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 2009-06-20

00:04:43 Quit edhelas ("MOVIM - Le rĂ©seau social libre de demain -")
00:05:02LambdaCalculus37sudo apt-get install more_beer
00:06:27 Join ocean__ [0] (i=d59c23f7@gateway/web/freenode/x-b2e13640cd8cbbcc)
00:09:19 Join Thundercloud [0] (
00:12:16ocean__if i upload rockbox to sd card, then do usbtool, if i want to remove it and use the original firmware what should i do (i don't understand how usbtool works right now)
00:14:42 Nick fxb__ is now known as fxb (
00:15:57 Quit courtc (Read error: 113 (No route to host))
00:16:16 Quit LambdaCalculus37 ("Fwump")
00:17:09GodEaterwhat's usbtool ?
00:22:03funmanbertrik: : not exactly the same than r21320, but it's cleaner than using a macro
00:22:38funmanusing the same loop than c200v1 on c200v2 i get worse performance, and more important screen corruption (on partial updates it seems, didn't investigate)
00:22:56ocean__GodEater: the jz/chinachip utils
00:23:19funmanif i use the diff you had committed for the clip i get 181.5 fps unboosted instead of 140 in svn, and 147.5 in svn+use a tail loop in lcd_send_data()
00:23:27ocean__"Jz4740 USB tools"
00:23:43funmanoops lcd_send_data needs to be declared after lcd_send_pixel
00:24:12 Join CH360PO [0] (n=57a6fc3b@gateway/web/cgi-irc/
00:25:44 Quit CH360PO (Client Quit)
00:26:16 Join Ch360PO_ [0] (n=57a6fc3b@gateway/web/cgi-irc/
00:26:34CIA-70New commit by bagder (r21355): initial support for builds and build "scores" (how heavy a build is)
00:31:01 Quit flydutch ("/* empty */")
00:31:16 Join `VL [0] (n=vl@
00:31:23Ch360PO_Hi there, I have a H340 (replaced battery, replaced 60gb hdd). Since rockbox 3.2 i've got the problem that after several minutes of playback the WPS starts to continously fade in and out very fast. I did not find anything in the reported bugs section, so what do i have to do? Report the bug?
00:31:27 Quit JdGordon (Read error: 104 (Connection reset by peer))
00:32:29ocean__i don't know how it works but i don't want to get my new player messed up :P
00:33:00`VLhello. just compiled my first plugin for rocbox (going to make work this one: running in sim, i'm able to select it in apps menu, but on start it reports 'incompatible model'.. where to look?
00:33:42B4gderVL: did you make install?
00:34:05 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
00:34:11`VLB4gder: i've done make fullinstall
00:36:23`VLhm..found nice ifdef ;-) #ifndef SIMULATOR
00:36:25PaulJamCh360PO_: you should first try if the issue is still present in a current build ( ).
00:39:28`VLwow...looks like it works...
00:44:20***Saving seen data "./dancer.seen"
00:46:52kugelfunman: Why don't we have the stack in IRAM?
00:46:56kugelon AMSes
00:47:35funmani didn't think about it, and bertrik says there is no performance gain
00:47:58`VLand even works in hardware... rockbox rulez +)
00:48:32saratogathe no performance gain part sounds really odd
00:48:45saratogaunless dram is just really fast?
00:48:46kugelit does indeed
00:49:09kugelthe AMS iram isn't the fasted, but it should be still faster than DRAM
00:51:07funmanwe should measure performance again
00:52:43ocean__damn have a problem building jz usb tools
00:55:06ocean__make: *** No rule to make target `../../rbutil/sansapatcher/bin2c.c', needed by `bin2c'. Stop.
00:55:13ocean__... :(
00:56:02funmanocean__: try make clean in tools/
00:56:48 Quit n1s (Remote closed the connection)
00:56:56ocean__i'm in j4740_tools dir got snapshot from git and tried make clean-linux but give errors with rm
00:57:51ocean__make linux give the precedent error
00:59:34ocean__i also have libusb (slackware)
00:59:43ocean__i can't understand why this error pop up
01:00:10CIA-70New commit by funman (r21356): jz4740_tools : bin2c.c location has changed
01:00:23funmanocean__: git pull ^^
01:00:38B4gderor svn update
01:00:52ocean__already got svn tree checked out 1 hour ago :P
01:01:13ocean__i was building up the vx747p firmware/bootloader
01:01:20 Part toffe82
01:01:26ocean__the need usbtools to upload but they don't compile
01:01:56 Quit Ch360PO_ ("CGI:IRC (EOF)")
01:03:37CIA-70New commit by bagder (r21357): delete client when client disconnects
01:03:39CIA-70New commit by zagor (r21358): Send HELLO with mostly proper params.
01:04:20ocean__lol update :P
01:04:27 Join markun_ [0] (
01:04:46ocean__make: *** No rule to make target `jz_xloader.bin', needed by `jz_xloader.c'. Stop.
01:05:03 Join Bagder [241] (n=daniel@rockbox/developer/bagder)
01:05:52 Join Domonoky1 [0] (n=domonoky@
01:06:30 Quit kugel (Nick collision from services.)
01:06:38 Join kugel [0] (
01:06:58 Quit Domonoky (Read error: 104 (Connection reset by peer))
01:07:03ocean__a new error popped up :P
01:07:05 Join perfectdrug [0] (
01:07:33 Quit PaulJam (Nick collision from services.)
01:07:39 Join PaulJam_ [0] (
01:08:37kugelfunman: how can we messure stack performance best?
01:08:52ocean__maybe that one is my fault let me double check that mips-gcc is in my path
01:09:02funmankugel: i think test_codec is fine since codecs are quite intensive
01:09:22ocean__yes mipsel-elf is in path i don't know why this error
01:11:31ocean__CROSS variable in jz_xloader Makefile is bad "mips-linux-" should be mips-elf- with latest rockbox/gcc mips :)
01:11:48funmankugel: or you could use a recursive function ?
01:12:05ocean__howewer still same error :\
01:13:25funmanocean__: i don't know what is that file, perhaps you have to provide it yourself ?
01:14:03funmanyou said you had cloned the jz_xloader.git tree, look in there
01:14:48ocean__to build jz4740 from svn jz_xloader is needed as stated in the wiki, but there's not a rule to make those files (and makefile inside jz_xloader latest git snapshot have bad CROSS variable), maybe it should be added to the usbtools makefile
01:15:39ocean__yes after modifying CROSS variable it builds and i'll copy them to jz4740_usbtools dir but it would be nicer if those works with regular make
01:16:37funmanocean__: well if you have a patch just send it
01:16:57ocean__i'm not good with Makefiles, i'll try to get it working though
01:18:07 Join mcuelenaere [0] (
01:18:20ocean__damn usbtool says device not found :(
01:18:29 Quit BHSPitLappy (Connection timed out)
01:18:42mcuelenaereocean__: having problems with usbtool?
01:18:59ocean__got to compile svn but now says it doesn't find the device
01:19:11mcuelenaereyou're using sudo?
01:19:14mcuelenaerewhat does lsusb give you?
01:19:29ocean__yes i used su i'm on a slack never use sudo :P ahah
01:20:09 Quit markun (Read error: 110 (Connection timed out))
01:20:30 Quit B4gder (Read error: 110 (Connection timed out))
01:20:35*mcuelenaere checks some source code
01:21:13mcuelenaereocean__: uhh, are you in usb mode?
01:21:21ocean__lol maybe not
01:21:32mcuelenaereyou do know of the reset hole, do you? :)
01:21:42ocean__i read it somewhere in the wiki i think
01:21:51mcuelenaerereset the player while holding the MENU button
01:21:59mcuelenaereit'll show up as an other VID/PID combination
01:22:07mcuelenaerethat should be in the wiki I think :)
01:22:27funmanwell i get the exact same performance for read/write to iram or dram on sansa c200v2
01:23:01 Quit GodEater ("Terminated with extreme prejudice - dircproxy 1.0.5")
01:23:18ocean__ah mcuelenaere i got this dmesg sdc don't get a valid partition and no errors (while on windows works) you have similar problems?
01:23:20funmani.e. X loads take 108 ticks for iram or dram, X writes take 298 ticks for iram or dram
01:23:23 Join GodEater [0] (n=yeahrigh@rockbox/staff/GodEater)
01:23:49saratogahow long is a tick? 1 clock?
01:23:50mcuelenaereocean__: the device shouldn't show up as a MSC
01:24:05funmansaratoga: 1 second / HZ (= 10ms)
01:24:21mcuelenaereocean__: did you reset the device (reset hole) while holding the MENU button?
01:24:22ocean__yeah for usbtools, i mean i cannot use the 8 gb samsung flash regularly on linux :P
01:24:28 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
01:24:29 Quit bertrik (Read error: 110 (Connection timed out))
01:24:29saratogafunman: how much do you copy during that time?
01:24:37mcuelenaereocean__: you won't, the OF won't expose it
01:24:50funman for(x = 0; x < max ; x++)
01:24:50funman i_test[x%512] = x;
01:24:54mcuelenaereyou can view it with Rockbox (and a hack) though
01:24:57funman y = i_test[x%512];
01:25:15saratogafunman: would you post a full diff? i'd like to play with this
01:25:48ocean__usbtool done :D
01:25:56funmansure, but i just thought perhaps i should disable interrupts to be really precise, and use timer2_value for precise timing
01:26:03ocean__howewer on windows the 8gb fat partition show up on linux not :(
01:26:23mcuelenaereocean__: that's something with your computer/linux setup I guess, here it works fine (Linux too)
01:26:37ocean__mcuelenaere: Onda VX747+ rockbox bootloader
01:26:41ocean__loading firmware
01:26:45ocean__error:file not found
01:26:46funmansaratoga: (in debug menu -> view I/O ports
01:26:53mcuelenaereocean__: you did put it on your SD, did you? :)
01:26:59mcuelenaereand that proves the ATA driver works
01:27:02mcuelenaereSD* on the sd right?
01:27:12 Quit Thundercloud (Remote closed the connection)
01:27:22ocean__ah lol
01:27:32mcuelenaereRockbox doesn't support zip files ;)
01:28:13ocean__lol sorry :P
01:28:16Unhelpfulit (sometimes) supports tars though :)
01:29:00saratogafunman: try stepping through 32 bytes at a time so that you don't hit the cache
01:30:03ocean__mcuelenaere: loading firmware lenght FFFFFFF8 error: read failed (chksum)
01:30:19mcuelenaereehm, try re-unzipping it and try again
01:30:23funmansaratoga: but the whole buffer can be cached anyway no ?
01:30:30*mcuelenaere never had a checksum failure before
01:30:39saratogafunman: what do you mean?
01:30:59Tornelenth fffffff8 would be, er, -8
01:31:07Tornei think the problem may be before the checksum :)
01:31:09funmanit takes a few time to load the whole 512 bytes in the cache
01:31:25saratogacache lines are only 32 bytes, so the whole buffer isn't loaded
01:31:41saratogajust 32 bytes, which results in 8 cache hits, then a miss, then a 32 byte fetch, ...
01:32:05 Join bertrik [0] (
01:32:26funmanbut there is several lines in the cache
01:32:38ocean__mcuelenaere: i can also try to recompile i have done a build from svn without errors i can recompile it or use the one on wiki
01:32:57ocean__mmm in fact there's something wrong with the zip
01:33:06mcuelenaereocean__: you mean a usbtool build? If you're running Linux, I would advise you to build it your own
01:33:19mcuelenaereocean__: try downloading one from
01:33:26ocean__yes i've builded everything rockbox included
01:33:28saratogafunman: cachelines are only loaded after an access to one of them
01:33:39saratogathe CPU has no way to know that the buffer spans more then one cache line
01:33:45ocean__usbtool have given some problems from jz_xloader
01:33:57saratogaalthough perhaps if its very clever it can speculatively load sequential cachelines, though i don't know if arm does this
01:33:59ocean__but got it to build somehow :P
01:34:11mcuelenaereocean__: what problems? what toolchain are you using?
