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 2012-07-21

00:41:16 Quit n1s (Quit: Ex-Chat)
00:59:23 Quit TheLemonMan (Read error: Connection reset by peer)
01:03:51 Join TheLemonMan [0] (~LemonBoy@unaffiliated/thelemonman)
01:04:17bertrikfunman, this describes how the ogg container is used for opus:
01:04:44funmanbertrik: i just learn that opus got accepted by ietf
01:05:19funmana developer will send patches for VLC shortly, so perhaps we'll start seeing files in the wild
01:05:56 Quit GeekShadow (Ping timeout: 245 seconds)
01:06:46 Join GeekShadow [0] (
01:09:15 Join factor [0] (~factor@
01:27:22 Quit bertrik (Ping timeout: 248 seconds)
01:28:35 Quit ender` (Quit: How do you generate a random string? Put a new user in front of VIM and tell him to save and quit.-- Duane Morin)
01:35:04***Saving seen data "./dancer.seen"
02:48:58 Quit pamaury (Ping timeout: 248 seconds)
03:04:16 Quit lurch (Ping timeout: 255 seconds)
03:07:31 Quit anewuser (Ping timeout: 248 seconds)
03:10:08 Join anewuser [0] (~anewuser@
03:10:08 Quit anewuser (Changing host)
03:10:08 Join anewuser [0] (~anewuser@unaffiliated/anewuser)
03:30:46 Quit XavierGr ()
03:35:07***Saving seen data "./dancer.seen"
03:41:07 Quit mirak (Ping timeout: 248 seconds)
03:45:48 Quit TheLemonMan (Quit: WeeChat 0.3.8)
03:46:54 Quit [Saint] (Ping timeout: 246 seconds)
03:48:35 Quit agumonkey (Ping timeout: 252 seconds)
03:59:35 Join [Saint] [0] (~Saint]@
03:59:35 Quit [Saint] (Changing host)
03:59:35 Join [Saint] [0] (~Saint]@unaffiliated/saint/x-8516940)
04:03:17 Join TheSphinX_ [0] (
04:05:03 Join user485763 [0] (~aoeu@
04:05:15 Quit TheSphinX^ (Read error: Operation timed out)
04:11:49 Quit TheSeven (Disconnected by services)
04:11:58 Join [7] [0] (~quassel@rockbox/developer/TheSeven)
04:49:16 Join lurch [0] (
04:50:23 Join amiconn_ [0] (amiconn@rockbox/developer/amiconn)
04:50:23 Join pixelma_ [0] (pixelma@rockbox/staff/pixelma)
04:50:24 Quit amiconn (Disconnected by services)
04:50:24 Quit pixelma (Disconnected by services)
04:50:24 Nick pixelma_ is now known as pixelma (pixelma@rockbox/staff/pixelma)
04:50:26 Nick amiconn_ is now known as amiconn (amiconn@rockbox/developer/amiconn)
05:16:01 Quit scorche (Disconnected by services)
05:16:05 Join scorche` [0] (~scorche@rockbox/administrator/scorche)
05:16:52 Quit prof_wolfff (Ping timeout: 246 seconds)
05:27:53 Join mikeryan [0] (
05:28:17mikeryanhi, i'm having a problem with a sansa fuze v2
05:28:36mikeryanwhen i plug it into a computer via USB it locks up
05:28:45mikeryanthe graphics get corrupted similarly to this image
05:29:35mikeryani'm running 3.11.2, flashed using the automatic installer
05:29:55mikeryanUSB worked fine about a month ago, then i didn't touch the player at all
05:30:02mikeryanturned it on, still had about 75% battery
05:30:21mikeryanevery time i plug it into usb i get the same problem
05:30:33mikeryani've tried different USB ports on two different machines
05:30:58 Quit uwe_ (Ping timeout: 245 seconds)
05:31:34 Join uwe_ [0] (
05:32:12 Quit TheSphinX_ (Ping timeout: 245 seconds)
05:35:10***Saving seen data "./dancer.seen"
05:37:17 Join TheSphinX^ [0] (
05:37:41mikeryani've solved it!
05:37:52mikeryanif i switch the theme, it does not freeze
05:37:55mikeryanquite odd
05:38:01 Part mikeryan
06:20:40 Quit Totalled (Quit: PETTAN PETTAN, TSURUPETTAN!)
