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 2012-01-08

00:00:23 Join T44 [0] (
00:00:27kugelas I said
00:00:28jhMikeSI wouldn't worry about the table, it's never going to change
00:00:48n1skugel: you didn't exactly explain anything
00:00:52 Quit Topy44 (Read error: Connection reset by peer)
00:01:12jhMikeSabout the only advantage would be one in IRAM
00:01:22n1swhy can't you stick a table in a .c file and have it include din the build?
00:03:04JdGordonjhMikeS: I have a working proof of concept fr overlays, its stalled becasue every single lcd driver needs updating and testing
00:03:47JdGordonfunman: pong
00:03:59funmanJdGordon: can you look at this debug menu?
00:05:04kugeln1s: I have some framework locally which selects arch specific .S and .c files and chose that over generic .c code
00:05:14jhMikeSJdGordon: I'd like to be able to sub any lcd function with the hw-based one, and overlays would be nice to include if hw ones are supported
00:05:25kugelthe SOURCES entry is the same whether or not there's specific code
00:05:59kugeli.e. there's "coldfire/ffs.S" and "./ffs.c". make will pickup the .S automatically
00:06:01jhMikeSbitmaps should include their metadata so one bitmap function can be called
00:06:36kugelin this SOURCES there can't be other .c files e.g. for such a table
00:06:53kugelthus the .c file for the table would need to be in a different SOURCES file
00:07:07n1skugel: so why can't there be an ffs_table.c file incuded everywhere and ffs.c ffs-cf.S and ffs-arm.S that implement the function (and −−function-sections will get rid of the table on arm where it isn't used?
00:07:16kugeljhMikeS: like lcd_bmp() ? :)
00:07:38n1swhy is that a problem?
00:08:03n1sor shouldn't that limitation be fixed perhaps?
00:08:06kugelwell it's not a *problem*, just awkward
00:08:11jhMikeSkugel: yeah, lcd_bmp, lcd_bmp_part. has it been changed since I last looked (years ago) ? :)
00:08:29kugelwell, there could be other .c files indeed. but then the table would be duplicated for arm
00:08:42kugeljhMikeS: they're new :P
00:08:51JdGordonfunman: I said sunday night my time didnt I? :)
00:09:06*jhMikeS literally hasn't written graphics for what seems like eons
00:09:20funmanJdGordon: you did, i just didn't remember ;)
00:09:23JdGordonjhMikeS: well, overlays arent really what I've implemented... more underlays
00:09:44jhMikeSsomething has to go over them!
00:09:45JdGordonI changed the lcd_* code to draw to a different framebuffer which is then used for the backdrop of the real buffer
00:09:53kugeln1s: let me ask the other way around. why is it important to have it in an extra file?
00:10:44n1skugel: this table isn't important but it seems to me like you are fixing the wrong thing
00:11:28jhMikeSkugel: mono/color bitmaps use the same api now?
00:14:12kugelyea still passes stride and fb_data directly rather than a bitmap struct or some such
00:15:36kugelnot lcd_bmp{_part}
00:16:10jhMikeSI'm looking at lcd-16bit.c right now
00:16:45kugellook at lcd-16bit-common too
00:17:28jhMikeSit's all different now
00:29:42 Quit liar (Remote host closed the connection)
00:35:48 Join perrikwp [0] (
00:38:12 Quit perrikwp_ (Ping timeout: 244 seconds)
00:39:17kugelenabling asm for the ypr0 works out of the box
00:39:39lebelliumsorry kugel, what is asm? :p
00:42:47lebelliumwhat opportunities does that give you?
00:42:56 Quit ender` (Quit: You this read wrong.)
00:44:00funmancodecs optimizations
00:44:21kugelno women, though
00:44:29saratogamakes sense, assuming you've got an arm CPU it should just work
00:45:26kugelsaratoga: well, in svn some asm is in firmware/target/arm which also includes stuff only for native targets
00:45:42saratogaah thought you just meant for codecs
00:45:44kugelthat needs care
00:45:50kugelno, globally
00:46:01kugelyou can't turn it on just for codecs iirc
00:46:47kugeljhMikeS: lcd-memframe isnt really faster for lcd_update/rect
00:47:14kugelsaratoga: oh well, it does not work. crashes if I run a codec
00:47:54kugeljhMikeS: 1230 vs 1240fps
00:48:00saratogayou could try wav to rule out the codec's own ASM code
00:50:08jhMikeSkugel: fast than what?
00:50:21kugeljhMikeS: c/memcpy vs asm
00:50:41kugelyuv is substantially faster though
00:50:56jhMikeSit's the same speed roughly if the copies are comparable but if it's small partial updates it's quite alot faster
00:52:04jhMikeSon gigabeat S it's exactly the same speed for full width, for 1/4 screen updates it blew the memcpy away since it really only copies 1/4 of the display rather than full width
00:52:42 Quit dfkt (Quit: -= SysReset 2.55=- Sic gorgiamus allos subjectatos nunc.)
00:52:43kugelwhy does memcpy update the full width?
00:53:47kugelthe memcpy loop I wrote updates also only 1/4
00:53:51jhMikeSelse then you have to call memcpy for each line
00:54:03 Quit perrikwp (Read error: Connection reset by peer)
00:54:39kugelbut that's faster than doing full width updates for sure
00:55:50kugelsaratoga: mp3 crashes; ogg, flac and m4a do work
00:55:52jhMikeSit was implemented as full-width at one time, just paying attention to the y and height, which wasn't my doing.
00:56:20funmanwhy some PP force an ADC scan at each read and some don't ?
00:56:50funmani started replacing all adc_scan by adc_read but noticed adc_read() returned a cached value for PP5020
00:57:38 Join perrikwp [0] (
00:58:13saratogakugel: i can't think of anything special about libmad
00:58:20saratogait does the same stuff as vorbis basically
00:59:30saratogaonly really weird thing about it is the dual core support, but assuming NUM_CORES isn't 2, that shouldn't matter
00:59:47jhMikeSkugel: I suppose the ASM routine will be better for taller rectangles than short ones (squeezing out whatever is possible to get)
01:00:05***Saving seen data "./dancer.seen"
01:00:59kugeljhMikeS: the numbers probably don't count on this device
01:01:14kugelI don't know when the update is actually performed
01:02:18jhMikeShow's that?
01:07:16 Quit Strife89 (Ping timeout: 248 seconds)
01:07:21CIA-88New commit by funman (r31620): button-target.h : move prototypes to button.h ...