01:34:23mcuelenaerebut if you got as far as running the bootloader, I don't think jz_xloader is the problem
01:34:27Tornesaratoga: it doesn't on most current ARMs, but this is what PLD is for
01:35:09funmansaratoga: but there is so much loops than all the buffer will have been quickly fetched in separate lines
01:36:00 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
01:36:09saratogawell yeah thats another problem :)
01:36:19funmanso what i measure actually is the cache speed :/
01:36:37saratogausing a 16KB buffer and stepping through 32 bytes at a time would help
01:37:10funmanor disabling cache :)
01:37:11Unhelpfulassuming there's no auto-prefetch :)
01:37:29Torneyeah, ote for "use an uncached region"
01:37:30ocean__mcuelenaere: i've followed the info on wiki, mips elf gcc 4.1.2
01:37:32Torneor "disable teh cache" :)
01:37:48ocean__and regular x86 toolchain 4.3.3
01:37:56TorneARM implementations are allowed to have their cache do whatever the hell it likes pretty much
01:38:01Torneso it might well prefetch
01:38:17Tornethe ones i've used don't seem to prefetch in L1, but it's ahrd to prove :0
01:38:44mcuelenaereocean__: try re-extracting the zip
01:38:49ocean__mcuelenaere: jz_xloader makefile had a bad CROSS variabile (mips-linux- should be ,mips-elf-) and included makefile in usbtools don't generate jz_xloader
01:39:11saratoga32 bytes prefetch would probably hurt more then it helped, since it'd lock the memory controller for quite a few cycles
01:39:19mcuelenaereno, you should generate jz_xloader.bin yourself
01:39:27mcuelenaereI think the README points that out
01:39:38funmandata cache disabled : same exact performance
01:39:52ocean__mmm i didn't understand the README then
01:40:05ocean__i think my english is badder thank i think :P
01:40:10funmanon c200v2, perhaps it's different on fuze
01:40:33Tornesaratoga: yah. the tendency is to just depend on the cpu issuing PLD
01:40:33saratogai get 7 ticks IRAM, 18ticks DRAM
01:40:44Tornesaratoga: quite a few of the L2 controllers prefetch though
01:40:53funmansaratoga: increase max to get more precision
01:40:58Tornebut i don't know of any mp3 players with L2 :)
01:41:01saratogaDRAM write being one slower
01:41:08saratogaTorne: Gigabeat S
01:41:28Torneand have you got it to work? :)
01:41:29saratogasame as Zune actually
01:41:33 Quit __lifeless (Read error: 60 (Operation timed out))
01:41:35saratogayeah its fast as hell
01:41:43 Join roolku [0] (
01:41:46saratogacan actually play APE files
01:41:58Tornei gues sit's an L210?
01:42:15funmaniram is a tad faster on fuze
01:42:33funmanfor writes, and a tiny tiny tad slower than dram
01:43:00Tornesaratoga: which remidns me, there's no preloads in the ARM mem* functions :)
01:43:09funmanhum it seems to depend if the cpu is boosting or not ?
01:43:41 Quit roolku (Client Quit)
01:44:05saratogahuh i do get the same for I and D, but writes are much slower on both
01:44:10funmanif cpu is boosted the iram loop is faster,
01:44:15 Join roolku [0] (
01:44:21Tornethe iram is probably clocked at the cpu speed
01:44:23funmansaratoga: ~3 times slower
01:44:30Torneat some divisor of it
01:45:07funmanthe as3525 datasheet says it comes from pclk
01:45:27funman"the free running AHB/APB clock (PCLK) for the 1-TRAM must always be present"
01:45:35ocean__mcuelenaere: the sd card in wich FS format should be FAT 32 or FAT 16?
01:45:39Torneooh, imx31 is quite nice isn't it
01:46:05mcuelenaereocean__: it used to be FAT32, but since some revisions ago FAT16 will work too
01:46:16 Quit gevaerts (Nick collision from services.)
01:46:25 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
01:46:28 Quit roolku (Client Quit)
01:46:33UnhelpfulTorne: it is. it has vector extensions, too. i've been working on asm IDCT of late, and it can, for the most part, work two columns at a time. :)
01:46:52TorneUnhelpful: bah. obsolete sstill though :)
01:46:54funmansaratoga: so can we conclude than the iram is a very tiny bit faster than dram ?
01:46:57TorneVFPv2 is icky
01:47:14gevaertsmarkun_: Bus 001 Device 011: ID 0492:0140 Samsung SemiConductor, Inc.
01:47:38UnhelpfulTorne: is that the vector coprocessor? i'm using tnhe CPU partitioned integere math instructions.
01:47:42saratogafunman: I get that one iteration boosted takes 4 clocks
01:48:01Torneoh, armv5e? yah those are helpful too
01:48:08Tornevfpv2 is the vector floating point unit, yah
01:48:18saratogathat seems to fast
01:48:24saratogai wonder if GCC is outsmarting us
01:48:28saratogatoo fast
01:48:36Unhelpfulthose are armv6. armv5e gives you the very nice top/bottom 16x16->32 multiply instructions.
01:48:51funmansaratoga: well using current_tick is not very precise because the cpu is interrupted to run through the tick tasks
01:48:57Tornei forget. i'm not great with math :0
01:49:04Tornei only have to deliver the OS support for it )
01:49:14TorneNEON is infinitely more hilarious though
01:49:44Tornesaratoga: got perfcounters on that chip? :)
01:49:45Unhelpfulthat's the coprocesser with wide vector registers, 64- and 128-bit i believe?
01:50:10Torneyah, though they are technically the same registers as VFP so if you use then as 128bit there aren't very many
01:50:19Tornebut NEON has integer vector ops as well as floating point, is the trick
01:50:31 Quit jgarvey ("Leaving")
01:50:38saratogafunman: interrupting the thread would make things slower, n ot faster
01:50:41Unhelpfulah... so i can do the partitioned adds and such across wider vectors.
01:50:55ocean__mcuelenaere: used the build you linked doesn't work, maybe because is the 747 and i builded the 747p bootloader
01:51:00Torneyup, 64x32bit or similar
01:51:06ocean__i'll build new bootloader then
01:51:14mcuelenaereocean__: ah yes, forgot about that part
01:51:24UnhelpfulTorne: the beast also has perfcounters. i don't know about the AMS targets.
01:51:37Tornearmv6 usually has them, armv5 may or may not
01:51:49Tornethey are not documented for ARMv5 but the same registers as int eh armv6 manual generally work
01:51:52saratogacan i mark the entire function as volatile in c?
01:51:59Tornelikewise onchip debug registers
01:52:07TorneARM hadn't decided whether they were done changing it yet in v5 I think :)
01:52:14mcuelenaereocean__: hmm yes, doesn't offer any builds for ondavx747p
01:52:21funmansaratoga: i don't think, but you can check disassembly to be sure
01:52:33Unhelpfulsaratoga: i think there's an optimize level pragma for gcc. "volatile" only applies to inline asm, doesn't it?
01:52:34saratogai was afraid you'd say that
01:53:37Unhelpfulsaratoga: why not just make the array access volatile asm? array[N] is pretty easy.
01:53:54Tornevolatile should work for C too
01:54:01Tornein gcc, anyway
01:54:14funmanbut for functions?
01:54:15ocean__mcuelenaere: with 747 bootloader gets loaded but isn't working really good i think
01:54:25Torneoh, not on the function
01:54:31Torneyou need it on the data object
01:54:37mcuelenaereocean__: how so?
01:54:45ocean__mcuelenaere: no no it's working my fault :P
01:55:02ocean__now it's time to try if i can build a working one for vx747p
01:56:29ocean__the "M" button seems not to work
01:56:35ocean__volume buttons works
01:57:11ocean__at least the "-" one
01:57:22saratogaadding volatile makes them all take 40 ticks
01:57:48saratoga163 for all unboosted
01:58:35Tornedoes the asm look right? :)
01:58:37saratoga12 and 49 clocks respectively
01:58:47saratogai haven't checked yet
01:59:03ocean__mcuelenaere: read capacity failed
01:59:11ocean__lol i think i'm messing up something
01:59:29mcuelenaereocean__: ?
01:59:39ocean__somehow the sd isn't working on linux
01:59:49mcuelenaerethrough Rockbox?
01:59:54ocean__no through linux
02:00:03ocean__original firmware sees it
02:00:04mcuelenaereyou mean it's in a card reader?
02:00:26ocean__no it's inside the dap
02:00:33ocean__i think i'll get a reboot
02:00:34mcuelenaereand the DAP is running the OF?
02:00:41ocean__the OF
02:01:00ocean__i'll reboot and try to upload them with windows then back to linux and use usbtool
02:01:06mcuelenaeretry using another usb port?
02:01:09ocean__mmm i try
02:01:25mcuelenaerebtw usbtool is available for windows too
02:01:34ocean__sd 10:0:0:0: [sdc] READ CAPACITY failed
02:01:51mcuelenaeresounds like something wrong with the OF usb stack
02:01:57funmansaratoga: with interrupts disabled, same r&w speed when unboosted, same r speed when boosted, iram w 10% faster
02:02:35saratogafunman: I'd be most interested in figuring out exactly how many clocks a single read and write to each takes
02:02:51ocean__no ok are my usb ports
02:02:54ocean__under linux
02:03:02funmanand i see iram w 10% faster for fuze when boosted or unboosted
02:03:05ocean__maybe there are some usb stack problems but not on the OF but on linux :P
02:03:10ocean__reboot! :)
02:03:11 Quit ocean__ ("Page closed")
02:03:14saratogatheres a way to get gcc to spit out the assembly annotated with the c code right?
02:03:20funmansaratoga: we don't have enough precise clocks :/
02:03:25funmanthe timer is 1.5MHz
02:03:30funmancurrent_tick is 100Hz
02:03:54saratogafunman: 100Hz is fine if we average over many seconds
02:05:06Unhelpfulsaratoga: -save-temps -fverbose-asm gives you assembly annotated with variable names
02:05:20Unhelpfulor from a debug object your can use objdump -S -d
02:05:38Unhelpfulwhich will interleave C source lines with asm
02:06:14Slasherimany tagcache bugs found & fixed, probably committing a big fix tomorrow :)
02:07:48CIA-70New commit by bagder (r21359): first basic BUILD support
02:08:27UnhelpfulSlasheri: i'm guessing that late-index-open fix doesn't address the speed problem i was seeing?
02:09:16SlasheriUnhelpful: nope, that is a whole different issue
02:09:23SlasheriUnhelpful: what kind of speed problem do you mean?
02:10:06Slasherinow i have fixed some pure data retrieval problems from the db
02:10:07 Quit Nico_P (Remote closed the connection)
02:10:09UnhelpfulSlasheri: reading all filenames took 1s with ramcache off vs 6s with it *on*, on my beast. very weird.
02:10:10 Quit perfectdrug ()
02:10:27funmani think the %512 is badly optimised :D
02:10:45SlasheriUnhelpful: ah, that actually might be fixed on my next commit
02:11:20Slasheriramcache + dircache (+ statefile on H1xx BL v7) had problems at least
02:11:30saratogayeah this should probably be done in assembly
02:11:31 Quit funman ("free(random());")
02:12:58SlasheriUnhelpful: did you have disk or flash player?
02:13:31Unhelpfuli was also going to take a look at having the string tag reads return snprintf-like values, it would be useful in pictureflow which wants to know exactly how much more buffer it needs if it doesn't have enough.
02:14:27Unhelpfulbeast, so disk.
02:15:03 Join ocean [0] (i=d59c23f7@gateway/web/freenode/x-9c7b01be350b2494)
02:15:27oceanmcuelenaere: vx747p build seems not to go bad checksum
02:15:49mcuelenaereso it works?
02:16:04oceanlenght: 911DC checsum: 801F3C sum: 25F532B
02:16:14ocean747 build you've linked to me go
02:16:27oceanmine 747p build don't with that checksum message
02:17:57oceani can try to build another time Normal and Bootloader and upload
02:19:24mcuelenaerethat's weird
02:19:28oceanbut don't know if it could be of any help :P
02:19:37mcuelenaerewhere did you get your toolchain, did you use tools/
02:19:46oceanor can try to build a 747 one and see if it works and 747p not
02:20:17 Join Thundercloud [0] (
02:22:37 Quit Sajber^ (Connection timed out)
02:23:09mcuelenaeresame versions as me
02:23:16mcuelenaereI haven't tried a vx747p build though
02:23:39mcuelenaereI'll look at it tomorrow, I'm going to bed now
02:23:53 Nick fxb is now known as fxb__ (
02:23:57oceanyes i too should go to sleep
02:24:24 Quit mcuelenaere ("Zzzzzzzzzzzzzzzzzz")
02:26:38 Quit ocean ("Page closed")
02:29:03CIA-70New commit by bagder (r21360): fix the delete again when I changed to ->fileno
02:29:52saratogabagder: any idea when the new build system will be ready to try out?
02:30:15Bagderno, but I hope we'll get some first testing tomorrow
02:31:19CIA-70New commit by zagor (r21361): First BUILD support.
02:34:04saratogaany chance it'll be able to do sftp tranfer over the mean ssh tunnel ?
02:34:54Bagderatm we're not even using ssh...
02:37:08saratogajust direct TCP connections?