06:20:59 Join Totalled [0] (
06:35:41 Join Keripo [0] (
06:40:24 Quit Keripo (Ping timeout: 260 seconds)
06:46:17 Join Keripo [0] (
07:22:06 Quit Keripo (Ping timeout: 245 seconds)
07:26:54 Join [Saint_] [0] (~Saint]@unaffiliated/saint/x-8516940)
07:27:48[Saint_]kugel: I'm done playing with the database/autoresume directory selection ui, ...its awesome.
07:30:07[Saint_]On another note: How difficult would it be to use the awesome effect that pictureflow uses in its menus in the main menu on color targets?
07:31:05[Saint_](the fade effect where entries further away from the current selection are dimmed slightly)
07:32:23 Join kevku [0] (
07:35:12***Saving seen data "./dancer.seen"
08:11:21 Quit [Saint_] (Remote host closed the connection)
08:28:35 Quit dionoea (Ping timeout: 244 seconds)
08:28:42 Join dionoea [0] (
08:49:27 Join lebellium [0] (
08:55:51 Quit lurch ()
08:57:14 Quit marcellods (Read error: Connection reset by peer)
08:57:28 Join marcellods [0] (
09:05:10 Join pretty_function [0] (~sigBART@
09:05:18 Join TheLemonMan [0] (
09:05:18 Quit TheLemonMan (Changing host)
09:05:18 Join TheLemonMan [0] (~LemonBoy@unaffiliated/thelemonman)
09:07:31 Quit wtachi (Ping timeout: 245 seconds)
09:31:27 Join n1s [0] (
09:31:27 Quit n1s (Changing host)
09:31:27 Join n1s [0] (~n1s@rockbox/developer/n1s)
09:35:01 Join Rower85 [0] (
09:35:15***Saving seen data "./dancer.seen"
09:36:42 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
09:37:44kugelp[Saint]: so it worked for you?
09:38:07kugelpwhat target did you test on?
10:12:11 Quit bzed (Ping timeout: 248 seconds)
10:13:43 Join bzed [0] (
10:16:26 Quit pretty_function (Ping timeout: 255 seconds)
10:37:21 Join prof_wolfff [0] (
10:56:11 Join brendyn [0] (brendyn@2a01:7e00::f03c:91ff:fedf:65b4)
11:03:51 Join wtachi [0] (
11:06:15 Join bertrik [0] (
11:06:15 Quit bertrik (Changing host)
11:06:15 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
11:06:30 Join Horscht [0] (
11:06:30 Quit Horscht (Changing host)
11:06:30 Join Horscht [0] (~Horscht@xbmc/user/horscht)
11:07:09 Join Totalled_ [0] (~Totalled@
11:10:26 Quit Totalled (Ping timeout: 245 seconds)
11:10:26 Nick Totalled_ is now known as Totalled (~Totalled@
11:12:20 Join ender` [0] (
11:24:55 Join Horschti [0] (
11:24:56 Quit Horschti (Changing host)
11:24:56 Join Horschti [0] (~Horscht@xbmc/user/horscht)
11:27:31 Quit Horscht (Ping timeout: 245 seconds)
11:32:39 Join einhirn [0] (
11:35:18***Saving seen data "./dancer.seen"
11:37:31 Join amithkk [0] (u4289@2buntu/writers/amithkk)
11:47:42 Join [Saint_] [0] (~Saint]@unaffiliated/saint/x-8516940)
11:55:18 Join pamaury [0] (
11:55:18 Quit pamaury (Changing host)
11:55:18 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
11:58:51 Join stoffel [0] (
12:00:00 Quit pamaury (Ping timeout: 245 seconds)
12:01:01 Join pamaury [0] (
12:01:01 Quit pamaury (Changing host)
12:01:01 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:06:57 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
12:09:40[Saint_]kugel: is there something keeping you from committing the directory selection UI stuff? Iirc did it just need testing?
12:10:15[Saint_]I've been playing with the three remaining patches on gerrit, very impressed.
12:10:20 Quit TheLemonMan (Quit: WeeChat 0.3.8)
12:11:19 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
12:12:48[Saint_]I'd personally rather a tap open a directory, and a long press select it, but...that's the only (even then its only slight) issue I have with it, even then I don't really care too much as its a "set and forget" thing mostly.