01:07:35jhMikeSthat seems odd if it's actually a normal memory framebuffer display with LCD DMA refreshing
01:08:43kugeljhMikeS: it's linux framebuffer device
01:09:01kugelI haven't looked into the linux sources to find out when the actual update is performed
01:09:20 Join perrikwp_ [0] (
01:09:22 Quit Newguy1 (Ping timeout: 260 seconds)
01:09:40CIA-88r31620 build result: 50 errors, 5 warnings (funman committed)
01:09:51 Quit perrikwp (Ping timeout: 252 seconds)
01:10:00jhMikeSok, not native then
01:10:08amiconnfunman: You cannot unify cached and uncached adc read, the method depends on the target
01:10:41amiconnSome targets have slow adcs, or other constraints which forbid uncached reading
01:11:26 Nick perrikwp_ is now known as perrikwp (
01:11:40kugelfunman: you didn't like the XX-driver.h idea did you?
01:11:51 Join MethoS- [0] (~clemens@
01:12:09kugelwhy not?
01:12:30funmanwe don't need a gazillion of drivers, xxx.h should define the target API (as it does already)
01:12:56kugelone XX-driver.h for all targets
01:13:12kugeljust to make it more obvious what's to be implemented by a new target
01:13:55funmanrun make, fix errors; that's obvious already
01:14:49jhMikeSshould there be a new target wizard that generates a skeleton implementation? :D
01:15:18 Join jlbiasini [0] (
01:15:48kugelfunman: I don't really like prototypes in headers that are not implemented in the corresponding .c file
01:16:01funmanthat's what we do already
01:16:14 Join domonoky1 [0] (
01:16:31kugelwe do many things I don't like :P
01:16:47 Quit domonoky (Ping timeout: 240 seconds)
01:17:52jlbiasinihello, It seems I cannot change anything to the partition of my player anymore. (I try copy/delete file, reformat, delete partition both from linux and windows... Is there anything else I could try?
01:18:02amiconnkugel: Why is that? I'd rather have relatively few header files
01:18:08jhMikeSyou should've been here when we had to shovel coal into our players just to keep them running!
01:18:44CIA-88New commit by funman (r31621): fix previous commit ...
01:20:53CIA-88r31621 build result: 4 errors, 0 warnings (funman committed)
01:21:08funmankugel: drivers/button.c use this interface already
01:21:19funmanbut we can't (obviously) move all definitions in this file
01:21:19 Part jlbiasini
01:21:47kugelThat's not what I'm suggesting..
01:22:33kugelI'm suggesting a single button-driver.h which has the prototypes for all button-XXX.c
01:22:40CIA-88New commit by funman (r31622): fix fuze+ build: we need stdbool.h for bool
01:22:44kugelincluded by button.h that is
01:22:46funmanwhy do we need a separate header again?
01:23:28jhMikeSthen it would be like system.h already is
01:23:46jhMikeSand some others
01:24:41kugelto keep the general protypes and the driver ones logically seperate. IMO button.h shouldnt define prototypes that button.c doesn't implement
01:24:51 Quit n1s (Quit: Ex-Chat)
01:24:57CIA-88r31622 build result: All green
01:29:40kugelalso, the driver's button functions are only called from within button.c, so they don't need to be exposed to the entire tree
01:30:35jhMikeSwhat about bootloaders?
01:35:08funmankugel: anyway it can be done in parallel of what i do
01:35:14funmani just make it easier for you
01:35:57funmanlooks like i caused ondio red but it was hidden by a random variation :P
01:37:28kugelsaratoga: unfortunately I get no other message than segfault
01:39:32CIA-88New commit by funman (r31623): adc-target.h: cleanup ...
01:41:34CIA-88New commit by funman (r31624): remove cargo-cult empty i2c-target.h
01:41:40CIA-88r31623 build result: All green
01:41:58 Join perrikwp_ [0] (
01:43:48CIA-88r31624 build result: 6 errors, 0 warnings (funman committed)
01:44:27 Quit perrikwp (Ping timeout: 252 seconds)
01:46:55funmanperhaps build clients should run make V=1
01:47:43kugelplaese no
01:47:46funmanmrobe500 has lcd_enable() but not HAVE_LCD_ENABLE
01:47:55funmankugel: why?
01:48:09funmanshouldn't we know exactly what goes wrong when there's a problem?
01:48:10kugelmakes spotting errors even harder
01:48:17funmanhm true
01:48:41funmanif we can display red/yellow we could have markers at the top (error #1, #2 ..)
01:48:44kugelV=1 doesn't show anything useful even in case of errors most of the time
01:49:18kugelespecially not when the build client is running with -jX
01:50:39kugelfunman: there's the "Goto problem 1 2 3 N" at the top
01:52:14 Quit Brijesh222 (Ping timeout: 240 seconds)
01:52:57CIA-88New commit by funman (r31625): lcd-target.h: remove some duplicate declarations
01:53:14funmankugel: Oo i had never seen that
01:54:28jhMikeSit's neat :)
01:54:49CIA-88r31625 build result: All green
01:56:28kugelhow do I go find out what's crashing mp3 on the ypr0?
01:56:39kugelhosted target -> no UIE
01:57:38kugelnever used those
01:58:32funmanor plain gdb
01:59:54kugelcan't access this with gdb
01:59:54 Part domonoky1
02:00:02kugelat least we don't know how yet
02:00:59kugelI don't suppose one can do anything useful from a singal handler?
02:02:13kugelI could compile gdb to run on the thing, but can't interactively control it
02:02:28kugelor run it through some *trace
02:02:49 Quit lebellium (Quit: ChatZilla 0.9.88 [Firefox 10.0/20111228055358])
02:07:22funmanyou don't have a shell?
02:07:38kugelno sh.rock :P
02:08:21funmanperhaps you can tell gdb to make a coredump when rockbox crashes
02:09:12kugelgdb −−batch with a .gdbinit should work
02:10:19 Quit pamaury (Remote host closed the connection)
02:13:05funmankugel: the ypro0 also copy/pastes existing as3514 bits :'(
02:13:40kugelfunman: where?
02:14:07funmanadc_read() at least
02:15:06funmanah i'll do it
02:17:42 Quit Zarggg (Quit: Rebooting client...)
02:23:13 Join Zarggg [0] (
02:24:45 Nick perrikwp_ is now known as perrikwp (
02:35:37kugellets see if that works
02:43:19CIA-88New commit by funman (r31626): ascodec-target.h: remove ...
02:44:10CIA-88New commit by funman (r31627): remove now unused prototype
02:44:23CIA-88r31626 build result: 1011 errors, 68 warnings (funman committed)
02:44:30*funman runs
02:45:07CIA-88r31627 build result: 1015 errors, 71 warnings (funman committed)
02:46:39funmanhum what did i commit that i didn't test ..
02:49:39 Quit Zarggg (Quit: Rebooting client...)