02:39:46Bagderyes. we haven't yet decided exactly how to do it for real, but we want to create a system where we can have admins that can add new clients
02:40:02Bagderso creating ssh accounts for each client doesn't feel very good
02:41:18Unhelpfuli'm thinking of putting some of this into a pluginlib header file. it's based on bench_jpeg_mem.c and bench_scaler.c, but offers flexible output, so you can use logf, or a pre-opened file, or whatever else as long as it's a printf replacement. it's working pretty nicely for benchmarking IDCT functions... i'm not sure if the busy-wait for the tick is really worth bothering with, though?
02:42:02Mikachuthe wiki page said ssl, not ssh
02:42:56Bagderexactly, just to avoid needing regular accounts
02:43:04saratogathat'd be nice
02:43:08Bagderbut really, we can start with plain tcp
02:44:22***Saving seen data "./dancer.seen"
02:44:26saratogagetting rid of the requirement to let people SSH in would make the build client a lot more palatable around here
02:44:59Bagderwell we've already reversed the roles, so we're running build clients that connect to the server
02:45:14Bagderand keep the connection
02:45:27saratogai could reasonably ask for permission to run something like that on our machines
02:45:37saratogawhereas before they'd never let me setup remote ssh accounts on cluster boxes
02:46:11Bagderit will also be easier to control when they run, so you can stop them and start them again if they ever conflict with anything
02:49:47 Join Jiremi [0] (n=627693fa@gateway/web/cgi-irc/
02:50:34Jiremii'm trying to install rbutil under linux
02:50:48Jiremii downloaded the archive file that contains rbutilqt-v1.2-64bit
02:51:05Jiremibut i'm not sure what to do with rbutilqt-v1.2-64bit
02:51:19Mikachutry chmod +x rbutilqt-v1.2-64bit; ./rbutilqt-v1.2-64bit
02:51:34Mikachu(it is probably already executable)
02:52:14Jiremiawesome, thanks
02:54:02 Quit PaulJam_ (".")
02:54:09 Quit jds_ ()
02:54:26Unhelpfulsaratoga: it appears that the second operand is taken whole, and the early termination happens on the size of the third operand. 128 muls of 4096 * 120 is a bit over 2x faster than 128 muls of 120 * 4096. just for future reference :)
02:54:28 Quit Jiremi (Client Quit)
02:54:56Unhelpfuli have no idea if a second non-dependent instruction can execute during mul's extra cycles?
02:55:12CIA-70New commit by bagder (r21362): send a more complete BUILD command
02:55:34Bagdertime to sleep
02:55:37 Quit Bagder ("It is time to say moo")
02:56:51 Quit petur ("Zzzzz")
02:56:51 Quit Zagor ("Leaving")
02:57:19 Part Domonoky1
02:57:22saratogaUnhelpful: i doubt it can proceed to the write back stage, but maybe it can get as far as decode?
02:57:57saratogayou could try doing a load immediately after a mul and see if its any faster
02:58:18Unhelpfuli'll have to try using all multiplies in the armv4 asm, and see if i can slot some of the other operations between the mul and the two mla's that want its output.
02:58:46saratogai bet you can swallow at least one clock off the load time if you put it after the mul
03:00:06 Quit markun_ ("sleep")
03:02:04Unhelpfulthe multiply cycles are "internal". i've not yet gotten to the part that explains how the various types of cycles interlock, but mul takes 1S + mI cycles, with m being the multiplier size in bytes.
03:03:15saratogaits just a question of when the subsequent instructions stall, after decode or before write back
03:03:24 Quit kugel (Read error: 60 (Operation timed out))
03:05:42 Quit Thundercloud (Remote closed the connection)
03:06:17Unhelpfulthe arm11 docs seem much more clear on this, they specify things in terms of result latency... aside from extra bits like "early" registers, anyway.
03:18:13 Quit bertrik (Read error: 110 (Connection timed out))
03:22:22 Join jfc [0] (
03:33:12 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
03:44:35 Quit amiconn (Nick collision from services.)
03:44:36 Join amiconn_ [50] (n=jens@rockbox/developer/amiconn)
03:44:52 Join pixelma_ [50] (n=pixelma@rockbox/staff/pixelma)
03:44:52 Quit pixelma (Nick collision from services.)
03:44:54 Nick amiconn_ is now known as amiconn (n=jens@rockbox/developer/amiconn)
03:45:08 Nick pixelma_ is now known as pixelma (n=pixelma@rockbox/staff/pixelma)
03:49:51Unhelpfulsaratoga: i played a bit with moving other instructions around the multiplies. the third function here is almost, but not quite, as fast as the first two. (<1% after accounting for the time needed to load the cofficient buffer with data before each IDCT)
03:50:46Unhelpfulthe variable-length multiply thing makes benchmarking this on arm7tdmi a bit touchy, since having weird data in the buffer can change the time to execute :/
04:07:04 Quit _Auron_ ("Infinity repeatedly denies rumours of plotting with zero to bring down the Universe.")
04:22:26 Join Xerion_ [0] (
04:27:53 Quit dmb (Read error: 104 (Connection reset by peer))
04:28:05 Quit efyx (Client Quit)
04:28:06 Join dmb [0] (n=dmb@unaffiliated/dmb)
04:33:46 Quit dmb (Read error: 104 (Connection reset by peer))
04:36:50 Join dmb [0] (n=dmb@unaffiliated/dmb)
04:39:32 Quit Xerion (Read error: 110 (Connection timed out))
04:39:32 Nick Xerion_ is now known as Xerion (
04:44:26***Saving seen data "./dancer.seen"
05:09:29 Quit dmb (Client Quit)
05:11:19 Join dmb [0] (n=dmb@unaffiliated/dmb)
05:14:12 Quit froggyman (Read error: 113 (No route to host))
05:18:22 Quit dmb (Client Quit)
05:33:35 Join dmb [0] (n=dmb@unaffiliated/dmb)
05:36:22 Join kkurbjun [0] (
05:43:34 Join perrikwp [0] (n=perrikwp@
05:52:30 Quit dmb (Client Quit)
06:05:15 Join courtc [0] (n=court@unaffiliated/courtc)
06:07:18 Quit intrados_ (Connection timed out)
06:23:12 Join dmb [0] (n=dmb@unaffiliated/dmb)
06:30:20 Join _Auron_ [0] (
06:44:29***Saving seen data "./dancer.seen"
07:01:54 Quit krazykit (Read error: 113 (No route to host))
07:06:26 Join _lifeless [0] (n=lifeless@
07:15:03 Join webguest44 [0] (n=4cf16035@gateway/web/cgi-irc/
07:16:52webguest44I encoded my video but the file doesnt come up in the file browser
07:20:03advcomp2019webguest44, did you look at the mpegplayer wiki
07:21:11webguest44Yes I have it open
07:22:38advcomp2019did you follow it? where did you put the mpeg at?
07:22:53webguest44I made a folder called Movies in my root.
07:23:30advcomp2019do you see that folder in the file/folder view?
07:23:40webguest44I just dont see the file.
07:24:22advcomp2019are you sure you converted the file to mpeg right?
07:26:20advcomp2019what steps did you used if i ask
07:27:11webguest44I used WinFF to convert it to Rockbox. With Ipod nano
07:29:15advcomp2019have you tried showing all files?
07:32:36 Quit scorche|sh (Remote closed the connection)
07:32:39 Join scorche|sh [0] (
07:33:54webguest44Oh, no I havelt I will
07:36:30 Join intrados [0] (
07:37:55webguest44Works, Thank you.
07:37:58 Quit webguest44 ("CGI:IRC")
07:49:52 Join Horschti [0] (n=Horscht2@xbmc/user/horscht)
07:55:52 Join D480 [0] (
08:06:19 Quit Horscht (Read error: 110 (Connection timed out))
08:08:52 Quit D480 ("Ex-Chat")
08:22:10 Join PaulJam [0] (
08:22:51 Quit scorche|sh (Remote closed the connection)
08:22:53 Join Rob2222 [0] (
08:22:56 Join scorche|sh [0] (
08:35:07 Quit _Auron_ (Read error: 54 (Connection reset by peer))
08:40:19 Quit Rob2223 (Read error: 110 (Connection timed out))
08:40:32amiconnUnhelpful: How are you benching just the idct?
08:40:51amiconnThis might be interesting for benching my mpeg2 idct as well
08:41:34Unhelpfulamiconn: i can post a patch for my test-plugin branch if you want, but the bulk of the "benchmark framework" stuff i put on pastebin a few hours ago, let me find the link :)
08:42:15Unhelpful <- basically call DO_BENCH or DO_BENCH_CUSTOM with the function or macro and the arguments.
08:43:49Unhelpfulby the way, it tells me that the zero-AC test in the C horizontal IDCT is a loser on ARM. i'm still coding it into the ASM, to see if it wins when it's done a bit more cleverly.
08:44:31***Saving seen data "./dancer.seen"
08:45:02Unhelpfulit also tells me that mla beats shift/add in the four-point idct... probably only because that one is small enough to keep the constants in registers.
08:46:57amiconnMy problem is that code paths do depend in input values in a way I don't fully understand yetz
08:47:37amiconnWhen trying the idct in mpegplayer on the beast, I can't see a significant difference
08:48:33Unhelpfulthat's why i started benchmarking IDCTs by themselves. it's quite difficult to see the difference benchmarking jpeg file decode, if the difference is small.
08:48:38amiconnI can think of two reasons for this. (1) my armv6 code sucks and ins't any faster than the shift/add thing we're using for all arms now
08:49:17amiconn(2) the idct time is almost insignificant compared to other decode stages
08:50:25Unhelpfulwell, i can't say how it'll play out on the 8-point idct yet, but on the 4- and 2-point sizes i saw much larger gains on ARMv6 from doing two columns at once with the SIMD instructions.
08:51:08Unhelpfulthe horizontal IDCT sees a *tiny* gain from some SIMD use, but it's nowhere near the 15-25% that the vertical passes get by doing two columns at once
08:52:00amiconnI'm using the half-reg multiplies, and even smuld(x)/smlad(x) where possible
08:52:30amiconn...and usat16 and sadd16 in the output stages
08:53:58Unhelpfulthat's basically the strategy i took on the 4-point. at that size the constants can all fit into registers, although this may mean saving an extra register or two. i'm only tackling armv4 on 8-point now. it's a good deal more complex, and i'm having to spill values to stack :/
08:55:09amiconnmpeg2 is only 8-point. I can't keep the constants in regs, but I load them 4 at once (using ldrd)
08:56:11Unhelpfulcan you use the ldr reg, =constant notation with ldrd as well?
08:57:02 Quit dmb (Read error: 113 (No route to host))
08:59:55amiconnI don't know
09:02:44Unhelpfulhrm... nope, can't get auto-generated constant pool for ldrd. the nice thing about that notation with ldr is that it generates a mov instruction for valid immediates.
09:03:40amiconnWell, those constants definitely aren't valid immediates
09:04:03*amiconn needs to check whether he can switch register allocation, saving loads this way
09:07:48amiconnSeems doable... I'll try this when we're back at Dekimo
09:09:30 Join flydutch [0] (
09:10:48Unhelpfuldoes the mpeg-2 IDCT use *exactly* the same matrix as JPEG? and is the C version the same algorithm as in the ASM?
09:12:59amiconnThe C code seems to work quite different
09:13:39amiconnWhat I did so far was basically porting the coldfire version of the idct (which is optimised for a mac unit) to armv5(+)
09:16:48Unhelpfulthe C IDCT we have for JPEG is afaict lifted from IJG almost verbatim. there aren't all that many MAC operations in it, really... two in the four-point, four in the 8-point and one product-of-sum that could use the multiply-add operation if i double up the constant multiplier.
09:18:19 Quit _lifeless (Remote closed the connection)
09:18:35 Join _lifeless [0] (n=lifeless@
09:23:08Unhelpfulrasher: seriously though, the major core use for image loads is album art, and that should never need a large static buffer - we do just fine grabbing needed decode/scale space from the buffer as needed, buffering.c was basically already handing the bmp loaded the whole buffer to use, and then completing the allocation after getting a size from the loader.
09:27:37Unhelpfuli'd really like to see a C implementation, or a readable-to-normal-people-not-involved-deeply-in-FFT-theory representation of a MAC-optimized IDCT. which probably means a C implementation, anything else i've read on DCT optimization has gone quickly over my head :/
09:33:17Unhelpfulfor the ARMv4 version i'm using the one spare register i have to do shift/add sequences for multiplies, and whenever one of those ends with several instructions not using the spare, i ldr a constant into it and take care of one of the few MAC operations :/
09:40:36 Join mcfock [0] (
09:41:32 Join bmbl [0] (
09:47:54mcfockHello! Owning a rockboxed sansa e260, I'm going to record a birthday party including live performances this evening - so I'd like to know which is the 'best' format; taking into account my sansa's disk size (approx. 4GB) and the duration of this event (ca. 6 hours of total recording time)
09:52:12CIA-70New commit by bluebrother (r21363): Update japanese translation for Rockbox Utility (FS #10349 by Yoshihisa Uchida).
