way to make iap work on a new target?
21:18:43kevin128so if the espeak folder has a MakeFie, do I stick it in /rockbox/firmware or rockbox/apps/plugins?
21:22:14 Join robin0800 [0] (~robin0800@149.254.61.29)
21:23:14 Quit amiconn (Remote host closed the connection)
21:23:14 Quit pixelma (Remote host closed the connection)
21:23:39 Join amiconn [0] (quassel@rockbox/developer/amiconn)
21:23:42 Join pixelma [0] (quassel@rockbox/staff/pixelma)
21:25:35Lalufubertrik: I am trying to get it to work on a iPod Mini 2G.
21:26:22bertrikDid you just change SER0 to SER1 so far?
21:26:33Lalufuto make serial basically work: yes.
21:27:05bertrikoh cool, then iap should work nearly automatically too I guess
21:27:52 Quit Buschel (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110902133214])
21:27:54LalufuEnd goal is to make it talk to an Audi car entertainment system
21:27:56bertrikcan you both tx and rx?
21:28:02Lalufuyes
21:28:14 Join petur [0] (~petur@rockbox/developer/petur)
21:29:11Lalufuit looks to me as if support in IMA for retrieving playlist information is a bit patchy.
21:30:19LalufuI'm currently reverse engineering what the Audi is doing, using an iPod touch as the reference
21:30:37 Join saratoga [0] (9803ec71@gateway/web/freenode/ip.152.3.236.113)
21:31:46 Part WalkGood
21:31:47 Join funman [0] (~fun@rockbox/developer/funman)
21:32:26 Quit benedikt93 (Quit: Bye ;))
21:38:11 Quit robin0800 (Read error: Connection reset by peer)
21:38:21 Join robin0800_ [0] (~robin0800@149.254.61.29)
21:39:40LalufuI'm not sure if the SER0/SER1 mapping can be fixed by poking some registers on the PP5020
21:39:51Lalufuif not the patch is going to be rather ugly.
21:40:06saratogacan't just preprocessor it?
21:40:20Lalufubut from a hardware perspective everything seems to work, now I'm wrestling with ima.c
21:40:24bertrikI was thinking along the line of using something like SER_LCR and define it as either SER0_LCR or SER1_LCR depending on the target
21:41:16bertrikThe other way is to use a base register plus offsets (so by changing the base register you can switch between ser0 and ser1) instead of direct serial port addresses
21:41:28Lalufucan be done, I think, but that's rather ugly.
21:41:46bertrikiap is already ugly ... :)
21:41:47 Quit kevin128 (Quit: Leaving.)
21:42:01funmanbertrik: so did you try code on a real clipzip?
21:42:15bertrikfunman, no I don't have a clipzip yet
21:42:32Lalufuiap is not touched by this, the changes are in serial.c and system-pp502x.c
21:42:52Lalufusystem-pp502x.c is only one line, for setting up the interrupt
21:43:37funmanbertrik: someone else maybe?
21:53:43 Join niekie [0] (~niek@CAcert/Assurer/niekie)
21:56:15 Join Keripo [0] (~Keripo@SEAS302.wlan.seas.upenn.edu)
21:59:40bertrikfunman, I'd rather not let a non-developer try experimental code, and there's no rockbox dev with the device AFAIK
22:00
22:00:11 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
22:02:57 Quit y4n (Quit: PANTS OFF!)
22:03:47 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
22:05:28 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
22:07:26 Join kevin128 [0] (~kevin@69-12-206-218.dsl.dynamic.sonic.net)
22:07:45kevin128ok, I've compiled rockbox for my D2, but the espeak plugin didn't build with it
22:08:18kevin128if i go into the espeak plugin directory and type make, i get makesrc.inc not found
22:10:18gevaertskevin128: plugins should build together with the rest of the system. Running make in random directories is going to have unpredictable and interesting results. Your problem most probably is that the espeak plugin/patch on flyspray is rather old and got out of sync with modern rockbox
22:32:06 Quit petur (Remote host closed the connection)
22:32:30 Quit Keripo (Quit: Leaving.)
22:37:49kugelaha, I think I know what's causing FS #12265
22:37:50fs-bluebothttp://www.rockbox.org/tracker/task/12265 Sansa Clip+ cannot display cabbiev2 wps at r30478 (bugs, unconfirmed)
22:38:16 Join reuttles [0] (~5cNed@69-165-173-252.dsl.teksavvy.com)
22:40:04kugelI suspect: since buflib reserves space for handles after alloc_maximum() there's free space after the audio buffer
22:40:21kugelthat's sufficiently large for the tiny monochrome bitmaps
22:40:38 Quit Stummi (Quit: Bye!)