02:51:52CIA-88New commit by funman (r31628): fix cpp condition
02:54:11CIA-88r31628 build result: All green
02:55:37funmankugel: the only header on my list is backlight-target.h
02:56:08kugelfunman: can adc-as3514.c be outside of the target tree?
02:56:31funmanvery likely (and i expected you to ask :))
02:57:15funmanthere's no target specific code here, only as3514 specific
02:57:17 Join Zarggg [0] (
02:59:23 Quit MethoS- (Remote host closed the connection)
03:00:09***Saving seen data "./dancer.seen"
03:09:48 Join Keripo [0] (
03:14:18 Quit antil33t| (Ping timeout: 252 seconds)
03:17:14 Quit krabador (Quit: Sto andando via)
03:17:26 Join antil33t [0] (~Ahurhurr@
03:28:49kugelsaratoga: crashes in synth_full_arm.S
03:29:01kugelline 174 gdb says
03:29:06kugelI cannot give more info unfortunately
03:29:22 Join saratoga_ [0] (9803ec71@gateway/web/freenode/ip.
03:29:37saratoga_kugel: makes sense, thats where almost all the asm in the codec is
03:30:56saratoga_maybe the ICODE_SECTION_MPA_ARM stuff isn't defined right on the r0?
03:31:01kugelcompiler error is also unlikely isnt it?
03:32:15saratoga_i think we were able to compile rockbox with codesorcery before, so probably not unless theres something weird about the one you're using
03:33:28saratoga_if you comment out the call to synth_full in synth.c does it still freeze? (audio will obviously not sound right)
03:33:44CIA-88New commit by funman (r31629): adc-as3514.c : move to drivers
03:33:57funmankugel: done
03:34:48kugelfunman: nice
03:35:24saratoga_is it just me or does III_aliasreduce in the asm file never restore the link register?
03:35:44CIA-88r31629 build result: All green
03:35:46saratoga_sorry, III_overlap
03:36:11kugelsaratoga_: btw, gdb cannot show the call stack properly because of the use of sp for calculations
03:36:29saratoga_yes, debugging this ASM was a lot of fun :)
03:37:46kugelsaratoga_: section icode
03:38:55kugelbut on android also so that's not the problem
03:38:55saratoga_is that not defined on your target?
03:39:11kugelshouldnt be, anyway
03:39:19saratoga_we use it on flac too, so if flac works its probably not the issue
03:39:42saratoga_funman: you wrote the ldmpc stuff right?
03:40:38kugelffs, it does "#define ICODE_SECTION_MPA_ARM .icode"
03:40:47kugelshould do "#define ICODE_SECTION_MPA_ARM ICODE_ATTR" instead
03:41:23kugelah hm I guess it being empty doesnt help
03:41:26saratoga_whats the difference?
03:41:39kugelICODE_ATTR is empty on hosted
03:42:43saratoga_and .icode isn't?
03:43:13kugel".icode" is a string literal
03:43:18funmansaratoga_: yeah
03:43:39saratoga_kugel: but the linker doesn't ignore it?
03:43:57saratoga_funman: what happens if a function ends in "ldmpc regs=r4-r7" without specifying "lr"
03:44:24kugellr is implicit IIRC
03:44:45funmanyeah i think it's implicit
03:44:47kugelldmpc means ldm the regs then pc to lr
03:46:04funmanon armv5+ or armv4 without thumb it pops the register into pc directly (doesn't need lr)
03:47:20funmanhmm, ldmpc's "order" argument is unused.
03:47:43saratoga_oh of course
03:48:09funmansaratoga_: i had to lookup the definition in config.h
03:49:02funmanwell, 'order' has been unused ever since this macro was added
03:52:51kugelsaratoga_: no sound and no crash with the call commented out
03:59:13kugelwould be nice if we can find out why it crashes there
03:59:58kugelas I don't think segfault on smlal is possible, perhaps the ldr just above is the culprit?
04:00:18kugelor maybe the line gdb tells is entirely wrong too
04:06:34 Join bharvey [0] (
04:11:31 Quit bharvey (Quit: CGI:IRC (Ping timeout))
04:13:52 Quit amiconn (Disconnected by services)
04:13:53 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:14:14 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:14:57 Quit efyx (Remote host closed the connection)
04:16:00 Quit pixelma (Disconnected by services)
04:16:02 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:16:04 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:16:10saratogafunman: did you end up making all the updated mkamsboot files ?
04:18:19saratogaah i see that you did: funman/mkamsboot-1.5/">
04:18:29saratogado any of these still need to be tested?
04:21:38 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
04:24:32 Join dys`` [0] (
04:26:11 Quit dys` (Ping timeout: 252 seconds)
04:26:41 Quit [7] (Disconnected by services)
04:26:56 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
04:29:02saratogawhy is the clip zip manual so screwed up
04:29:40saratogathe html version is truncated, and the pdf version has a big list of install directions for a couple different ams targets all added together
04:37:55 Quit amithkk (Remote host closed the connection)
05:00:10***Saving seen data "./dancer.seen"
05:03:25funmansaratoga: all are tested but ppc binary is missing
05:03:48saratogaIMO we should get bagder or zagor to upload them as soon as possible
05:03:57saratogaright now the wiki links to the old binaries which don't work at all
05:03:58funmanthey are all in Bagder's inbox too
05:04:03saratogaok cool
05:04:20saratogaPPC still doesn't compile?