09:52:40Galoisrockbox only does mono audio recording. 4GB is like 10 hours of mono audio.
09:52:55Galoisif your disk is partially full, scale accordingly.
09:52:57 Quit BHSPitLappy (Remote closed the connection)
09:53:09mcfockGalois: on the Sansa, it also does stereo mode
09:53:26Galoiswell, yeah, it does, but what's the point?
09:53:31Galoisdo you have a stereo mic?
09:53:39mcfockno, only the internal
09:53:51Galoisthe internal mic is rather small, and mono. There's no point in stereo.
09:54:41mcfockGalois: any recommendations concerning the file format, regarding power consumption?
09:55:16Galoisdefinitely go with wav to minimize the cpu
09:56:05mcfockbut could the sansa's batteries also do wavpack, 6 hours of recording time?
09:56:35GaloisI think they could new, never tried
09:56:49mcfockwell, then I'll stick to wav
09:57:52Galoiseasiest solution is to buy an 8GB micro SD card, they cost like $20 these days. Rockbox can record on those.
09:58:47martian67microsd is so tiny
09:58:49martian67its awsome
09:59:09martian67i wonder if storage will get any smaller
09:59:56mcfockGalois: I am considering this option as I wouldn't have to move my music and stuff over to my pc
10:01:24Galoisas a bonus you'll have more storage for your songs too
10:01:52mcfockyes %)
10:10:43mcfockGalois: thank you, you were right - the mic indeed does only mono, just tried it
10:19:05 Join Zagor [242] (n=bjst@rockbox/developer/Zagor)
10:19:09 Join Bagder [241] (n=daniel@rockbox/developer/bagder)
10:19:24Bagdergoood devcon morning
10:23:05 Join ocean [0] (i=d59c23f7@gateway/web/freenode/x-8328ca7cd57aef80)
10:25:05oceanhello, i need to understand how scramble.c works, is there someone who knows how it works? i have modified the "configure" from tools because checksum wasn't calculated for vx747+, now i get also a checksum error, but checksum and sum differ only by 1
10:25:55Bagderscramble.c is made to scramble the binary according to the given options
10:26:03Bagderit knows a bunch of different ways
10:26:28oceanyes i understood already that by giving a look at the code, and mode add is a simple "addition" for crc
10:26:33mcfockwell, thanks then to Galois! keep up the good hacking rockbox, have a nice day!
10:26:42mcfock*good work
10:26:51 Join bertrik [0] (
10:27:04oceani have seen a modelnumber, is it somehow put somewhere in the firmware?
10:27:06 Join bertrik_ [0] (
10:27:39oceanfor vx747 modelnumber is 45, this model numer is stored somewhere and doublechecked?
10:27:44 Quit mcfock ("Lost terminal")
10:28:22 Quit bertrik_ (Client Quit)
10:28:29oceanon configure tool vx747 is target_id 44 has this something to do with the modelnumber?
10:30:24oceanchksum = modelnum; /* start checksum calcs with this */
10:30:26Bagderno you're right, I think target_id is that number
10:30:39oceanok but it differs by one
10:30:54Bagderdiffers by one compared to what?
10:31:09ocean747 is 45, 767 is 44, 747+ is 54 on configure
10:31:26oceaninstead in checksum only have 44 fo x747
10:32:29BagderI understand that
10:32:35Bagderbut who says the checksum is wrong later?
10:32:52oceanbut build for 747 is ok, when building 747+ checksum is wrong
10:33:02oceanthe firmware says it when it starts on the player :P
10:33:14oceanthe rockbox bootloader i think
10:34:38Bagderso that's the same bootloader for all those models?
10:35:11oceani think yes
10:35:31Bagderthen it should be fixed
10:36:02Bagderbut aren't you use scramble for x747?
10:36:11oceanno add
10:36:13 Join Domonoky [0] (n=domonoky@rockbox/developer/domonoky)
10:36:25Bagder-add=x747 ?
10:36:26 Join markun [50] (n=markun@rockbox/developer/markun)
10:36:28oceanat least from what i can understand from the code
10:36:37 Join kugel [0] (n=kugel@rockbox/developer/kugel)
10:37:00 Join ender` [0] (
10:37:22oceanis there some place in the firmware/bootloader where the modelnumber is stored, if i can compare them and find the 747p i could fix it
10:37:38Bagderfirmware/bootloader loads the image, it doesn't store it
10:37:51Bagderscramble is what puts the checksum to the image
10:38:58oceanok but somehow there's that difference of "1" from checksum and sum and i think it's related to the modelnumber
10:39:17oceani think just changing by 1 the modelnumber it should probably work
10:39:30oceansince checksum starts adding the modelnumber
10:39:45Bagderthe model number is selected from the -add string
10:39:50 Join petur [50] (n=petur@rockbox/developer/petur)
10:39:52Bagderthus you get the same for all x747
10:40:15oceanmmm but that doesn't explain the different checksum
10:40:17Bagderand no it isn't related to target_id in configure
10:40:24Bagderocean: exactly
10:40:55oceani already changed the configure because checksum wasn't calculated at all for vx747+
10:41:07oceanchecksum and sum where completely different
10:41:13oceannow they differ only by 1
10:43:29*Bagder thinks he spots it
10:44:19Bagdertry changing the MODEL_NUMBER in firmware/export/config-ondavx747.h
10:44:34***Saving seen data "./dancer.seen"
10:44:57Bagderbut really, I think scramble should have use number 44 for the 747+ if they're not compatible with each other
10:45:42oceanah here is it! thanks
10:45:54oceani'll change it in scramble adding 747p
10:46:02oceanas model
10:46:05Bagderyes, makes sense to me
10:48:11Slasherithe reason for db duplicates have been found (small tagcache file corruption near end of the file). But the cause for that is not yet known
10:48:47bertrikSlasheri, at least good to hear you have a handle on it
10:52:27 Join funman [0] (n=fun@rockbox/developer/funman)
10:54:54funmansaratoga: i measure DRAM writes 2 timers faster than IRAM writes
10:55:06funmanreads are the same speed
10:56:57 Join matsl [0] (
10:57:16funmanhm .. read test is fucked
10:59:38kugelfunman: that seems unlikely. how are you messuring, on what target?
10:59:39Slasheribertrik: hehe, i think i found it :) i just forgot to truncate the existing index file before rewriting it (there could be left some junk at the end of the file)
11:01:25funmankugel: i measure on fuze: now with a proper test READ is a bit faster on iram, and WRITE is twice faster on dram
11:02:18bertrikfunman, can you also measure the clip, it has a differnent kind of DRAM IIRC
11:02:24funmankugel: : my clip deadlocks and the c200v2 shuts down
11:03:57funmanalso shows extra pixels on the fuze screen, not sure what happens here ..
11:04:21funmanhm .. i'm dumb
11:04:59funman~0xff => 0xff else it writes anywhere
11:05:49*bertrik is still desperately hacking on the e200v2 radio problem
11:07:13funmanwell the IRAM is faster but it seems to depend on CPU freq
11:08:11 Quit Tristan (Remote closed the connection)
11:09:08funmanClip: Write 0% faster (same speed), Read 36% faster, another test Write 50% faster, Read 60% faster
11:09:35*ocean happy to have get 747+ working :P
11:09:39kugelso iram is always faster?
11:10:08funmannot always: unboosted i get the same write performance on the fuze
11:10:44CIA-70New commit by rasher (r21364): Language updates: ...
11:10:46 Quit Bagder (Read error: 60 (Operation timed out))
11:11:41kugelok, let me rephrase. iram never slower? :)
11:11:51funmanbertrik: doesn't look like ^^
11:12:12funmanbut with the data cache it really looks the same
11:13:51*ocean submits patch to flyspray and goes to stud
11:13:55oceanstudy lol bye
11:14:09oceanthank Bagder_ :)
11:14:16 Quit ocean ("Page closed")
11:14:29linuxstbSlasheri: Some people have noticed that when building very large databases, the speed (i.e. number of files processed per second) drops to 3 or 4 per second as the number of files processed gets large (sorry that's a bit vague) - do you know why?
11:14:38funmanbertrik: no idea about the e200v2 radio ..
11:15:11bertrikfunman, I have dominic's e200v2 radio here, so I can do a lot of tests now
11:15:16 Join FOAD_ [0] (
11:15:35kugellinuxstb: the database progress doesn't tell how many files it processed. I don't know what the numbers are but they're not the files.
11:18:45 Quit PaulJam (Nick collision from services.)
11:18:52 Join PaulJam_ [0] (
11:18:58 Quit FOAD (Read error: 60 (Operation timed out))
11:18:58 Nick FOAD_ is now known as FOAD (
11:21:05 Join Bagder [241] (n=daniel@rockbox/developer/bagder)
11:24:45Slasherilinuxstb: yep, that's because of heavy disk i/o involved and no hash maps used
11:25:00Slasherilinuxstb: we should create hash table for filenames at least
11:25:47Slasherikugel: indeed, that could be chaned. Currently the number is any found file entries
11:26:26linuxstbWhat do you mean by "any found file entries" ?
11:26:45funmanbertrik: should get the old performance, the produced code for the loop is not the same but looks ok
11:26:56funmanbertrik: for c200v1 lcd performance
11:26:58Slasheriyep, the progress indicator is not so useful :)
11:27:08CIA-70New commit by bagder (r21365): respond to COMPLETED
11:27:16bertrikfunman, ok, I'll try that (in a while)
11:27:19CIA-70New commit by amiconn (r21366): Fix a few HID contexts to use proper preconditons (no more pause-before-stop and stop event flooding).
11:27:28funmanbertrik: if you see 50.7 fps you can commit
11:28:22Slasherilinuxstb: just any other files than music files too. I don't remember if directories are counted too
11:29:12 Join p3tur [50] (n=petur@rockbox/developer/petur)
11:30:03CIA-70New commit by zagor (r21367): Added pseudo builds, and COMPLETED.
11:30:21linuxstbSlasheri: So it is "number of files", but that includes _all_ files and maybe directories, rather than just files added to the database?
11:30:44linuxstbThat sounds fair enough.
11:32:20funmanbertrik: do you remember in which OF library block the FM code is located ?
11:32:36 Quit petur (Nick collision from services.)
11:32:42 Nick p3tur is now known as petur (n=petur@rockbox/developer/petur)
11:33:07 Quit Horschti ("Verlassend")
11:33:16bertrikfunman, wav_codec I think
11:33:32 Join kugel_ [0] (
11:33:34 Quit kugel (Nick collision from services.)
11:33:36 Nick kugel_ is now known as kugel (
11:33:39pixelmalinuxstb: it does even say "xx found", not "xx files found" IIRC
11:36:40gevaertslinuxstb: good news! The DAX bootloader still compiles and runs :)
11:38:35 Quit funman ("free(random());")
11:39:50Unhelpfulamiconn: i'm slightly confused about stm/ldm... if i spill some values to stack with stmdb sp!, {r4-r7}, and load them later with ldmia, are the going to be in the same registers? or, more clearly, is one of those, presumable stm, reversing its reglist?
11:40:06Slasherihmm.. anyone reporting duplicates in the db, should state at least the following (so that the report would be useful at all): The player used, is dircache enabled, is db loaded in ram
11:40:36amiconnUnhelpful: If your're using the same register list, values will end up in the original registers
11:40:37 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
11:40:57Unhelpfuland if i'm using some other register list, they'll be in the same *order*?
11:42:14 Join planetbeing [0] (
11:42:43 Quit planetbeing (Remote closed the connection)
11:42:43amiconnNothing is reversed here - both ldm and stm are reading/writing forward. The difference between 'ia' and 'db' is just when the base register gets updated
11:44:32 Quit kugel (Nick collision from services.)
11:44:35 Join markun_ [0] (n=markun@
11:44:38 Join kugel [0] (
11:44:48 Join bertrik_ [0] (
11:45:02 Join perfectdrug [0] (
11:45:05 Join Domonoky1 [0] (
11:46:51Unhelpfulso DB moves the base register down N, before stepping up and reading/writing registers? and IA steps up from the original value to read/write, and then writes the incremented value back?
11:48:14amiconndb == "decrease before", ia == "increase after". The new value is only written back if you specify it using "!")