12:15:32 Join TheLemonMan [0] (
12:15:33 Quit TheLemonMan (Changing host)
12:15:33 Join TheLemonMan [0] (~LemonBoy@unaffiliated/thelemonman)
12:17:28[Saint_]kugel: JdGordon: (?): whoops...
12:17:40[Saint_]Just noticed a bit of a blocker, heh.
12:18:09 Join einhirn [0] (
12:18:27[Saint_]Selecting some directories lock up the ui, with only one way out (hard reset).
12:19:29[Saint_]Examples being: /cache, /data, /root
12:20:29[Saint_]I didn't notice this as I had no reason to select these, knowing no audio lay therein. I tried just now and it locks the whole thing up.
12:25:05[Saint_]Accessible directories: /dev, /etc, /lib, /proc, /mnt, /sdcard, /sys, /system,
12:25:25[Saint_]Everything else locks up the app.
12:25:46kugelgood catch
12:25:54[Saint_]...I guess that kinda screws up devices where "sdcard" lives in /data
12:26:11kugelyou didnt test on non-app yet did you?
12:26:27kugelsdcard in data?
12:26:31[Saint_]kugel: no, sorry. Just app targets.
12:26:47[Saint_]kugel: yeah, some devices do this.
12:27:27kugelnever seen one
12:27:52[Saint_]Ones where "sdcard" is actually still internal storage and not a *real* sdcard.
12:28:07[Saint_]The real sdcard gets yet another hilarious path :)
12:30:19[Saint_]Sorry I didn't catch this earlier, btw. But as I said earlier I never selected these dirs as I knew there was no audio in them. Then I selected one accidentally and the ui locked up...initially I didn't know what I did but some poking proved it was reproducible.
12:30:38[Saint_]At least I caught it before commit :)
12:31:12[Saint_]Apart from this slight hiccup, very nice, very very nice.
12:33:02 Quit user485763 (Quit: Leaving.)
12:34:07 Join user485763 [0] (~aoeu@
12:34:13JdGordonkugel: I see you're using it for the autoresume setting, does that handle more than one folder?
12:34:30JdGordondo you want to add a param to limit it to one folder?
12:35:27 Quit user485763 (Read error: Connection reset by peer)
12:35:34 Join user485763 [0] (~aoeu@
12:36:09JdGordon[Saint_]: that doesnt sound like a horrible problem
12:36:37[Saint_]Locking up the ui isn't a horrible problem?
12:37:15[Saint_]Also, why would you want to limit autoresume to one dir?
12:38:39 Join pretty_function [0] (~sigBART@
12:39:26[Saint_](It can select multiple dirs, btw)
12:44:00*bertrik is struggling with codec makefiles
13:00:39n1sbertrik: what's the problem? they should be very simple
13:00:52lebelliumIs it normal that with the official cabbie theme for H10 there is no space between the bottom of the album art and "next: track title", not even one pixel while there are several pixels rows between this text and the playing bar
13:01:57kugelJdGordon: it does
13:02:15kugelthe default is even multiple folders
13:03:10[Saint_]lebellium: "normal"?
13:03:40[Saint_] didn't happen by accident, but I guess you're actually asking if its intentional.
13:04:00[Saint_]In which case I'll go with "no".
13:04:29lebelliumis that intentional?
13:04:38[Saint_]Yeah, cool. I'll fix it up, thanks.
13:04:47[Saint_]See above.
13:05:54lebelliumactually it's from the UI simulator, I'll get my H10 next week but I assume it looks the same in the simulator and on the device?
13:06:10n1sbertrik: should be pretty straight forward copying the makefile from one of the other codec libs (except mad and speex since thy're special) adding a SOURCES file for the lib, adding a line to codecs/SOURCES for the codec c file with the entrypoint and adding a few things in codecs.make
13:06:45[Saint_]Thanks for pointing it out. I nerd to do a bit of work all round on the non-app cabbies.
13:06:46bertrikn1s, yes I'm doing that now
13:07:24lebelliumthank you for the coming fix :)
13:07:39bertrikI'm having a couple of headerfiles in lib/rbcodec/codecs/libopus and I can't seem to #include them from lib/rbcodec/codecs/libopus/celt
13:09:42bertrikin the lib/rbcodec/codecs/libopus/SOURCES file I have celt/bands.c for example
13:10:41n1sdoes adding OTHER_INC += -I$(RBCODECLIB_DIR)/codecs/libopus/celt to your .make file help?