05:04:33funmanwith SVN perhaps
05:04:48funmani was betting that nobody would notice if PPC/OSX was missing
05:05:02saratogayeah seems safe enough to depreciate at this point
05:05:07funmanbut the first user who tried was running exactly OSX on PPC :/
05:05:08saratogapeople can still use rbutil
05:05:48funmanno idea for the manual - i was thinking missing screenshots would be the first problem
05:06:17saratogaat present we have several sets of install directions, none of which will work
05:08:51 Join Strife89 [0] (
05:10:52 Quit t0rc (Quit: WeeChat 0.3.6)
05:14:14funmando we have the 'go to ABI forums'? :)
05:14:31saratogai don't think so
05:14:54saratogai fixed the sansa ams wiki page by changing the link to your webspace for now
05:15:19saratogathe manual is still broken though, and i guess no sense fixing it until we have a download link anyway
05:15:31saratogaat least now its obviously broken so most people probably realize its not to be trusted
05:15:54saratogawe really just need a new rbutil release
05:29:30 Quit Strife89 (Quit: Heading home)
05:34:12 Join amithkk [0] (u4289@gateway/web/
05:34:12 Quit amithkk (Changing host)
05:34:12 Join amithkk [0] (u4289@2buntu/writers/amithkk)
05:34:59 Quit ps-auxw (Ping timeout: 252 seconds)
05:39:08 Join ps-auxw [0] (
05:55:13 Join Rob2223 [0] (
05:58:32 Quit Rob2222 (Ping timeout: 240 seconds)
05:59:21 Quit amithkk (Remote host closed the connection)
06:04:32saratogaFS #12514 seems to be another report of get_token_value failing, probably due to memory corruption
06:04:32fs-bluebot Sansa ClipV1 Non-Deterministic Crashes on Vorbis Playback (bugs, closed)
06:12:17 Join perrikwp_ [0] (
06:14:08 Quit perrikwp (Ping timeout: 240 seconds)
06:14:45 Quit perrikwp_ (Read error: Operation timed out)
06:15:28 Join perrikwp [0] (
06:17:34 Quit ps-auxw (Quit: leaving)
06:17:56 Join ps-auxw [0] (~arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
06:19:45 Quit perrikwp (Read error: Connection reset by peer)
06:20:55 Join perrikwp [0] (
06:28:41 Quit perrikwp (Read error: Connection reset by peer)
06:30:09 Join amithkk [0] (u4289@2buntu/writers/amithkk)
06:30:09 Join perrikwp [0] (
06:37:54funmanhm it's always clipv1
06:39:36saratogasmall buffer
06:41:04saratogaso probably its due to compaction or some other kind of double allocation thats much less likely on devices with larger buffers
06:42:12funmankugel: is there any debug we can enable to spot bugs in buffer allocation ?
06:42:51funmankugel: clipv1 still show bugss : FS #12514 / FS #12475
06:42:52fs-bluebot Sansa ClipV1 Non-Deterministic Crashes on Vorbis Playback (bugs, closed)
06:42:52fs-bluebot Crash while playing audio (bugs, unconfirmed)
06:43:18funmani dont remember causing the abort in mmy report
06:43:31funmandont remember the address causing it i mean
06:47:16saratogadebugging it in the sim with an artificially small audio buffer (e.g. 64kb) might work
06:49:07 Quit antil33t (Ping timeout: 255 seconds)
06:55:07 Join antil33t [0] (~Ahurhurr@
06:56:03 Quit factor (Ping timeout: 240 seconds)
07:00:12***Saving seen data "./dancer.seen"
07:04:48 Quit antil33t (Read error: Connection reset by peer)
07:08:10 Join antil33t [0] (~Ahurhurr@
07:09:27 Quit captainkwel (Quit: Ex-Chat)
07:12:14 Join factor [0] (~factor@
08:05:59CIA-88New commit by saratoga (r31630): Clarify that you must reboot into the Sandisk firmware when uninstalling the bootloader.
08:06:44 Quit Keripo (Quit: Leaving.)
08:08:26CIA-88r31630 build result: All green
08:11:23 Join perrikwp_ [0] (
08:11:28pocketprotectorcheck out kaskade, that music dj is good.
08:14:06 Quit perrikwp (Ping timeout: 252 seconds)
08:32:41 Quit perrikwp_ (Read error: Connection reset by peer)
08:34:20 Join perrikwp [0] (
08:54:49 Quit factor (Read error: Connection reset by peer)
09:00:13***Saving seen data "./dancer.seen"
09:05:35 Join factor [0] (~factor@
09:28:59 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
09:31:51 Join RusselKubes [0] (
09:55:26CIA-88New commit by bertrik (r31631): sudoku: make local functions static and add missing #include
09:57:17CIA-88r31631 build result: All green
09:59:36 Join martinst_ [0] (
10:07:34 Join ender` [0] (
10:15:14 Join Horscht [0] (~Horscht@xbmc/user/horscht)
10:17:41 Quit Horschti (Ping timeout: 248 seconds)
10:21:04 Join quick- [0] (73f1b42a@gateway/web/freenode/ip.
10:21:24 Join pamaury [0] (
10:21:25 Quit pamaury (Changing host)
10:21:25 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
10:23:38 Quit quick- (Client Quit)
10:37:52 Quit gevaerts (Ping timeout: 268 seconds)
10:40:56 Quit Horscht (Quit: Verlassend)
10:45:36 Join n1s [0] (~n1s@rockbox/developer/n1s)
10:48:58kugelsaratoga_: I tried that already
10:49:44saratogasim with small buffer?
10:54:19 Join Horscht [0] (~Horscht@xbmc/user/horscht)
11:00:16***Saving seen data "./dancer.seen"
11:00:37n1shmm, we now have 6 patches open for updating the german translation
11:03:13 Join JdGord [0] (~AndChat@
11:05:27 Quit perrikwp (Read error: Operation timed out)
11:05:36 Join perrikwp_ [0] (
11:06:00 Quit n1s (Read error: Connection timed out)
11:06:44 Join n1s [0] (
11:06:45 Quit n1s (Changing host)
11:06:45 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:10:54 Join perrikwp [0] (
11:13:45 Quit perrikwp_ (Ping timeout: 260 seconds)
11:21:40 Quit amithkk (Remote host closed the connection)
11:22:12 Quit RusselKubes (Quit: Ex-Chat)
11:25:19bertrikn1s, nice ...
11:29:40 Quit saratoga (Quit: Page closed)
11:31:11 Join lorenzo92 [0] (~chatzilla@
11:31:30 Join Lonely_Dragon [0] (
11:31:32CIA-88New commit by bertrik (r31632): FS #12406 - Updated Galician translation by Ismael Castiñeira Álvarez
11:32:28 Quit Lonely_Dragon|2 (Ping timeout: 240 seconds)
11:33:27CIA-88r31632 build result: All green
11:37:44bertrikn1s, I'd say we have plenty of german-speaking contributors to handle that
11:38:19n1sbertrik: yeah i think so too
11:45:04kugelall are bad IMO
11:45:55kugelI don't understand why almost all want to invent a new translation for sleep timer. there exists one already (another lang string)
11:46:25 Join lebellium [0] (
11:46:48n1sSomeone should just update the translation in svn and close all the patches then
11:48:16 Join prof_wolfff [0] (
11:52:32 Join amithkk [0] (u4289@gateway/web/
11:52:32 Quit amithkk (Changing host)
11:52:32 Join amithkk [0] (u4289@2buntu/writers/amithkk)
12:12:04 Quit anewuser (Ping timeout: 252 seconds)
12:12:40 Quit JdGord (Quit: Bye)
12:21:13 Quit lorenzo92 (Quit: ChatZilla 0.9.88 [Firefox 9.0.1/20111228084940])
12:25:53 Join nick_p [0] (
12:29:08lebelliumbertrik: did you read my replies yesterday? If you send me a compiled version of your patch I can try it ;)
12:36:44 Join mortalis [0] (~mortalis@
12:37:25 Join Kiwi_CAM [0] (~Kiwi_Cam@
12:49:58CIA-88New commit by bluebrother (r31633): Fix wrong includes.