11:48:57amiconnThe necessary address stepping is presumably done in the memory addressing unit and has no effect on the register
11:49:51 Quit matsl (Read error: 60 (Operation timed out))
11:51:25Unhelpfulhrm, so, if i don't need to update sp, since i won't be doing other stores via it or calling to other functions, i could spill to stack with stmdb, and retrieve with ldmdb, without using writeback?
11:53:29 Quit Domonoky (Read error: 60 (Operation timed out))
11:53:41bertrik_funman, c200v1 is now at 49.7 fps, so only very slightly slower than it used to be
11:54:52 Quit at0m (Read error: 110 (Connection timed out))
11:56:04 Join at0m [0] (
11:59:01CIA-70New commit by zagor (r21368): More debug output.
12:02:00CIA-70New commit by bagder (r21369): better COMPLETED support and now hands out the next build
12:03:00 Nick fxb__ is now known as fxb (
12:03:45 Quit bertrik (Read error: 110 (Connection timed out))
12:04:02 Quit markun (Nick collision from services.)
12:04:06 Nick markun_ is now known as markun (n=markun@rockbox/developer/markun)
12:04:27Unhelpfuland i guess when i read in a whole row of data with ldmia, i can use writeback to update the pointer and save an add.
12:19:32HorschtI am looking at the commits for the last 4 weeks and I notice 21261 and 21262 are missing...
12:19:39Horschtwhat happened to those?
12:21:37Mikachuprobably committed on the 3.3 branch
12:21:43Mikachusvn repos have global revision numbers
12:21:52 Join robin0800 [0] (
12:21:55Mikachucould be /www too
12:23:21Horschti see
12:23:51AlexPHorscht: It was in the 3.3 branch
12:24:12CIA-70New commit by zagor (r21370): Fixed pipe closing.
12:24:28CIA-70New commit by bagder (r21371): changed the logic so that we only hand out builds to clients according to what ...
12:25:03AlexPThe note in pictureflow that it'll stop playback is currently \opt{hwcodec} - does it make more sense to have this on plugin buffer size? (As I'm using that to add the playback starting text)
12:26:34 Join mcuelenaere [0] (
12:27:12mcuelenaereBagder, Zagor: as you're doing the build server remake, could you also add a checkwps target? Last time I checked it was broken again
12:27:16Horschtok, I am asking because I am trying to find what breaks FS #9708. I noticed it first on rev 21318 after updating from 21175 or something that "old". So, I tried to narrow the revision down and I noticed that it seems to work in 21260 but not in 21262.
12:27:47CIA-70New commit by zagor (r21372): Fixed one-off in socket read.
12:27:47Bagdermcuelenaere: sure, but it's not strictly build server related, is it?
12:27:56mcuelenaereBagder: no, not really
12:27:57Horschtso either 21261 or 21262 would break FS #9708, but since they are in the 3.3 branch that confuses me
12:28:20mcuelenaereI just saw it was nowhere mentioned on the BuildServerRemake wiki page (while sims etc are)
12:28:54*amiconn wants an additional register :\\
12:29:45*Domonoky1 hands some registers to amiconn :-)
12:31:53 Join robin0800_ [0] (
12:32:57CIA-70New commit by pixelma (r21373): Finally commit a few improvements to the c200 (last version of FS #8824): In lists - long instead of short Rec to switch to the recording screen for ...
12:35:15 Join AndyIL [0] (i=AndyI@
12:35:30CIA-70New commit by zagor (r21374): Added build countdown.
12:36:09CIA-70New commit by pondlife (r21375): Display speed as relative to real-time, not timestretch amount.
12:39:32CIA-70New commit by alex (r21376): Document the ability to start playback from Pictureflow to the manual.
12:41:39*pixelma takes no responsibility for it though and blames everything on AlexP and Llorean ;)
12:42:34scorche|shpixelma: well, at least you can reach one of them now ;)
12:42:40LloreanI'll gladly get yelled at for a change that makes it make sense. :-P
12:43:01 Join Thundercloud [0] (
12:43:06pixelmaalright, I'll defer any complaints on you then
12:43:17CIA-70New commit by bagder (r21377): bugfixes for zagor
12:44:37***Saving seen data "./dancer.seen"
12:47:00 Quit AndyI (Read error: 110 (Connection timed out))
12:49:07 Quit robin0800 (Read error: 110 (Connection timed out))
12:49:44Mikachuis this the first time we have sound from devcon?
12:49:58LloreanNo, we had sound last time
12:50:00gevaertsMikachu: I heard sound from devcon last year as well
12:50:07Lloreangevaerts: You were there. :-P
12:50:10Mikachui didn't see that one
12:51:13Mikachumore dev than con right now :)
12:52:38CIA-70New commit by bagder (r21378): introduce a better concept of "build round" and get clients to start building ...
12:53:45LloreanMikachu: I think we had it muted except for certain talks
12:54:16CIA-70New commit by mcuelenaere (r21379): * Commit a modified FS #10354 (thanks to Davide Quarta) - add support the VX747+ in scramble and configure ...
12:55:55CIA-70New commit by zagor (r21380): Added multiline packet parsing.
12:56:15Mikachuno "who is who".jpg for the cams yet?
12:57:16 Join pyro_maniac [0] (
12:57:31BagderI tried to comment in picasaweb
12:57:36Bagderbut it failed me bigtime
12:57:49Mikachui figured out you and Zagor are close to the camera with the mic at least :)
12:57:49Bagderso I can't comment :-/
12:57:57Bagderwe are yes
12:58:08Zagorenjoying the conversation? :)
12:58:14Mikachuvery interesting ;)
12:58:54gevaertsMikachu: so what are the remaining issues with the build server rewrite? I have *no* idea :)
12:59:02Mikachusomething with multiple clients!
12:59:09Bagderyay, our first simulated build round with the new build system works
12:59:38AlexPMikachu: On cam one, one the left you can see amiconn and then to his left round the table is pixelma, Domonoky1, then scorche. At the end left to right is Bagder then Zagor. On the other side I'm in the stripy top - to my left is GodEater, right is petur then gevaerts. The others are hidden at the other end of the table, and can be seen on cam 2 :)
13:00:34Mikachuah :) thanks
13:02:54AlexPThe others being (left to right) bertrik_, markun and kugel
13:03:56CIA-70New commit by pixelma (r21381): Forgot to commit the change to keyboard.c to make the line edit mode acctually work on the c200. (belongs to r21373)
13:04:06 Join GodEater_ [0] (n=bibble@rockbox/staff/GodEater)
13:05:48 Nick bertrik_ is now known as bertrik (
13:07:12CIA-70New commit by zagor (r21382): Restart if server dies.
13:08:37mcuelenaereBagder: I have some more targets for the build table ;) VX777 + its BL
13:09:13Bagderthere's a whole range of new ones pending
13:09:40Bagderc200v2 boot/sim/build, Philips SA9200 (normal, bootloader, & sim), 3 Samsung sims, VX777 + its BL
13:10:27mcuelenaereand checkwps ;)
13:10:53Bagderthat takes a busload of fixes by mrsomeone first
13:12:43 Quit mcuelenaere ()
13:13:21 Join krazykit [0] (
13:14:21pyro_maniacBagder: you missed the sim builds for the samsung devices
13:15:01Bagder" 3 Samsung sims"
13:15:19pyro_maniacsorry my false
13:18:03 Quit linuxstb (Read error: 113 (No route to host))
13:20:38Zagorpixelma: what tools to you need to build the manual?
13:28:26 Quit ender` (" I remember being impressed with Ada because you could write an infinite loop without a faked up condition. The idea being t")
13:28:42 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
13:31:42Zagorrasher: yeah thanks. I was looking for an executable that rbclient I can run "which" for to see if it can build the manuals. htlatex seems to fit the bill.
13:33:18*Bagder runs 4 rbclient clients aginst rbmaster
13:38:12Unhelpfulamiconn: i realized something... i probably shouldn't say anything about whether zero-AC checks in the second pass are worthwhile, seeing as the benchmark plugin is always running on the same block :)
13:38:22rasherThe samsung sims don't compile, do they?
13:40:11 Join lee321987 [0] (
13:41:00lee321987On Ubuntu - I'
13:41:49lee321987On Ubuntu - I'm using synaptic to install Latex. Do I need all the "texlive" files?
13:42:18Bagderrasher: I have no idea really...
13:46:02 Join efyx [0] (
13:46:39 Quit GodEater_ (Read error: 104 (Connection reset by peer))
13:47:40rasherOh, they did
13:48:44CIA-70New commit by rasher (r21383): Add some more sims, and fix a bug which would make it also build c200 if you asked for c200v2.
13:50:32 Quit lee321987 ("ChatZilla 0.9.84 [Firefox 3.0.11/2009060215]")
13:52:32CIA-70New commit by bagder (r21384): figure out if we end up with no clients that can build the remaining builds ...
13:54:42 Join Zagor_ [0] (n=bjst@
13:55:03 Quit Zagor (Nick collision from services.)
13:55:06 Nick Zagor_ is now known as Zagor (n=bjst@rockbox/developer/Zagor)
13:55:09Unhelpfuleven better, the test block that i'm processing over and over has *no* zero-AC rows in the second pass.
13:58:55 Quit raps (Read error: 104 (Connection reset by peer))
14:07:18CIA-70New commit by rasher (r21385): Add the Onda V747 and V747+ sim builds as well.
14:08:11 Join __lifeless [0] (n=lifeless@
14:09:11 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
14:09:11 Nick efyx is now known as efyx`AFK (
14:11:57 Join Ubuntuxer [0] (
14:18:14CIA-70New commit by zagor (r21386): Require archlist parameter.
14:20:06pixelmais the auto-update of the database not finding deleted files dependent on - missing dircache / missing option to load it to RAM / both ?
14:20:46Slasheripixelma: nope, shouldn't be currently (it was previously)
14:21:01Slasherinow there is just a high penalty of running it without dircache
14:21:09Slasheri(on disk based players)
14:21:18pixelmaso it would detect deleted files, only slowly?
14:21:27Slasheriit should
14:21:50Slasherii will commit shortly a bunch of fixes into tagcache
14:22:15Slasherithat might not solve the duplicate issue yet, but i will need more detail and bug reports for that
14:22:25CIA-70New commit by bagder (r21387): support a disconnecting client properly
14:22:32pixelmaI'm just looking for the documentation...
14:23:22pixelmaI mean, update the manual a bit, so good to get rid of some misinformation therer
14:23:50Slasherithat would be good
14:24:38Slasheriprobably a next thing to do would be to adjust tagcache to handle the hvsid collection :)
14:27:14 Quit tessarakt ("Client exiting")
14:27:38 Quit _lifeless (Read error: 113 (No route to host))
14:28:46*amiconn concludes that the armv5+ mpeg2 idct won't be possible without reloading constants, and that it only makes sense on armv6
14:29:33 Join Tristan [0] (
14:32:44 Join ZincAlloy [0] (n=d9eec609@gateway/web/cgi-irc/
14:37:42CIA-70New commit by zagor (r21388): Fixed clientname
14:44:41***Saving seen data "./dancer.seen"
14:45:37 Join n1s [0] (n=n1s@rockbox/developer/n1s)
14:45:48 Join matsl [0] (
14:57:46 Join nibbler_ [0] (
15:01:30 Join n00b81 [0] (n=taylor@unaffiliated/n00b81)
15:01:47 Part n00b81 ("Leaving")
15:04:44 Quit ZincAlloy ("CGI:IRC (Ping timeout)")
15:10:31 Join ZincAlloy [0] (n=d9eec609@gateway/web/cgi-irc/
15:10:33 Quit ZincAlloy (Remote closed the connection)
15:10:58 Quit perfectdrug ()
15:14:33 Join fdinel [0] (
15:22:54 Join ocean [0] (i=d59c23f7@gateway/web/freenode/x-12ab51959d32e9f3)
15:29:30*Llorean thinks we need a DevCon bot so that questions can be addressed to "DevCon" and ping someone.
15:30:29 Join Sajber^ [0] (
15:30:59CIA-70New commit by zagor (r21389): Fixed clientname
15:33:15 Join Bunia [0] (
15:36:06 Part Bunia
15:41:37 Join dz [0] (
15:42:02 Quit matsl (Read error: 110 (Connection timed out))
15:44:53oceanthere's a checksum error again with the commit :P
15:47:45Bagderwith the commit?
15:49:05dzanyone care to give me (MattSmith) write permission on the wiki? I've got an entry to add to SansaAccessories
15:50:24 Join ZincAlloy [0] (n=d9eec609@gateway/web/cgi-irc/
15:50:26 Quit ZincAlloy (Client Quit)
15:50:30oceanone of the latest commit in svn
15:50:50 Join ZincAlloy [0] (n=d9eec609@gateway/web/cgi-irc/
15:51:14oceani put the patch on flysprayer and has been modified and committed to svn, but now gives another checksum error
15:51:43CIA-70New commit by bagder (r21390): if another client builds a build when a COMPLETED is received, CANCEL the ...