13:11:02[Saint_]...opus? wow...where have I been?
13:11:06lebelliumand is there any H3xx developer around or someone who could explain to me why the RB utility can't patch any firmware but only some of them and if someone could make the latest fw 1.31 by the RB utility patchable ?
13:11:43n1slebellium: because it only patches versions that are known to work
13:11:57 Quit TheLemonMan (Quit: WeeChat 0.3.8)
13:13:01n1sthe stand alone patching tool should have a switch for ignoring this check or just need a trivial patch to do so but at your own risk of course
13:13:10bertrikn1s, I can try, but I'm not trying to include stuff from the celt directory
13:14:20lebelliumnls: I tried the fwpatcher.exe tool too but it also says "unrecognised firmware"
13:14:34n1sbertrik: ah the other way around, drop the /celt at the end i guess
13:15:22n1slebellium: yes, it has the same check but i thought it should have aswitch to ignore that, that said though this is probably risky and quite pointless :)
13:15:54n1sor do you actually use the OF?!
13:16:07lebelliumactually I just wanted to try the new games introduced in the OF v1.31 without loosing the RB dualboot.
13:16:33lebelliumjust curiosity
13:16:38 Join agumonkey [0] (
13:17:26bertrikn1s, the line from 13:10 does not help, I'll try drop the /celt
13:18:30n1sbertrik: i think the later one should help, i misunderstood what you wanted at first
13:18:49bertrikthe later one does not help either
13:19:05bertrik/home/bertrik/rockbox.git/rockbox/lib/rbcodec/codecs/libopus/celt/arch.h:42:24: error: opus_types.h: No such file or directory
13:19:35bertrikit works if I add it to CFLAGS :|
13:19:58bertrikgod, I hate Makefiles, especially reverse engineering them
13:20:32bertrikOTHER_INC seems to be involved in the "codec helper library"
13:20:42n1sit seems broken too :)
13:21:11n1sseems the var is just set but never usede
13:21:37 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
13:23:16n1sbertrik: you can cheat and use paths in the includes but that's not very nice
13:24:10bertrikor pile all source files up in one directory, but I'd rather keep the original source file layout a bit
13:24:11 Join TheLemonMan [0] (
13:24:11 Quit TheLemonMan (Changing host)
13:24:11 Join TheLemonMan [0] (~LemonBoy@unaffiliated/thelemonman)
13:24:33bertrikopus is basically built on two other codecs: silk (used in skype) and celt
13:27:18n1sdoes adding $(OTHER_INC) to the INCLUDES line at the top of root.make help?
13:32:18bertrikWhere can I best add stuff like -DFIXED_POINT? or perhaps add a #define in a header file somewhere in the codec-specific directory?
13:32:27 Quit Szatan (Changing host)
13:32:28 Join Szatan [0] (principal@unaffiliated/blown-engine)
13:35:22***Saving seen data "./dancer.seen"
13:37:35 Quit pretty_function (Remote host closed the connection)
13:41:09n1sto do that in the makefile i think you need to do something like libmad does and use your own make rule and just set whatever flags you want
13:41:28bertrikok, got the celt/ part to compile, but now I have a bogus dependency in make.dep from silk/CNG.o to a header file from celt/
13:43:11bertrikI added $(OPUSLIB) : CODECFLAGS += -DFIXED_POINT to libopus.make and that seems to work
13:44:57bertrikah, there is a config.h file in SILK
13:46:31 Quit TheLemonMan (Quit: WeeChat 0.3.8)
13:48:38n1sbertrik: yeah, but i'm not sure if that will make -DFIXED_POINT be in CODECFLAGS for all codecs, looks like it will to me
13:49:32n1syou did make clean after trying the different include changes right?
13:50:38bertrikoh, there is a dependency from silk to celt after all
13:51:43 Join pretty_function [0] (~sigBART@
13:57:01 Join TheLemonMan [0] (~LemonBoy@unaffiliated/thelemonman)
14:07:26lebelliumI have a "Ubuntu Font Pack" on my PC. Is it from you [Saint]? If so, can I use it for my H10 theme?