12:50:06CIA-88New commit by bluebrother (r31634): Rockbox Utility: use libmp3lame for voice clips. ...
12:51:45kugelbluebrother^: no static linking?
12:51:52CIA-88r31633 build result: All green
12:53:52CIA-88r31634 build result: All green
12:54:57bluebrother^kugel: no, intentionally not
12:55:48kugelany reason for that?
12:57:01kugel(which one?)
12:57:10bluebrother^(1) the situation for mp3 encoding is a bit unclear in some countries. As a result f.e. Fedora doesn't even ship mp3 decoding so you need to install it yourself
12:57:35bluebrother^(2) I don't want to maintain an additional static library on Windows.
12:57:54kugelI thought you prefer static linking
12:58:16bluebrother^and while mingw-cross-env provides a libmp3lame.a MinGW by itself does not, so building on Windows would get more complicated
12:58:56bluebrother^no, I don't prefer static linking. We just use static linking for the release versions to make distribution simpler −− i.e. make sure the user only needs to run one binary
12:59:06bluebrother^without the need to care about additional DLL files.
12:59:12bluebrother^which mostly is a concern on Windows
12:59:25 Join y4n [0] (y4n@unaffiliated/y4ndexx)
12:59:26bluebrother^and on Linux ... not everyone has Qt installed. Or the required versions.
12:59:41bluebrother^I just need to remember the libpng issues on Linux because it wasn't statically linked.
13:00:18***Saving seen data "./dancer.seen"
13:07:18CIA-88New commit by nick_p (r31635): Simulator: Emulate a period of external power after charging. ...
13:09:10CIA-88r31635 build result: All green
13:25:13 Quit nick_p (Quit: Leaving)
13:25:48 Part martinst_ (" - Chat comfortably. Anywhere.")
13:27:08 Join nick_p [0] (
13:31:32 Join fyrestorm [0] (
13:34:22 Quit fyre^OS (Ping timeout: 240 seconds)
13:36:19 Join TheLemonMan [0] (
13:39:12 Join GermanMushroom [0] (
13:43:54 Quit bertrik (Remote host closed the connection)
13:44:30bluebrother^urgh. Encoding repeatedly fails on an i7 with W7. Seems that change causes the long standing encoding issue to show up more often.
13:45:02bluebrother^though that's not necesarily a bad thing. At least it makes the problem more easily to reproduce :)
13:46:59 Join bertrik [0] (
13:46:59 Quit bertrik (Changing host)
13:46:59 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
13:49:25 Quit Torne (Quit: rebooting server; wish me luck)
13:54:40 Quit orzo (Remote host closed the connection)
13:57:04CIA-88New commit by bertrik (r31636): Remove duplicate #include "button.h"
13:58:44CIA-88r31636 build result: All green
14:00:10 Join Torne [0] (~torne@rockbox/developer/Torne)
14:10:41 Join dfkt [0] (dfkt@unaffiliated/dfkt)
14:21:05 Join tessarakt [0] (
14:21:27tessaraktRockbox does not create my database correctly :-(
14:22:30 Join anewuser [0] (~anewuser@
14:22:30 Quit anewuser (Changing host)
14:22:30 Join anewuser [0] (~anewuser@unaffiliated/anewuser)
14:29:11 Join gevaerts [0] (~fg@rockbox/developer/gevaerts)
14:29:50tessaraktthe files were not copied ...
14:30:26tessarakt0 bytes
14:32:38 Quit gevaerts (Disconnected by services)
14:32:41 Join gevaerts_ [0] (~fg@rockbox/developer/gevaerts)
14:33:11 Quit gevaerts_ (Client Quit)
14:33:29 Join gevaerts [0] (~fg@rockbox/developer/gevaerts)
14:36:36tessaraktthanks, works
14:36:59gevaertsbluebrother^: we already have mp3 encoding code in rockbox itself
14:56:49 Quit pamaury (Remote host closed the connection)
14:59:18CIA-88New commit by amiconn (r31637): Fix several flaws in argument handling and reconfiguring, mainly related to voice building: (1) Properly retain arguments containing spaces by ...
15:00:00amiconngevaerts: That's not exactly high quality though. Iirc it's based on the 'shine' encoder
15:00:19***Saving seen data "./dancer.seen"
15:00:27 Quit n1s (Quit: Ex-Chat)
15:00:54CIA-88r31637 build result: All green
15:03:18gevaertsamiconn: true, but bluebrother^ was mentioning possible patent issues. Quality isn't going to be important there (although it is of course possible that some encoders are old enough to be safe)
15:04:37 Join MethoS- [0] (~clemens@
15:10:18 Quit Torne (Quit: argh rebooting again, stupid hdmi audio)
15:13:33 Join Torne [0] (~torne@rockbox/developer/Torne)
15:20:08 Quit tessarakt (Ping timeout: 240 seconds)
15:33:07 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
15:33:53 Nick Jack87 is now known as Jack87|Away (Jack87@nasadmin/admin/jack87)
15:39:15kugelnick_p: ping
15:41:00nick_pkugel: hi
15:42:08kugelnick_p: tried your embedded cuesheet work
15:42:11kugelwell done!
15:42:46nick_p cool!
15:43:30kugelnow I have one request! merge playlist viewer and browse cuesheet :)
15:44:57nick_pyeah, browse cuesheet is a bit odd, I'll look into it
15:45:51nick_pthough might take a while, work is a bit busy at the moment
15:47:18kugelnick_p: the idea that I have in mind is "inline" the cuesheet tracks into the playlist viewer, and but don't handle them as real tracks
15:47:59kugele.g. you can't move them, they're unaffected by shuffle
15:48:15kugelbut you could still select them individually for playback
15:48:22kugelI don't know if that's feasible
15:48:42nick_phmm, don't know if that's something I'd use myself
15:49:03nick_pie, the motivation isn't really there
15:52:05kugelnick_p: you don't use cuesheets you mean?
15:53:08nick_pkugel: yes, but mainly for long audiobooks as a single file - it's useful for chapter marking
16:00:25kugelah i see
16:21:24 Quit GermanMushroom (Ping timeout: 252 seconds)
16:22:38 Join jlbiasini [0] (
16:32:26 Quit nick_p (Quit: Leaving)
16:32:37 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
16:40:05 Quit chkktri_ (Ping timeout: 248 seconds)
16:45:34jlbiasinibluebrother: did you saw I updated FS #12492 ?