15:52:03CIA-70New commit by zagor (r21391): Added CANCEL
15:52:06oceanBagder: what was the file you spotted today that i don't remember i think is there the problem cause i remember a "46"
15:52:24oceanand in scramble and configure there 54
15:52:29Bagderit was the firmware/export/config-*.h file
15:52:33ocean(as model id
15:52:40Bagderthe configure number is not used
15:53:17oceanthe config- file seems ok :|
15:54:44Bagderso you updated scramble accordingly?
15:55:00oceanthere's a difference of 10 somewhere, so i think somewhere is defined as 44 vx747 where it should be 54 for vx747+
15:55:15oceanmmmm let me try to recompile scramble maybe i forgot :P
15:55:30Bagderyou should use the new flag for scramble, right?
15:55:38Bagder-add= whatever
15:56:07oceanyes maybe i had an old scramble binary
15:56:36Bagderuh? your build should use the new option so an old scrable would fail...
15:57:38oceanyes i forgot to build it :P
15:57:42oceanafter svn update
15:59:22 Join robin0800 [0] (
16:05:17CIA-70New commit by amiconn (r21392): Faster idct for ARMv6. Overall mpegplayer speedup is quite minimal though.
16:05:46CIA-70New commit by zagor (r21393): Kill 2, fixed confargs
16:09:25 Join moos [0] (i=mustapha@rockbox/staff/moos)
16:09:34oceanok works everything had to recompile bootloader also and forgot it
16:09:56*Llorean demotes himself.
16:10:42Lloreanpreglow: I think you're taking over my spot on the steering board, if the backups are in order.
16:12:42 Join LambdaCalculus37 [0] (n=rmenes@rockbox/staff/LambdaCalculus37)
16:13:35 Quit robin0800_ (Read error: 110 (Connection timed out))
16:14:13scorche|shwell, we will ave discussions in 45 minutes as well
16:17:52 Join ender` [0] (
16:26:39 Quit Ubuntuxer ("Leaving.")
16:28:13 Join robin0800_ [0] (
16:29:23CIA-70New commit by bertrik (r21394): Enable charging on the Sansa Fuze (charging/discharging curves are still uncalibrated)
16:33:59*LambdaCalculus37 decides to start looking into setting up a few plugin keymaps for the SA9200
16:34:29CIA-70New commit by amiconn (r21395): No need to run a command here.
16:35:59CIA-70New commit by bagder (r21396): initial PING support from the master when a client's been silent for 10 secs
16:41:44 Join froggyman [0] (
16:44:45***Saving seen data "./dancer.seen"
16:45:22 Join dmb [0] (n=dmb@unaffiliated/dmb)
16:46:54 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
16:47:39oceanmmm something strange with latest build i've done, update of the database works fine, but opening the "files" menu shows nothing
16:47:53oceani mean shows no files :P
16:48:31linuxstbWhat is your "show files" setting?
16:48:34 Quit n1s (Read error: 60 (Operation timed out))
16:48:38 Quit krazykit (Read error: 113 (No route to host))
16:48:50 Quit robin0800 (Read error: 110 (Connection timed out))
16:50:09oceanit was on playlist :P
16:51:06 Join krazykit [0] (
16:51:10oceanput on "all" still doesn't show FLAC files
16:51:25oceanonly ogg/mp3 and the .rockbox folder
16:52:09oceanmaybe there's something it didn't write them at all
16:52:41oceanbut that could be my linux usb stack that gives strange problems :P
16:53:37oceanmmm ogg files aren't playing :((
16:54:12CIA-70New commit by zagor (r21397): Better handling of server disconnect.
16:55:52oceanok they weren't written, i suspect that there are some kind of errors with sd vx747 driver of rockbox or with linux usb stack :P
16:57:15Bagderor you just pulled the plug
16:57:25Bagderso the error was you :-)
16:58:42oceanmaybe PEBKAC :P
16:59:05oceani should get the volume keys working too high volum :P
17:00:17 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
17:01:01CIA-70New commit by zagor (r21398): Kill all running builds if server connection is dropped.
17:01:11oceanit seems they works but the volume starts way too high
17:01:20oceanand no way to get it down
17:01:45 Join D480 [0] (
17:02:55 Join n1s [0] (n=n1s@rockbox/developer/n1s)
17:03:34 Join mcuelenaere [0] (
17:03:52mcuelenaereocean: that's normal behaviour, the hardware doesn't support volume control
17:04:02mcuelenaereyou'll need to setup a precut in the eq
17:04:15oceanah ok :)
17:04:23oceanthis should be put on the wiki :P
17:04:25oceani think
17:04:44mcuelenaereno it shouldn't, it should get fixed/worked around
17:05:13mcuelenaereand playing some songs can crash the device atm
17:05:17oceanah ok :)
17:05:19Lloreanmcuelenaere: Hardware doesn't support period, or "we don't have support for that in the hardware yet"?
17:05:33mcuelenaereLlorean: Hardware doesn't support it
17:05:43mcuelenaerethe datasheet advises doing SW volume control
17:06:00mcuelenaereit only has some headphone boost stuff
17:06:03Mikachucan't you set the replaygain pregain too?
17:06:06LloreanSo now a decent range for software volume is needed.
17:06:10Mikachumaybe that is more roundabout than eq though
17:06:58mcuelenaereMikachu: I don't know what's best, I'm not really familiar with that part of Rockbox :)
17:07:15LloreanI'd say EQ precut is best for people trying to use it right now
17:07:23dzI wish my FM transmitter had hardware volume control on it, I either have to crank the volume up on my sansa or on the radio
17:07:56dzbut I suspect I'm just being too picky :)
17:08:12mcuelenaerethere's hardware volume control for line in though
17:08:49mcuelenaereocean: do all buttons work correctly on the VX747+?
17:09:48dzanyways, could someone kindly give me write permissions on the wiki so I can add the FM transmitter I got to the Sansa accessories page?
17:10:35mcuelenaeredz: what's your wiki name?
17:10:50oceanmcuelenaere: seems yes
17:11:14 Join robin0800 [0] (
17:11:20oceanalso touch seems to be working good
17:12:02mcuelenaeredz: all done!
17:12:48CIA-70New commit by zagor (r21399): Basic upload
17:13:04*mcuelenaere spots a deleted user in TWikiUsersGroup
17:13:43oceanmcuelenaere: so, now there's something specific to help to get working? or i can try and wander around and get something done :P
17:19:59 Quit tmzt (Connection timed out)
17:21:21mcuelenaereocean: you can look into pictureflow or the PCM/codec issues, but those are pretty hard to debug
17:21:41mcuelenaereyou could also try getting FM working, I did some groundwork there but can't seem to initialise it
17:21:57mcuelenaereor you could look into reversing the NFTL ;)
17:24:15CIA-70New commit by zagor (r21400): Added command check.
17:26:27 Quit robin0800_ (Read error: 110 (Connection timed out))
17:27:42 Quit LambdaCalculus37 ("Fwump")
17:27:53CIA-70New commit by zagor (r21401): Don't overwrite files.
17:31:34 Quit PaulJam_ (Nick collision from services.)
17:31:39 Join PaulJam [0] (
17:34:32 Quit mcuelenaere (Remote closed the connection)
17:35:06 Join mcuelenaere [0] (
17:37:44*GodEater calls the DevCon Meeting to order
17:37:49petur#### meeting start #####
17:39:37amiconnCurrent rsb:
17:40:52 Quit robin0800 (Read error: 110 (Connection timed out))
17:43:08Slasherihuh.. now i have fixed countless tagcache bugs, too many to list those in the commit log shortly :)
17:43:43Slasherii think the db stability should be _much_ improved =)
17:43:59ZagorSlasheri: excellent!
17:45:09Slasheristill testing to find any new issues before committing all
17:45:11 Quit mcuelenaere (Remote closed the connection)
17:47:21Nico_PSlasheri: maybe just reference FS task number for those that have one
17:48:02 Quit PaulJam (Nick collision from services.)
17:48:09 Join PaulJam_ [0] (
17:53:01Lloreanpetur: Maybe we should have "RBUtil Supports", "We provide support" and "Usable but unsupported" categories, of which the first two get release-number versions, and the last one we don't list on the front page (just a link with "Some in-development targets may also already be usable") and provide limited support for
17:53:19*Llorean doesn't know who to ping to send word to devconners
17:53:45Bagderjust write!
17:54:37 Join DarkDefender [0] (
17:56:15LloreanThe Gigabeast is not supported because it's not reliable installation yet, right?
17:56:18 Join toffe82 [0] (
17:56:25AlexPlargely I think
17:56:27LloreanWhereas the e200R *always* works, the S still has issues with single boot
17:56:32*Llorean is saying this for the devconners
17:56:41GodEaterwe gather :)
17:56:47LloreanThey just asked why the R is "supported" but the Beast is not. And I think it's outside the "install" process
17:57:04Bagderthe S is clearly unstable so its not the same
17:57:09GodEateryes, we agree
17:57:10LloreanI think the manual should be complete before something's supported.
17:57:23GodEaterso what about the iAudio M3 then ?
17:57:27LloreanWhereas coding RBUtil requires special skills, the manual is fundamentally text editing. It's just a time investment.
17:57:39LloreanGodEater: That's a "special limitations" exception
17:58:04GodEaterre: manual editing, tell that to AlexP :)
17:58:51Bagderaudio is recorded
17:58:56Bagderfor the record
17:59:20Nico_Pgreat :)
17:59:31LloreanBagder: Maybe two tables on the front page: "Rockbox is release on <list> and is runs on <longer/extra list>" where the release is only ones with a version number release (which indicates a manual) so we accomplish a lot more on the front page *and* quickly letting people know which ones have "gold star" support
17:59:49obosome people come through louder and clearer than others
18:00:38gevaertsLlorean: I think that's mostly what wer're saying here
18:01:05Lloreangevaerts: Ah, it's a *little* hard to hear. It sounded like the idea was to just lower the overall bar, not create two tiers.
18:01:22GodEaterno no, we want two tiers
18:03:26 Quit nibbler_ (Read error: 110 (Connection timed out))
18:04:20LloreanThe difference between the two tiers is that the top tier should *in most cases* only have software issues (any time someone has a problem you don't need that specialist support you're asking for) whereas the other tier is one where you get "limited" support in the sense that there's a more limited number of people who *can* support you, rather than that we intentionally limit your support
18:05:32Lloreangevaerts has my intent with what the "Expert" badge was meant to mean. The name was something i picked after asking around for a while for suggestions, I don't think it's ideal
18:06:48GodEaterthanks Llorean :)
18:07:21LloreanI think the basic checklist should be "can you install it without a high chance of bricking, does it play music without freezing, and can you adjust all the playback related settings"
18:07:39LloreanWho cares about plugins, secondary features (even radio and recording count as these to me, they're nice, but not a bar for support)
18:08:20LloreanIt's "supported" when people can use it for daily use for music listening, for a lower bar, I think.
18:08:30GodEaterI agree :D
18:08:33Bagderme too
18:08:41*Domonoky1 agrees too
18:08:49*ocean too
18:09:45 Quit ocean ("Page closed")
18:10:00Lloreangevaerts: If you can't load plugins, you *probably* can't load codecs either.
18:10:12LloreanNot to mention, isn't the playlist editor a plugin (which is a key part of daily playback)?
18:10:52ZagorI don't consider playlist editor a critical part of daily playback
18:11:02Zagor"critical" as in showstopper
18:11:39LloreanZagor: It's a little weird that you can add things to playlists, but not remove them from them, then.
18:11:53LloreanIt's not obviously a plugin, and it's available from "core" menus, so users won't understand that technicality, I think
18:12:05Zagorit is certainly an undesired situation, but it's still release-worthy imho
18:12:26LloreanI imagine if you've got playback working, getting the plugin editor working is just a keymap issue though.
18:13:36LloreanMaybe half as often, but a 2-week freeze + 1 week branch
18:13:45LloreanTimestretch is after 3.3, it's not in 3.3
18:14:08GodEateryes, I've been thoroughly corrected there
18:14:21LloreanI think the really short freeze is also bad because people will just take a break for a week. It's too short to notice.
18:14:31LloreanA slightly longer freeze, half as often, but still strict limits, might help
18:15:04 Quit bmbl (Read error: 104 (Connection reset by peer))
18:17:31LloreanMaybe make the RC period longer then?
18:17:57CIA-70New commit by miipekk (r21402): A bunch of stability fixes into tagcache engine and database browser. Mainly data retrieval problems, races, data corruption of sorted index files at ...