14:12:30 Quit stoffel (Ping timeout: 245 seconds)
14:13:51 Join liar [0] (
14:28:56bertrikhm, now libfaad is interfering because both libfaad and silk has a structs.h include file
14:30:41bertrikthe OTHER_INC thing seems like a bad idea, because it globally adds an include path
14:43:43 Quit user485763 (Read error: No route to host)
14:44:36 Join user485763 [0] (~aoeu@
15:10:29bertrikok, I now have the bulk of opus compiling
15:12:41bertriknext up: 1) codec main that parses opus packets from the ogg container and feeds them through the opus decoder 2) metadata parser
15:27:32bertrikopus has its own specification of how it is wrapped in the ogg container, I wonder if we can re-use existing code or if we need to write somthing new
15:27:53 Quit pretty_function (Remote host closed the connection)
15:33:59[Saint_]lebellium: yep, sounds like it's from me.
15:34:30[Saint_]And yes, the Ubuntu Font License 1.0 is compatible
15:35:26***Saving seen data "./dancer.seen"
15:36:02 Join LambdaCalculus37 [0] (~LambdaCal@rockbox/staff/LambdaCalculus37)
15:36:20 Quit LambdaCalculus37 (Client Quit)
15:36:37bertrikvorbisfile.c implements seeking for vorbis in an ogg stream. This could be vorbis-independent (but isn't) and be re-used for other codecs that use an ogg container, like opus.
15:51:06 Quit mikroflops (Ping timeout: 240 seconds)
15:51:26n1sbertrik: i think that only makes sense if we have a separate ogg library like we do for some container formats
15:54:14 Quit n17ikh (Ping timeout: 252 seconds)
15:56:13lebellium[Saint_] ok thank you. Could you make one or several more font sizes between 10 and 14?
15:57:29[Saint_]Yeah, sure. It wont be tonight, but sure.
16:02:27lebelliumcool :) > I find the 14 Ubuntu [Bold].fnt pretty cool for the track title, artist and album but I'm looking for other fonts for the other information. Maybe an ubuntu font between 10 and 14 will fit it well :)
16:07:53 Join n17ikh [0] (~n17ikh@
16:07:54 Quit n17ikh (Changing host)
16:07:54 Join n17ikh [0] (~n17ikh@unaffiliated/n17ikh)
16:29:43 Join lebellium_ [0] (
16:31:53 Quit lebellium (Ping timeout: 260 seconds)
16:32:08 Nick lebellium_ is now known as lebellium (
16:36:05 Join mikroflops [0] (
16:45:33bertrikn1s, I guess vorbis is the only ogg container user so far, right?
16:45:43bertrik(which is supported in rockbox)
16:47:03bertrikopus compiles now (although with some warnings), stuff to fix: 1) find out what we really need (leave out the *en*coder stuff) 2) fix the lrint and lrintf warnings 3) fix malloc/free stuff used in some places
16:50:11[Saint_]Would the encoder stuff not be useful, for say, recording?
16:51:44bertrikyes, but I think that is of later concern
17:20:02 Quit user485763 (Ping timeout: 252 seconds)
17:23:03 Join stoffel [0] (
17:27:40 Quit Horschti (Quit: Verlassend)
17:30:26 Join gmaxwell [0] (~greg@wikimedia/KatWalsh/x-0001)
17:35:27***Saving seen data "./dancer.seen"
17:35:35 Join pretty_function [0] (~sigBART@
17:37:39 Quit agumonkey (Quit: Lost terminal)
17:44:14 Quit pretty_function (Remote host closed the connection)
17:54:56 Quit factor (Read error: Connection reset by peer)
18:07:55 Join factor [0] (~factor@
18:18:21 Join y4n [0] (~y4n@unaffiliated/y4ndexx)
18:38:40pamauryfunny how a completely documented chip can be so hard to find on google
18:46:01 Join pretty_function [0] (~sigBART@
18:48:08bertrikoh, which one?
18:53:18pamaurybertrik: em1 = Emma Mobile1
18:54:03pamauryit seems to be a chip uses in some packages like the em1-d512 (used in some sony nwz) or the (em1-)mp200/201 (used in some sony nwz too)
18:58:02 Join XavierGr [0] (~xavier@rockbox/staff/XavierGr)
19:07:14 Join quem [0] (
19:07:33quemsigh. the battery or something of my less than seven months old sansa clip+ seems to have failed.