16:45:35fs-bluebot add fuze+ manual (patches, unconfirmed)
16:58:16 Quit TheLemonMan (Quit: WeeChat 0.3.6)
17:00:20***Saving seen data "./dancer.seen"
17:05:40bertrikgevaerts, it appears MTP devices can report their battery charge level to the host
17:05:57gevaertsah, right
17:06:18gevaertsSo over to pamaury :)
17:08:11kugelhm, irssi doesnt compile
17:10:29 Nick Jack87|Away is now known as Jack87 (Jack87@nasadmin/admin/jack87)
17:10:35bertrikand apparently ubuntu 11.10 can report the battery charge level of the MTP connected device on the GUI
17:12:59 Quit Kiwi_CAM (Quit: Leaving)
17:13:47 Join n1s [0] (
17:13:48 Quit n1s (Changing host)
17:13:48 Join n1s [0] (~n1s@rockbox/developer/n1s)
17:22:59 Join TheLemonMan [0] (
17:24:33soapAre there new wiki pages for working with git? I don't see any changes in /wiki/DocsIndex
17:26:08 Quit TheLemonMan (Client Quit)
17:26:31soapyea, I feel like an idiot. I find that easily after bothering people with the question. So here's another dumb question - those are the current "best practices"? I don't recall anything Torne saying in the ML which conflicts with what I'm reading here...
17:26:34bertrikmtp-detect can query the battery level from a clip zip in MTP mode (OF), but I don't see any kind of graphical indicator in the GUI for the battery level
17:27:22Tornesoap: that should be fine; that's what i wrote for the migration
17:27:23n1ssoap: since Torne wrote the wiki page i hope he's put the stuff he conciders best practice there :)
17:27:42Tornei'm currently mirroring every 5 minutes so it should be reasonably current always :)
17:27:46soapI was more concerned Torne changed his mind or had new idea. ;)
17:27:56Torneno, we are all but good to go
17:28:04Tornejust need to corner the swedes and get stuff done :)
17:28:08soapWill the svn repository go dark soon?
17:28:13Tornehopefully :)
17:28:19Tornewell, readonly
17:28:37soapOk, that was the other "real" question. will go dark, not be an impotent mirror?
17:28:48Tornereadonly for now, in case we have to back out
17:28:54soapahh, so if I don't commit anything my SVN workflow need not change?
17:28:56Torneeventualy, i don't know
17:29:01Torneno, i mean, it won't change
17:29:05Torneit will just sit there being the last revision
17:29:10Tornenot a mirror.
17:29:34soapahh, forever at r31666 or whatever?
17:29:40Tornewell, not forever :)
17:29:43soapthank you very much.
17:29:44Torneeventually i expect it will go away
17:29:50n1sprobably only for a while to let people diff their local changes and see that the git stuff works
17:32:25Torneanyone around who was using the old git?
17:32:43Tornei'm curious to talk with such folks about how they plan to move their changes over and if there's anything i can do to help
17:34:33kugelI head git rebase −−onto works
17:35:01Torneit will, if you don't ahve merges, or can deal with them during a rebase without going mad
17:36:20 Join roneet [0] (
17:37:17Torneif you do then probably best to edit them out first :p
17:37:25bertrikTorne, if I have local branches against the current git mirror, can I keep them when moving to the final git archive?
17:37:40Torneyou need to rebase them
17:38:48bertrikok, that shouldn't be too big of a problem
17:40:42Tornegit rebase −−onto newremote/master oldremote/master somebranch
17:40:59 Quit roneet (Client Quit)
17:41:03Torneif you don't have merges it should be entirely trivial and require no manual intervention
17:41:20 Quit mortalis (Ping timeout: 252 seconds)
17:49:47kugelTorne: can I just add the new repo as a remote to the old one?
17:49:58kugelthe history isn't compatible
17:50:20Tornethat doesn't matter
17:50:27Torneyou can add anything you feel like to anything ;)
17:50:44Tornethere isn't only one history
18:14:06 Join Keripo [0] (
18:19:10amiconnkugel: Do I need to do something about the android sdk/ndk setup on my build box?
18:19:51kugelupdate the sdk to api level 15
18:20:12kugelI recommend updating the ndk too while at it, but it's not strictly necessary
18:20:26 Quit Keripo (Quit: Leaving.)
18:21:32kugelI'm actually surprised your build clients still makes builds. I thought i changed the target api level to 15 which the old sdk can't build for
18:22:44kugelapparently the command line tools aren't so strict and only the eclipse plugin
18:32:12 Quit bertrik (Remote host closed the connection)
18:35:27 Join bertrik [0] (
18:35:27 Quit bertrik (Changing host)
18:35:27 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
18:38:42CIA-88New commit by amiconn (r31638): Remove outdated (and buggy) configuration option for the 8MB mod. This is configured by selecting memory size directly.
18:40:43CIA-88r31638 build result: All green
18:49:53kugelamiconn: that 8MB mod that doesnt build anymore? :)
18:50:21amiconnIt does. The 'buggy' comment refers to the selection
18:51:02amiconn'make reconf' would have broken the command line, because it left a single '8', leading to an error when doing 'make reconf' a second time
18:51:43kugelis there another 8mb mod which was removed from the build system?
18:51:59amiconnThe option has been redundant for quite some time (ever since the meory size prompt has been added)
18:52:33amiconnIn theory the 8 MB mod is possible for all archoses, I don't know whether anyone tried it on an Ondio
18:53:20amiconnThe build system did only one, it has been disabled because the binary became too big *for that target* (due to the 8 MB mod including dircache, unlike the stock build)
18:53:45amiconnIf it was the plain recorder, we could re-add such a build e.g. for FM recorder (which has a higher binsize limit)
18:54:47 Quit Horscht (Quit: Verlassend)
18:58:46 Join Keripo [0] (
18:59:53CIA-88New commit by amiconn (r31639): Add missing languages from configure, and sort them in exactly the same order for easier maintenance.
19:00:21***Saving seen data "./dancer.seen"
19:02:01CIA-88r31639 build result: All green
19:08:44 Join dreamlayers [0] (
19:08:44 Quit dreamlayers (Changing host)
19:08:44 Join dreamlayers [0] (~dreamlaye@rockbox/developer/dreamlayers)
19:13:26dreamlayersre 8MB another issue is that the V1 Recorder (and also the Player) might have a flash chip that can't be programmed in-circuit, so Rockbox must be able to load via OF.
19:24:56dreamlayersIf HOME_DIR is "/" when running native, how could that cause problems for RaaA? In rbpaths.h, HOME_DIR is defined as "<HOME>" in that case.
19:30:38amiconndreamlayers: The binsize issue can be (and will be) worked around
19:30:52kugeldreamlayers: what about sims?