18:19:15 Quit n1s (Read error: 104 (Connection reset by peer))
18:19:25obocan always put out a point release if something really horrible is discovered
18:19:29LloreanWe've gotten feedback on RC branches in the forums before, when someone's made build available.
18:19:39 Quit D480 ("Ex-Chat")
18:21:23LloreanMaybe we need a release checklist.
18:21:33oboI thought there was one?
18:21:36Llorean"Does X target install, boot, play each format, run each plugin"
18:21:40Bagderwe have one
18:21:40LloreanA more fomulaic one.
18:21:46LloreanFor users to respond to in the forum
18:22:03LloreanSo we can put up the checklist, and get users involved in using the build so that they can be the one "who filled out the checklist for H10 3.3"
18:22:53LloreanInstead of just asking them "please use this RC" we should ask them "Try these things".
18:23:10GodEaternice idea, I like that
18:25:49LloreanWe haven't failed, certainly. We could probably do better with minimal work on our part (once the system is in place) to make sure it's harder for things to fall in the cracks.
18:26:55ZagorLlorean: I agree
18:27:50LloreanThere might be a benefit to making the RC period longer, just to provide more time for the feedback to come back. But that's sorta a neutral issue either way.
18:28:17GodEatermeanwhile, off camera, there is frantic lock picking action!
18:28:20gevaertswe can do that after improving the RC process first
18:28:29oboNightly RC builds? Or automated builds when the branch is changed?
18:28:38Lloreangevaerts: Yes, certainly. See how much feedback we get in a week, and then decide if we think more time would help
18:28:49Zagorobo: the latter
18:28:54*Llorean agrees
18:28:55gevaertsobo: mostly guding reporting a bit, like Llorean said
18:30:28LloreanWe could also close more "old" tasks that haven't received comments in a while
18:30:35LloreanTheres' a nice "Out of date" closing reason
18:33:36 Join n1s [0] (n=n1s@rockbox/developer/n1s)
18:34:08LloreanThere's really no reason someone can't open a new task if someone's going to try to get it into condition for commit again.
18:36:53LloreanI think a good standard for plugins is that they can go in when they "work" right, have keymaps for all targets, and someone is actively working on them and is willing to fix the code style (but hasn't yet)
18:37:09LloreanIf the code is in SVN it might inspire them to work on code style, especially if they've seen people often get SVN access to maintain their plugins.
18:37:41LloreanAnd then plugins should be disabled for Releases until they're in the manual for all targets.
18:38:01 Quit jordan` (Read error: 60 (Operation timed out))
18:40:03 Join robin0800 [0] (
18:40:05LloreanAfter the release seems good
18:40:35LloreanWe should be committing as many new features as we can right now, so there's time for shakedown before the next freeze.
18:40:39pixelmalike Superdom?
18:40:55LloreanSuperdom is kinda broken anyway, isn't it?
18:40:56pixelma(about undocumented plugins)
18:42:04pixelmanot sure, I always thought I can't control it correctly and also saw problems with a userfont in its menu - but haven't looked for a long while
18:42:26LloreanLast I heard it was also actually broken in the gameplay
18:44:46***Saving seen data "./dancer.seen"
18:45:04LloreanWhat's the question?
18:45:18LloreanSound is dropping in and out again
18:45:27obosomeone uploading?
18:45:28peturLlorean: next point on the agenda
18:46:01GodEaterRockbox logo / Lyre project
18:46:03GodEaterto be clear
18:46:08LloreanDidn't even notice the agenda
18:46:18GodEaterhow much we should charge them to license our logo :)
18:46:20LloreanI think use of the Rockbox logo is fine if it's like those "Intel Inside" logos and such
18:46:40GodEaterpretty much what we've said
18:47:22LloreanYeah, an official line is very hard to draw
18:47:43peturis the sound better?
18:48:09LloreanA good way would just be a recommendation. "We ask that you only use the Rockbox logo to inform people that our software runs on your product, and try to ensure it's clear that the Rockbox project is an independent project not associated with any specific hardware or software group."
18:48:13LloreanYeah, the sound is better now
18:48:39Bagderyes, a recommendation like that i fine
18:49:00CIA-70New commit by miipekk (r21403): Keep simulators happy.
18:49:31LloreanThere have been a few times where we've had real spammers in here and nobody around.
18:49:53LloreanI don't think we should make all committers ops. But I do think all committers can be trusted with it, we should just ask which of them are willing to volunteer to accept the responsibility
18:50:00gevaertsSlasheri: the bugs seem to have taken quite a bit of RAM ;)
18:50:43 Quit Thundercloud (Remote closed the connection)
18:51:36LloreanI don't think it's likely that anyone will ban people for most of the guidelines.
18:51:46Slasherigevaerts: oh, that's interesting, i thought ram consumption would have been increased :) at least amiconn is now happy :)
18:52:09LloreanI mean, you don't ever ban people for idle off-topic. you warn them and remind them. It's not until people are *willingly* violating things that an op-like actions should even be a question.
18:52:38Slasheriin fact it has increased but not declaring two bit structures as static has freed it back
18:52:55LloreanIf consistency becomes a problem, we can write clearer guidelines for when a ban is allowed, but we can start by trusting peoples' instincts.
18:54:24AlexPLlorean: agreed
18:56:11LloreanModeration should always be reserved for people *willingly* misbehaving
18:56:40LloreanBeing stupid / annoying / unable to get the point is different from getting the point and still doing it.
18:57:07 Nick efyx`AFK is now known as efyx (
18:59:57 Join ocean [0] (i=d59c23f7@gateway/web/freenode/x-f83d6da59640d5d4)
19:01:57 Quit kadoban (Remote closed the connection)
19:02:11 Quit dmb (Read error: 54 (Connection reset by peer))
19:02:26LloreanWhen I was running the forums, the "bar" for Expert (in my mind) was always approximately "Do they seem to research before answering, do they try to diagnose a problem before providing an answer instead of trying several answers and hoping one works, and do they seem to show an interest in helping people with players beyond simply those they own"
19:02:38AlexPLlorean: I think that works well
19:03:32LloreanIf someone had a history of jumping to conclusions on answers etc, I was very hesitant about them in the future (even when others said they'd improved) though. Probably too hesitant
19:03:51Galoisnot literally, I take it - otherwise someone who owns every target would be unable to help people with players they don't own
19:04:28LloreanGalois: "was always approximately" :-P
19:04:57LloreanWe need to decide what we want from a new website first, then
19:05:06LloreanFor example, it's pretty clear we want a user portal and a dev portal
19:05:39LloreanThe user portal needs to quickly tell a user "We're pretty active, and it makes your MP3 player awesome if your player is on this giant list"
19:06:28LloreanWe can drop *most* of the menus from the user portal. Instead just use the category headers, like "Documentation" can bring them to a friendlier documentation portal.
19:06:39LloreanBasically, split the main page into a bunch of friendly, simpler pages.
19:07:26AlexPI agree
19:09:04LloreanI'm still firmly in the "we need a list of updates" on the front page but it could be, for example, the last 10 MajorChanges and a note that says "Last minor change or bug fix on <date+time of last commit>"
19:09:52LloreanThat can be solved with a "Player Identification" page
19:10:05 Quit Nico_P (Remote closed the connection)
19:10:06LloreanWe have a list on the front page, and an "Aren't Sure? More info here..." link
19:10:27LloreanThat can have images for all the players we have images for, and descriptions for other unique things people can use for identifiers (firmware version, text color, etc)
19:12:06 Join saratoga_home [0] (n=41becb3b@gateway/web/cgi-irc/
19:12:38 Join Thundercloud [0] (
19:16:08 Join CPrompt^ [0] (
19:16:16 Nick zitune is now known as zitune[afk] (n=zitune@bearstech/zitune)
19:16:43 Nick J-23 is now known as noluv4mgorny (
19:16:48 Nick noluv4mgorny is now known as J-23 (
19:20:31 Quit fyrestorm (Read error: 104 (Connection reset by peer))
19:23:13 Join robin0800_ [0] (
19:23:17 Quit robin0800 (Read error: 104 (Connection reset by peer))
19:27:00 Quit ocean ("Page closed")
19:30:20 Join homielowe [0] (n=42b74819@gateway/web/cgi-irc/
19:31:08CIA-70New commit by rasher (r21404): More translations for arabic by Raafat Akkad (FS #10355).
19:31:14 Join fyrestorm [0] (
19:35:49*Llorean can't think of any new business.
19:37:23Bagdermeeting is over for now
19:37:25Domonoky1official meeting time ended now.
19:41:41 Join tmzt [0] (n=tmzt@
19:44:13 Join ocean [0] (i=d59c23e8@gateway/web/freenode/x-08a2d21539e7e08d)
19:54:28 Nick fxb is now known as fxb__ (
19:54:28 Quit homielowe ("CGI:IRC (EOF)")
19:56:36 Join pyro_maniac1 [0] (
19:57:04 Part pyro_maniac1
19:59:12 Join Hillshum [0] (n=chatzill@unaffiliated/hillshum)
20:05:36 Quit Zarggg ()
20:05:54 Quit ZincAlloy ("CGI:IRC (Ping timeout)")
20:09:17 Join tessarakt [0] (
20:12:37CIA-70New commit by amiconn (r21405): Plugin for testing LCD driver and greylib drawing performance.
20:14:28 Quit davidfg4 ()
20:14:43CIA-70New commit by bagder (r21406): don't die on unknown protocol keywords, better deal with (weird) clients ...
20:14:51 Quit pyro_maniac (Read error: 113 (No route to host))
20:15:48 Quit n1s (Read error: 60 (Operation timed out))
20:19:58 Join nibbler_ [0] (
20:23:42CIA-70New commit by zagor (r21407): Proper http codes.
20:24:27 Join Zarggg [0] (
20:24:33 Join engelzz [0] (
20:24:39 Nick engelzz is now known as Russel (
20:25:31HillshumRussel1: Hi
20:25:40Russelif i have something like playcount in my tagnavi_custom.config, do i need to enable something for this to work?
20:28:21oceanPANIC TLB Exception 0x80053F00 (stack at 0) [0x0000004C] x8000190C lol!
20:28:42LloreanRussel: By 'something like playcount' do you mean 'playcount' or something else?
20:29:10LloreanGather Runtime Data
20:29:55oceanoccurs with metronome - play
20:30:23CIA-70New commit by bagder (r21408): modified receive path
20:30:57 Join n1s [0] (n=n1s@rockbox/developer/n1s)
20:31:02 Part Russel
20:34:41 Join ZincAlloy [0] (n=d9eec609@gateway/web/cgi-irc/
20:37:31CIA-70New commit by amiconn (r21409): Make ATA poweroff an advanced build option for the Player.
20:37:56 Join mcuelenaere [0] (
20:38:10mcuelenaereocean: yeah, that's part of the 'PCM issue'
20:38:21oceanah ok :)
20:38:42mcuelenaeretry building with #define EXTENDED_EXCEPTION_DESC 1 in firmware/target/mips/ingenic_jz47xx/system-jz4740.c, you'll get more info
20:39:20oceanok thanks, strange thing is that it doesn't only panic but the video gets fucked up
20:39:37oceanlike it "shifts"
20:41:09*amiconn is cleaning up his tree a bit
20:41:54 Join webguest03 [0] (n=29e89f6c@gateway/web/cgi-irc/
20:42:18mcuelenaereocean: you'll need to reset
20:42:29mcuelenaereget some tooth picks or something sharp, you'll need it ;)
20:43:33oceanyeah i know already done it two times :P it's getting funnie to reset :P
20:43:56oceani always use the reader touch screen pen :P
20:44:47***Saving seen data "./dancer.seen"
20:45:03Unhelpfulamiconn: hrm, sleep(0) should be as good for sync-to-tick as the busy-wait that i used, then?
20:45:29oceani see a bunch of warnings when compiling firmware in fat.c maybe i can clean it a little :)
20:46:27mcuelenaerehuh? what compiler? (x86?)
20:46:41amiconnActually busy wait might be a little more precise, as it doesn't yield
20:46:41mcuelenaerewarnings should show up as yellow in the build table
20:48:07oceanconfigure -> vx747p -> bootloader -> make
20:48:14mcuelenaerehmm did you modify your Makefile?
20:48:38oceanno i don't think i have modified nothing
20:48:44oceanjust done svn up
20:49:31oceandone a make clean before though
20:49:44mcuelenaeretools/configure should've added -Wno-pointer-sign
20:50:06mcuelenaerecould you check your Makefile?