19:08:03quemphone's just too clumsy for my liking when it comes to listening to music and books.
19:08:07*quem sobs
19:33:39 Quit fs-bluebot (Ping timeout: 246 seconds)
19:34:55 Join fs-bluebot [0] (
19:35:31***Saving seen data "./dancer.seen"
19:35:38 Quit bluebrother (Ping timeout: 260 seconds)
19:37:07 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
19:39:41 Quit y4n (Quit: Assumption is the mother of all fuckups)
19:42:48 Quit TheLemonMan (Quit: WeeChat 0.3.8)
19:57:23 Join y4n [0] (~y4n@unaffiliated/y4ndexx)
20:13:54 Quit [Saint_] (Ping timeout: 246 seconds)
20:37:59 Quit liar (Quit: huiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii)
20:44:40 Quit quem (Remote host closed the connection)
21:02:08 Join MethoS- [0] (~clemens@
21:28:17 Join freqmod [0] (~quassel@2001:700:300:2120:5652:ff:fe70:1e36)
21:28:22kugel[Saint]: fixed
21:28:31freqmodHello, is there any documentation on the rockbox build/makefile system?
21:29:55freqmodi am trying to convice it that $(call preprocess, $(RBCODECLIB_DIR)/codecs/libopus/SOURCES) exists in the source tree (codecs/libopus/ogg/ogg.h
21:31:10pixelmahmm, someone else was also trying to port that codec, I believe I read... was it bertrik?
21:31:29freqmodthe build system looks for <build_root>/ogg/ogg.h instead of codecs/libopus
21:31:48freqmodpixelma: currently i am only looking to see if i can make makefiles for it
21:31:58freqmodi didn't find any info on other ports on google
21:32:30gevaertsbertrik is working on that, yes. Check today's irc logs
21:32:35freqmodhmm, ok
21:32:46gevaertsI'm sure he won't mind help :)
21:32:49freqmodthen i'll just wait and hope it gets finished :P
21:33:09freqmodit dind't look that diffecoult, I could not find that many mallocs
21:33:20freqmodand it has a fixed point configure option
21:34:41freqmodI am most likely off the internet next week, and then starting in a new job
21:35:02freqmodbut if some code is made available i might help a bit if needed
21:35:34***Saving seen data "./dancer.seen"
21:39:42pixelmaheh, I meant it in a "could be fun joining efforts" way, not in "someone else is already on it, you don't have to" one :)
21:42:19freqmodi'll fight my way trough the perl scripts
21:42:44freqmodand then we'll see if i get anything compiling before i can see what bertrik has done
21:44:04bertrikwell, the sources compile (with some warnings) and it sounds entirely doable, I think I can look at speex/vorbis to see how to unpack the ogg stream and feed the data to the opus codec
21:44:50bertrikthe biggest thing I worry about is the mismatch in standard frequency of 48 kHz in opus and 44.1 kHz in rockbox
21:45:17freqmodi wrote the speex ogg interface, and that code is not very stable/pretty
21:45:42freqmodthe tremor code is very integrated with the rest of the codec
21:46:00freqmodso ideally the ogg code should be rewritten from scratch
21:46:37freqmoddoesn't rockbox resample other sample rates?
21:46:50*freqmod thought he used lower sample rates in the speex codec
21:47:43 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
21:48:57freqmodehh, the problem with the speex ogg code is that it uses malloc
21:55:15freqmodand the seeking code is quite creative :S
21:59:10 Quit n17ikh (Read error: Operation timed out)
21:59:24gmaxwellit's not terribly hard to make an ogg reader not use malloc... so long as you're willing to impose a maximum packet size.
22:00:19freqmodprobably not,
22:00:52freqmodand that is most likely the reason it works with a simple malloc that does not retrieve unused allocations
22:01:08 Join n17ikh [0] (~n17ikh@
22:01:09 Quit n17ikh (Changing host)
22:01:09 Join n17ikh [0] (~n17ikh@unaffiliated/n17ikh)
22:10:02 Quit kevku (Ping timeout: 248 seconds)
22:13:58 Quit pretty_function (Remote host closed the connection)
22:16:53 Join [Saint_] [0] (~Saint]@unaffiliated/saint/x-8516940)
22:20:03bertrikfreqmod, would it help if I give you my stuff so far?