19:31:06 Join stripwax [0] (
19:32:26dreamlayersamiconn: the solution is a small /ajbrec.ajz which functions as a bootloader and loads the larger main binary, right?
19:32:50kugeldreamlayers: re: FS #12403: I understand you're afraid of changing lots of linker scripts, but I would prefer we take the opportunity and handle all targets equally. I think wodz has some lds unification going
19:32:51fs-bluebot INIT_ATTR for SH Archos HWCODEC targets (patches, new)
19:33:08amiconnThis transition needs to be cordinated because of the (optional) flash loader though
19:33:21 Join tessarakt [0] (
19:35:26kugelalthough it's not that much linker scripts to be edited. I think just 2 or 3
19:40:34 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
19:42:13 Quit bluebrother^ (Read error: Operation timed out)
19:43:40dreamlayerskugel: // is ok as long as it does not start a path that's given to the OS. The sim's files are supposed to be in ./simdisk, so it shouldn't generate paths starting with //, even if it's running in a root directory.
19:44:08 Quit fs-bluebot (Ping timeout: 240 seconds)
19:46:34 Join fs-bluebot [0] (
19:49:17kugeldreamlayers: the ypr0 target is RaaA and doesn't use "<HOME>"
19:49:30kugelit doesnt remap paths
19:50:01amiconnHigher quality voice files (which don't fit into RAM as a whole) are working fine on Ondio :)
19:52:22dreamlayerskugel: ypr0 defines HOME_DIR to "/mnt/media0". Also, Linux should be able to deal with // properly. I guess it's only a problem on a theoretical target like ypr0 which runs Windows and has HOME_DIR set to "/".
19:52:39kugelyes you're right
19:52:48 Quit domonoky (Quit: Leaving.)
19:52:49dreamlayersAnyways, for Windows, we would also have to handle drive letters.
19:53:01 Join liar [0] (
19:53:16kugeldreamlayers: according to amiconn, cygwin could be problematic
19:53:32kugelyou don't have drive letters there but // is still specially treated
19:53:53amiconn// starts an UNC path
19:54:37 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
19:55:15gevaertsSo this entire issues is about RaaA on windows?
19:55:48kugelnot only. IMO we should handle /. correctly anyhow
19:55:54*amiconn doesn't know whether RaaA even works on cygwin
19:56:06kugelsince we do it for non-/ already
19:56:36dreamlayersWhen building simulators in Windows, Cygwin can be used as a development environment, but the simulators themselves are ordinary Windows applications, without dependence on Cygwin.
19:57:06gevaertsAre simulators actually affected?
19:57:26dreamlayersNope, because they put their files in a simdisk subdirectory.
19:57:45dreamlayersIt's only a problem if the path given to the OS starts with "//".
19:57:59kugelI think they can possibly pass /path/to/simdisk//foobar
19:58:08gevaertsSo we're keeping trunk broken for actual supported targets because // might break a setup that isn't even known to work at all?
19:58:23kugeliirc io.c only prepends the real path without fixup
19:58:32 Quit Xerion (Read error: Connection reset by peer)
19:58:33dreamlayers// in the middle of paths is ok
19:58:51dreamlayersIt only has special significance at the start (like amiconn said).
19:59:18kugelgevaerts: a fix to handle "/." is available
19:59:24kugelit's not committed yet
20:00:08 Join Xerion [0] (
20:00:52gevaertsWhy not?
20:03:48dreamlayersI think the proper fix is to set HOME_DIR to "/" for native targets, but it seemed others disagreed. I only see a problem for hypothetical non-existant targets like ypr0 but running cygwin. It should be ok for RaaA on Cygwin even.
20:03:48 Part pocketprotector ("Leaving")
20:04:48kugelthe fix can be applied regardless how HOME_DIR ends up, no?
20:06:26gevaertsI think so
20:08:20dreamlayersopendir_uncached_recognize_dot.patch can be applied, and it is sufficient to fix the issues I observed. It's not sufficient for "/./<volumename>" and if you want ".." to fully work, that adds a lot of complexity because it can be used to get back to the root and go into a different volume.
20:10:36gevaertshm, wait
20:11:00gevaertsHOME_DIR is only set to /. for non-application builds?
20:11:17gevaertsIn that case the // situation on windows can't occur *at all*, can it?
20:16:15kugelRaaA with /. is possible
20:16:16 Join lorenzo92 [0] (
20:16:32gevaerts#if !defined(APPLICATION) || defined(SAMSUNG_YPR0)
20:30:14 Join tessarakt2 [0] (
20:30:30 Quit Keripo (Quit: Leaving.)
20:32:18dreamlayersTo what extent should "." and ".." be supported? Do they need to work in all cases, like in familiar operating systems?
20:33:15 Quit tessarakt (Ping timeout: 240 seconds)
20:34:34dreamlayersCurrently, they're only supported by code in dircache. In opendir_uncached() they only work because FAT directories other than the root actually contain them as real directory entries.
20:35:32lorenzo92kugel: removed all // comments, reviewed GPIO list and added multiple keystrokes ;)
20:42:15kugellooks good
20:43:19 Join z180 [0] (
20:43:25lorenzo92kugel: lebellium told me you tried to enable arm optimizations, rigth?
20:44:01kugelthat's right
20:51:39 Join TheLemonMan [0] (
21:00:25***Saving seen data "./dancer.seen"
21:05:51lorenzo92kugel: success?
21:06:05kugelwell yes but mp3 crashes
21:11:11bluebrotherhmm, CIA on vacation?
21:15:27lorenzo92kugel: okay ;)
21:19:37 Part jlbiasini
21:23:24 Quit dreamlayers (Quit: Bye)
21:31:19 Quit stripwax (Quit:
21:36:51 Join Kiwi_CAM [0] (~Kiwi_Cam@
21:42:35 Quit benedikt93 (Quit: Bye ;))
21:44:11 Join BRkSYs [0] (~t7DS@
21:45:22kugellorenzo92: at least I managed to compile gdb and run rockbox within
21:45:50lorenzo92ah! debugger hehe
21:46:25 Part BRkSYs
21:48:40 Quit anewuser (Ping timeout: 255 seconds)
21:52:35lorenzo92kugel: so you have run rockbox on R0 using gdb?
21:54:48lorenzo92seems good, but I still never used it :)
22:11:12 Quit y4n (Quit: 6,000,000 ways to die — choose one.)
22:16:14 Nick T44 is now known as Topy44 (
22:22:02 Quit saratoga_ (Quit: Page closed)
22:24:56 Join Thra11 [0] (~thrall@
22:27:35 Quit n1s (Quit: Ex-Chat)
22:35:02 Quit factor (Ping timeout: 252 seconds)
22:37:47 Join factor [0] (~factor@
22:38:23 Join saratoga [0] (9803ec71@gateway/web/freenode/ip.