22:41:05kugelthus, a tiny allocation sneaks in behind audiobuf
22:41:17kevin128can rockbox be set up to only turn on the backlight if I press a certain button? for example not having it turn on if I turn up/down the volume or advance to a next track
22:42:39kugelall fine, except that only audiobuf can shrink on its back to make room for handles. the tiny alloc can't shrink, and thus no room can be made for handles
22:43:24reuttlesHi, I am having problems with a Sansa Clip v1. It was playing fine, reached the end of the album and stopped, so I put it down. The next time I went to use it it would not turn on at all (no display). Connected to USB, the screen turns on as normal and is detected by the PC. I have tried the standard reset (holding power switch), updating to latest Sansa FW, reinstalling RB bootloader, and
22:43:24reuttlesdoing complete reinstall of RB, no change. I thought it might be the power switch, but when USB is attached and switch is held, Charging/Connected screen will turn off after ~10 seconds. Any suggestions?
22:43:34kugelnot sure how to solve
22:45:08reuttlesI've also tried booting to the Sansa FW using power+left dir key, that doesn't do anything either.
22:47:13kugelthe problem is that the audiobuf is unmoveable so compacting stuff behind it is likely to fail
22:47:50***Saving seen data "./dancer.seen"
22:54:32kugelbut otoh, rockbox clearly shouldn't crash if the skin engine runs OOM
22:58:04 Quit funman (Quit: leaving)
22:59:07 Join viperfang_ [0] (~viperfang@cpc2-nrwh9-2-0-cust228.4-4.cable.virginmedia.com)
22:59:27 Quit ender` (Quit: If the experiment works, you must be using the wrong equipment. -- Patrick's Theorem)
22:59:27viperfang_Good evening all
22:59:50viperfang_I have an IPod 6G that crashes whilst in USB mode... anyone heard of this?
23:00
23:04:31pixelmaI believe n1s had a similar problem and knows more but am not sure
23:07:04n1syep
23:07:33n1sviperfang_: see FS #12083
23:07:35fs-bluebothttp://www.rockbox.org/tracker/task/12083 Add delay to usb-s3c6400x init to fix usb hang on some ipod classics (patches, new)
23:07:39viperfang_I have the 128GB model, I believe Classic G2, you?
23:07:54n1ssame i think
23:08:09viperfang_Ah, so if I were compiling from svn that should be fixed?
23:08:20n1snope
23:08:47viperfang_dang
23:08:48n1si really should commit that fix though, needs a bit more testing on nano2g i think
23:09:11n1sviperfang_: if you can compile, just apply the patch and it should work fine
23:10:20viperfang_heh, just compiled, looks like another run :-)
23:17:52 Join nick-p [0] (~nick@82-69-105-120.dsl.in-addr.zen.co.uk)
23:19:34 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
23:28:53 Part Strife89 ("Vamoose!")
23:30:37 Quit ReimuHakurei_ (Read error: Connection reset by peer)
23:30:55 Join ReimuHakurei_ [0] (~kudo@wireless.sit-co.net)
23:33:45 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
23:40:32 Join nameless [0] (b8980af8@gateway/web/freenode/ip.184.152.10.248)
23:41:06 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
23:41:52 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
23:44:38namelessHello everyone. I have a slight problem, my iPod Nano 1G is stuck at "Divide by zero at 01F8055C (0)" after being in the metronome app.
23:46:03 Quit God_Eater (Ping timeout: 252 seconds)
23:52:49 Quit ReimuHakurei_ (Read error: Connection reset by peer)
23:53:05 Join ReimuHakurei [0] (~kudo@wireless.sit-co.net)
23:53:22bluebrothernameless: reset it
23:54:21bluebrothernameless: http://www.rockbox.org/wiki/IpodFAQ#How_do_I_reset_the_iPod_63
23:56:46 Quit kevin128 (Quit: Leaving.)
23:56:46namelessThank you! I shall remember this ( i think it will happen again). Bye!
23:56:49 Quit nameless (Quit: Page closed)
23:58:21 Quit Unhelpful (Read error: Connection reset by peer)

Previous day | Next day