22:20:44freqmodif you do a git diff and put it somewhere i can look into it
22:22:27freqmod(given that all the files are added to the repo)
22:29:33freqmodthanks, the patch applied cleanly
22:30:33bertrikthe metadata parser does not work, neither does the codec main file, and malloc/free is currently stubbed in os_support.h
22:30:51freqmodis it correct that the build system complains about headers not found?
22:31:13bertrikno, it shouldn't do that
22:31:27freqmodi deleted the directory and started to build again
22:31:45freqmodwhich worked
22:32:15freqmodi'll start trying to fix the parser based on the speex code
22:33:16gmaxwellbertrik: what you did with malloc and free is fine.
22:36:36gmaxwell+// HACK bertrik
22:36:36gmaxwell+// int *value = va_arg(ap, opus_int32*);
22:36:42gmaxwellerp. You're using some old opus code there.
22:37:02gmaxwellThe stupid type there was fixed a long time ago.
22:38:00*freqmod hoped the code could be ported without much changes
22:38:09freqmodas the codec doesn't look like it is completly finished
22:38:13bertrikI just took the latest tar ball, not the latest git stuff
22:39:31gmaxwellThere shouldn't be _any_ changes required to libopus itself, at least if you can either leave alone or define out the malloc/free calls. (it won't use them if you don't call the _create() API).
22:40:01gmaxwellbertrik: the tarfiles are old and just stuck waiting on the IETF to assign the RFC number.
22:41:11gmaxwellWould it help you if I add #ifdef OPUS_NOMALLOC to easily turn off the _create() / _destroy() calls that call malloc so there would be no malloc symbol reference in the source anywhere?
22:42:04bertrikyes, I think that would be nice. What do you think freqmod ?
22:42:32bertrikfreqmod, I won't hold a grudge if you just start over .. :)
22:43:07freqmodI see no reason for not keeping your code as a base
22:43:46freqmodand if it has done significant changes to the upstream opus code they just have to be reverted/fixed
22:44:24freqmodfor metadata and codec interface it doesn't matter much how the opus library code is internally as long as the api has not changed
22:44:51freqmodat least you seem to have figured out the build system issued i had
22:44:52gmaxwellAh, you can just set -DOVERRIDE_OPUS_ALLOC -Dopus_alloc -Dopus_free and basically get that, except it doesn't remove the then-non-functioning create and free calls.
22:45:18bertrikso far, it seems we actually need very little interaction with the main opus code, just 3 or 4 calls to decode
22:45:23freqmoddoes the gerrit stuff in rockbox support separate "branches"?
22:45:37*freqmod are committing bertiks patch as a separate commit
22:45:46gmaxwellbertrik: Thats all you'll need, plus a call to set the playback gain from the header.
22:46:38gmaxwell(which is another reason why you shouldn't be using the outdated code, otherwise handling header gain field is somewhat more work)
22:47:14freqmodhmm, ok
22:47:42bertrikgmaxwell, at first I think we'll be happy with playback-only for opus, but I have no idea which source files are needed for decode-only
22:48:37freqmodbertrik: tell me if you want to update to the most recent opus code and i will keep out of those files
22:48:50freqmodotherwise i will probably do it after i get the metadata stuff working
22:48:52bertrikopus/src and opus/include have been copied to libopus/ . I copied opus/celt to libopus/celt and opus/silk to libopus/silk
22:50:03gmaxwellbertrik: Unfortunately for someone who only cares for decode, encode+decode share a lot of code in opus. opus/silk/float is droppable and a few other files here and there if you're not decoding.
22:50:40bertrikok, I already left out silk/float and some of the obvious encoding-related files
22:52:08freqmodencoding could be useful on rockbox too
22:52:22freqmodbut I would probably not port it myself
22:53:04[Saint_]Another possibility for recording wouldn't hurt.
22:56:11gmaxwellfreqmod: shouldn't be ~any harder than decoding, at least from the codec perspective.
22:56:30freqmodbut I have'nt implemented an encoding codec before :P
22:56:36gmaxwellFwiw, CFLAGS="-DOVERRIDE_OPUS_ALLOC -DOVERRIDE_OPUS_FREE -D'opus_alloc(x)=NULL' -D'opus_free(x)=NULL' " make -f Makefile.draft works fine to get you a build that has no malloc/free symbols in it.