22:39:09saratogaFS #12521 looks interesting
22:39:10fs-bluebot iPod Classic/6G: cpu freq scaling (patches, unconfirmed)
22:43:51kugelsaratoga: checked without asm threads (changing threading scheme can always be a bit dodgy). still crashing mpa though
22:44:35saratogakugel: can you try commenting out the call to synth_full in synth.c and see if it works better?
22:45:17kugelsaratoga: yes. no crash but also no sound
22:51:23saratogakugel: not sure if this really matters, but the register constraints in synth.c look odd on the ASM blocks
22:51:46saratogai think line 815 should be : "=&r" (lo), "=&r" (hi), "+r" (__p) \
22:53:12 Join webguest28 [0] (
22:53:20 Quit webguest28 (Client Quit)
22:54:10saratogaalthough i'm not sure how that would crash things, i would expect just corrupted audio
22:54:47kugelI'm amazed it works on android just fine
22:55:01kugeleven if I configure for armv6 too
22:55:09kugelsaratoga: anything else to change?
22:55:27saratoganot yet, but i'll keep looking
22:57:39saratogai guess worst case theres really only 3 ASM blocks in synth.c, so you could bisect by replacing PROD_O and PROD_A with the c versions and narrow the problem down
22:58:09kugelremember the crash is in the .S file
22:58:25saratogaoh you were able to debug it?
23:00:07kugelkugelsaratoga: crashes in synth_full_arm.S
23:00:08kugel03:29kugelline 174 gdb says
23:00:16saratogamissed that
23:00:29***Saving seen data "./dancer.seen"
23:00:57saratogathe only weird thing i see in that code is saving the stack pointer
23:01:18 Join maffe [0] (
23:01:28saratogaif the assembler messes up the ".section IBSS_SECTION_MPA_ARM,"aw",%nobits" it would crash pretty quick
23:01:42saratogato be honest i don't really understand this kind of gcc stuff though
23:02:00saratogai just copied and pasted that code around since it seemed to work
23:06:00kugelsaratoga: is r4 (D1ptr) perhaps wrong?
23:06:26kugelon line 173 it does ldr from r4 if it's wrong (perhaps unaligned) it could crash
23:07:16saratogahow would it be wrong?
23:07:41kugelit's a parameter passed via stack
23:07:52saratogaso stack overflow?
23:07:53kugelmaybe the #36 is off?
23:08:02kugelor stack overflow
23:08:07kugelalthough I think I ruled that out
23:08:10saratogathe algorithm itself is correct
23:08:21kugel(/me tried a giant stack and it fails the same way)
23:09:12saratogai went through and checked the alignment of every access in each loop iteration a while back when i was working on the armv5 version of that code, theres no way for it to be unaligned
23:10:10 Quit lorenzo92 (Quit: ChatZilla 0.9.88 [Firefox 9.0.1/20111228084940])
23:10:31 Join lorenzo92 [0] (
23:10:38kugelwhat if a) r4 doesn't have what the caller passed as D1ptr or b) the caller passes a wrong D1ptr?
23:11:16saratogawe'd have noticed b) by now
23:11:55kugelnah, it loads from r4 multiple times. it would crash earlier
23:13:51kugelbut it cannot possible segfault in the smlal instruction can it?
23:15:39saratogaactually i'm going to check the arm_arm
23:15:54 Join krabador [0] (
23:16:29saratogayes, no exceptions are defined
23:19:14Thra11After upgrading my ipod classic to the most recent rockbox release, it seems the random freezes may have been fixed... (fingers crossed!)
23:19:36Thra11If so, well done to the rockbox devs!
23:21:44kugelsaratoga: meh, it's really random now.
23:21:52kugelI change the line a bit and get the crash a few lines above
23:22:40 Quit lorenzo92 (Quit: ChatZilla 0.9.88 [Firefox 9.0.1/20111228084940])
23:23:44 Quit maffe (Quit: IRC ist obsolet!)
23:24:07saratogaThra11: bunch of patches for the classic posted on the tracker today, might want to check back in a week
23:24:20 Quit krabador (Ping timeout: 248 seconds)
23:25:27kugelsaratoga: it's called successfully 6k times before failing
23:26:54saratogathats almost a second of audio i think, does it sound right or is it just static?
23:27:29saratogahmm no stereo, so half a second
23:27:47 Join Scromple [0] (~Simon@
23:31:26kugelsaratoga: no, not static
23:34:23 Join t0rc1 [0] (
23:34:28kugelsaratoga: grml
23:34:34kugelI don't know what's wrong
23:34:49kugelthe functions are called with the same parameters multiple times just fone
23:35:14saratogahave you tried commenting out synth_full_odd_sbsample but leaving synth_full_even_sbsample?
23:35:59kugelI get crashes in both
23:37:47funmanTheSeven: did you see ?
23:38:12kugelheh, it sounds funny with the calls commented out
23:38:19TheSevenyes, but didn't have time to read through it
23:38:27kugelsaratoga: commenting both out -> no crash
23:39:25saratogai have no idea then
23:39:33saratogathat code just iterates over the same pattern endlessly
23:39:47saratogait should repeat something like every 100 calls
23:40:08saratogawith the entire internal state of the synth function resetting each time
23:40:20saratogai don't see how it can operate through one iteration and crash the next
23:40:29 Quit t0rc1 (Quit: WeeChat 0.3.6)
23:40:43 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
23:42:19 Quit TheLemonMan (Quit: WeeChat 0.3.6)
23:42:25saratogakugel: eventually i'll finish rewriting that code for armv5/6, at which point that function will be irrelevant anyway, so i suppose you could disable the asm in just synth_full for now
23:43:24kugelsaratoga: it with synth_full_odd_sbsample commented out it runs for a few seconds before crashing, with synth_full_even_sbsample commented out it crashes immediately
23:44:57 Quit z180 (Quit: ChatZilla 0.9.87 [Firefox 3.6.25/20111212142243])
23:45:09kugelgrr, with another file it runs longer
23:45:58saratogaif the files have the same sampling rate and stereo channels, it should run identically
23:47:11saratogathe code doesn't actually depend on the program data for flow control, just as arguments to those multiply instructions
23:47:15saratogais it possible something else is trashing the stack or corrupting memory?
23:47:25saratogathat would explain how the result is non-deterministic
23:54:49kugelI try a different toolchain
23:55:15 Quit domonoky (Read error: Connection reset by peer)
23:58:14 Join krabador [0] (

Previous day | Next day