22:57:10gmaxwell(then you just have to not call _create or _destroy in opus)
23:01:07 Quit stoffel (Ping timeout: 248 seconds)
23:06:44 Quit anewuser ()
23:19:19freqmodhope i don't make anybody unhappy :/
23:19:30freqmodif someone wants access tell me
23:23:21 Quit benedikt93 (Quit: Bye ;))
23:23:58kugelfreqmod: fwiw, there's a rockbox git mirror on github you could have forked from
23:24:30freqmodmaybe it is possible to rebase it
23:24:38freqmodmy checkout on top of that
23:24:51kugel(which is still not updated automatically, btw)
23:25:33freqmodi had it checked out automaticly, and plan to remove it once the codec is merged
23:25:40freqmodso i am not sure if i bother then
23:32:05 Quit linuxstb (Ping timeout: 245 seconds)
23:35:35***Saving seen data "./dancer.seen"
23:39:27freqmodit looks like the celt codec uses some scratch memory
23:39:43freqmodwith a stack based allocator on top of malloc
23:39:57freqmod(celt as part of opus)
23:40:43gmaxwellfreqmod: No. It optionally does this, but its not generally encouraged.
23:41:03gmaxwellIt's normally used for TMS320-c55 DSPs... and where it never frees.
23:41:11freqmodthen i'll have to find out how to disable it :P
23:41:31gmaxwelland if you were not running outdated code you you would get helpful messages.
23:41:44freqmodi retrieved the most recent code
23:42:05gmaxwellthen it shouldn't be enabled unless you've added a define for THREADNONSAFE_PSEUDOSTACK
23:42:22bertrikfreqmod, there is apparently some arm optimised code in opus already, it's not enabled yet in the patch I gave you
23:43:20freqmodfgrep -R NONSAFE * in the most recent opus git checkout gives no results
23:43:25freqmodbertrik: ok
23:43:35freqmodi used your makefiles and replaced all the code from upstream opus
23:43:43freqmod(in the libopus directory)
23:43:53gmaxwellfreqmod: sorry, lossy memory— NONTHREADSAFE_PSEUDOSTACK
23:43:59 Join linuxstb [0] (
23:44:15bertrikfreqmod, and it still compiles? :)
23:44:35freqmodwell, most of it compiles
23:44:45freqmodbut linking does not work until i fix the mallocs
23:45:17freqmodi'll push it to github when it compiles
23:45:25freqmodand then go to sleep
23:46:20gmaxwellfreqmod: so point me to the remaining malloc thats leaking through. :)
23:46:39freqmodit is only the celt stack stuff
23:46:46freqmodi try to use VAR_ARRAYS instead
23:47:13gmaxwellfreqmod: How did you end up with NONTHREADSAFE_PSEUDOSTACK defined? Yes. For your toolchain you should use VAR_ARRAYS.
23:47:29freqmodi don't think it is called
23:47:42freqmodbut i did not define the functions, therefore the compiler complains
23:47:59freqmodand i want to make sure where malloc is called before i adds an empty function
23:48:06freqmod or a funcion pointing to 0
23:48:33 Quit y4n (Quit: PANTS OFF!)
23:52:42freqmodgmaxwell: do you know how to replace MULT16_32_P16?
23:52:48freqmodit seems like it is undefined
23:53:32freqmodhmm, it seems like the fixed point define is missing
23:53:42gmaxwellI was about to say— :)
23:54:38freqmodor is it MULT16_32_Q16
23:54:41freqmodi.e. a typo
23:54:53gmaxwellNo. P and Q are _not_ the same.
23:55:21gmaxwell(and both are defined for both fixed and float)
23:56:03gmaxwell(see fixed_generic.h)
23:56:48freqmodhmm, it is defined in freqmod/rockbox-opus">
23:58:01gmaxwellI'm not sure what you've done to the build enviroment, but you can't just build opus by ripping it all out and not replacing some of it. You'll at least want -DVAR_ARRAYS -DOPUS_BUILD -DFIXED_POINT
23:58:50freqmodmaybe i have copyed the wrong directories :S

Previous day | Next day