20:50:35oceanremoved the build directory and works fine
20:50:51oceanbut i'm sure i didn't modify it
20:51:13oceanstrange maybe my fault
20:51:33 Quit Bagder (Success)
20:51:34oceanor something like make reconf / make clean
20:52:19 Join Bagder [241] (n=daniel@rockbox/developer/bagder)
20:52:32oceanremoving the dir solved everything :P
20:54:13Unhelpfulbetter still would be to use platform counters where possible, and suspend interrupts... beast has cycle counters, does anything else support some sort of timer or counter without interrupts?
20:54:53CIA-70New commit by zagor (r21410): Do build, upload log and zip.
20:55:03saratoga_homeis there a download link somewhere for the complete set of test_codec files?
20:55:32LloreanI don't think we have one yet.
20:55:46LloreanYou're expected to generate them from the source file, which is the opposite of ideal I think. :)
20:56:20saratoga_homei think it only generates some of the formats on the page, and even then its tricky to generate some formats like wma for some people
20:56:47*mcuelenaere would a like a download link for that set too
20:56:53saratoga_homei wonder how big the whole set would be
20:57:11Unhelpfuland some users might end up with different files :/
20:57:34Hillshumsound works better now
20:59:00LloreanUnhelpful: Yeah, it would definitely be ideal to have a set of exact files.
21:01:02Domonoky1there is some problem with the rbspeex Makefile. It wants to put its intermediate files into a rbspeex/build directory, but it doesnt exist and it isnt automatically created. :-/
21:03:04*bertrik thinks he almost fixed the e200v2 radio problem
21:03:42saratoga_homea zip with 96, 128, 256, 320 for mp3, aac, vorbis, wma, mpc would be pretty reasonable in size
21:04:04Bagderbertrik: go go go!
21:04:04saratoga_homemaybe add in AAC-HE and speeks too
21:04:16saratoga_homeand drop 256k for 192k
21:04:21saratoga_home100MB or so
21:09:02 Quit flydutch ("/* empty */")
21:09:06oceanmcuelenaere: with extended excp. desc define it simply reboot!
21:09:18mcuelenaerehmm that's odd
21:09:59mcuelenaereocean: try replacing the system_exception_wait(); call with while(1); at l.279
21:10:01oceanshows two lines for a second without rockbox logo and as always screwed up, and reboots after one second
21:11:53 Join _lifeless [0] (n=lifeless@
21:11:55oceanmcuelenaere: is there a way to make it ouput everything to a file?
21:12:07mcuelenaereocean: probably, but you'll need to code that yourself ;)
21:12:26mcuelenaerekeep in mind you're running in 'interrupt' mode at that point though
21:12:37mcuelenaereso you can't do much threading stuff without screwing up
21:13:29oceanmm ok
21:15:03 Join mc2739 [0] (
21:15:15 Quit saratoga_home ("CGI:IRC (EOF)")
21:17:09CIA-70New commit by alex (r21411): Prepare the manual to have two or three column button map tables so that remote keymaps can be added. ...
21:18:02 Quit `VL (Read error: 110 (Connection timed out))
21:18:28oceanmcuelenaere: not much only two lines not much readable
21:19:07oceanTLB Exception (Store) 0x000000004C at 0x80053fe4
21:19:33mcuelenaerethen the code probably gets overwritten by something, which only causes you to see the first lines
21:20:03mcuelenaerenormally you should see a register dump
21:20:12oceaneverything's black but these two lines
21:20:15CIA-70New commit by alex (r21412): The define is remote_keymap - good place for a typo that.
21:20:18 Quit Hillshum ("ChatZilla 0.9.83 [Firefox 3.0.3/2008092417]")
21:21:57oceanthe two lines are at the bottom of the screen
21:21:59 Quit ZincAlloy ("CGI:IRC (Ping timeout)")
21:23:40oceanmaybe there's something wrong with lcd_puts, i think it goes off the screen
21:25:24oceangraphics gets weird like a shifted down of a lot of lines and a little shifted at right
21:26:01 Join jordoex [0] (
21:26:44 Join matsl [0] (
21:29:12 Quit __lifeless (Read error: 110 (Connection timed out))
21:30:11oceanoh and also after it aumoatically reboots OF only shows the graphics but don't work at all
21:30:34 Quit Thundercloud (Remote closed the connection)
21:31:00mcuelenaereocean: that's really odd :)
21:31:12oceanmcuelenaere: you're the expert here :P
21:31:15mcuelenaereactually, I have to admit; I never tried using the OF after Rockbox rebooted it :)
21:31:19CPrompt^every r2XXXX release that I have put on my iPod Video, the screen never says "Do Not Disconnect" when plugged up via USB
21:31:24oceanno no i don't use it
21:31:53oceanafter metronome panic with the define it automatically reboots and start with the OF that is like "locked up"
21:31:53CPrompt^what happened to that? I also never get the check mark when I eject it
21:32:03LloreanCPrompt^: Those were apple screens, and have nothing to do with Rockbox.
21:32:18CPrompt^Llorean: really? ah, I didn't know that
21:32:25mcuelenaereocean: it doesn't respond to touchscreen presses nor physical buttons?
21:32:28LloreanRockbox's USB mode looks different. Big USB plug.
21:32:40mcuelenaereocean: then this isn't a 'proper' reset
21:32:43oceanmcuelenaere: yes right only reset button works :P
21:32:51CPrompt^Llorean: well, even that, I eject it and it still shows the big USB plug
21:33:39CPrompt^Llorean: it doesn't eject
21:33:48LloreanDoes your computer say it doesn't eject?
21:34:02mcuelenaereocean: weird, never encountered that before
21:34:06mcuelenaereocean: is it reproducable?
21:34:22oceanevery time :P
21:34:41CPrompt^yes. fdisk doesn't show it
21:34:53CPrompt^ actually
21:35:02LloreanSo your computer says it's ejected.
21:35:11 Quit robin0800_ (Remote closed the connection)
21:35:12LloreanThat means it's ejected. I'm not really sure why you're saying it doesn't eject then
21:35:17oceancompiled with the #define EXTENDED_EXCEPTION_DESC 1, boot, metronome -> play -> panic -> automatic "reboot"
21:35:27 Join robin0800_ [0] (
21:35:45LloreanCPrompt^: Are you having some sort of specific problem?
21:35:46CPrompt^Llorean: should the screen on the iPod not change?
21:35:51LloreanNo, it should not.
21:36:16CPrompt^well, then I guess it's work at it should
21:36:29CPrompt^i was just used to the screen changing
21:37:45oceanmcuelenaere: and instead of writing the strings on the first line it draws them on the bottom, that's why they aren't visible, and i suspect the reboot is because it goes to write to some memory zone wich shouldn't be accessed
21:37:55ocean...buffer overflow :P
21:38:10LloreanCPrompt^: If your computer says it's ejected, then it's ejected and safe to remove. A major point of ejecting is to make sure your computer is no longer accessing it.
21:38:54CPrompt^Llorean: as long as fdisk doesn't show it, it's not being written to. So, I'm good. I was just a bit confused
21:39:01mcuelenaereocean: that part is perhaps the DMA doing RAM->audio transfer getting messed up
21:39:56grimmanAnd now for something completely different:
21:40:23oceanmcuelenaere: but how does that affect "video ram"?
21:41:10mcuelenaereocean: well I'm just guessing here, that's the most unstable part used at that moment (and it also causes your TLB exception)
21:42:56oceani'll do a debug build and see if it changes something
21:43:05oceannow i was doing a "Normal" build
21:43:15mcuelenaerewhat do you mean with a 'debug' build?
21:43:54oceanwhen configure -> Advanced -> Debug
21:44:07oceanPlatform set to ondavx747p Build (N)ormal, (A)dvanced, (S)imulator, (B)ootloader, (M)anual: (Defaults to N) A Advanced build selected Enter your developer options (press enter when done) (D)EBUG, (L)ogf, (S)imulator, (P)rofiling, (V)oice
21:44:27mcuelenaereI don't think that changes much, AFAIK it only adds -g to the compiler arguments and sets some threading checks in Rockbox
21:45:10oceanmcuelenaere: if some memory location gets overwritten setting the -g gives less crahes (in my experience with C on x86 architecture)
21:45:29mcuelenaerex86 perhaps, with a full kernel running
21:45:41mcuelenaerebut these are embedded devices, and this is very low-level
21:45:43oceanmaybe i don't know :P
21:46:16Unhelpfulocean: that might make sense if it ends up writing debug data instead of code etc. i think we strip debug data from target binaries, though.
21:46:41oceanyes that make sense
21:47:12 Join prona [0] (
21:48:26*ocean is hungry and have to study
21:49:58 Quit nibbler_ (Read error: 110 (Connection timed out))
21:52:12oceani think debug build have some problems :P
21:54:02 Join fml [0] (n=4fd3ff08@gateway/web/cgi-irc/
21:54:02 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
21:55:31fmlShouldn't the last change to the manual (three column key map tables for DAPs with a remote) break the manual? But it builds just fine for H120. Hence I assume something is wrong.
22:06:39 Part CPrompt^
22:11:38 Join faemir [0] (n=faemir@
22:16:56 Join tvelocity [0] (
22:17:43 Quit mcuelenaere ()
22:18:37 Join Ubuntuxer [0] (
22:27:21 Quit Llorean ("Leaving.")
22:27:23rasherIs anyone against me removing the marquee theme from SVN, as it's the last one we don't have permissions to relicense to CC-BY-SA 3.0. If no one speaks up, I'll remove it and add a COPYING+AUTHORS file to the theme dir. At that point, we're just about ready to remove the themes and put them on the theme site instead. Not quite sure how that'd work out though - might need more theme site work
22:28:25 Join Llorean [0] (n=DarkkOne@
22:34:18 Join bluebrother [0] (n=dom@rockbox/developer/bluebrother)
22:36:22 Join saratoga_home [0] (n=41becb3b@gateway/web/cgi-irc/
22:36:39saratoga_homeocean: you probably shouldn't open FS entries for in progress ports
22:37:46pixelmarasher: I already wondered how (good) the theme site copes with themes named exactly the same but for different targets
22:38:38pixelmafml: the plan was to not break the manuals and give us the chance to add the remote buttons bit by bit
22:38:52rasherpixelma: It hates them with a passion. Or rather, it rejects them.
22:39:15 Quit petur (Remote closed the connection)
22:41:33pixelmaand speaking of which... what's with the H100/H300-themes-with-RWPS-problem? Some SVN themes include RWPSs as well...
22:42:02rasherAh yes.. I think I have a simple fix for that.
22:42:21fmlpixelma: but how could that be possible? You either have three columns (and then the manual should not build as it is now) or you have wo columns (but then there is no column for the remote keys)
22:42:56 Join petur [50] (n=petur@rockbox/developer/petur)
22:42:57pixelmayou can enable the third column when you need it
22:43:34 Quit bmbl ("Woah!")
22:43:37 Quit moos ("Rockbox rules the DAP world")
22:44:28 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
22:44:46fmlpixelma: I don't quite understand how it's supposed to work but it apparently works!
22:44:51***Saving seen data "./dancer.seen"
22:47:36oceansaratoga_home: sorry, i will pay more attention
22:49:30 Quit Ubuntuxer ("Leaving.")
22:51:45rasherpixelma: Should be fixed now. I simply made it not check rwpses (since they got listed on the wrong page because of the checkwps result for the rwps)
22:53:59AlexPfml: You need to add a define to the platform file to make it use the three column version
22:54:09AlexPfml: Otherwise it uses the two column version as before
22:54:42AlexPfml: If you don't add all the extra buttons for a target to the manual in one go it will break the manual
22:54:53AlexP(and at the same time add the define)
22:55:06fmlkugel: hello. By "not breaking config" do you mean that the config values are saved by text values (e.g. "off") rather than as numbers? Then indeed putting it at position 0 would be a more natural option.
22:55:35kugelyes, I think it works like that
22:55:57fmlAlexP: hey! But aren't all the defines and options defined in the .h files? That's what I thought and hence asked.
22:56:00kugelyou need to check the string anyway, since it could possibly contain invalid values
22:56:58matslBagder: Just tuned in to the live feed. You all seem more concentrated today. Got some tricky things going?
22:57:15pixelmafml: no, for the manual there are defines in the platform files - and additionally ones are parsed from the apps' features.txt (used for the language files as well)
22:57:26Bagderme and zagor continue to work on the distributed build server remake
22:57:51Bagderwe've done simulated builds fine, and we're now slowly getting the clients to do actual builds and uploads
22:57:53 Quit saratoga_home ("CGI:IRC (EOF)")
22:58:05fmlkugel: So it's stored as text values? Which are specified via ID2P in settings_list?
22:58:39fmlpixelma: ah, now I see. I thought everything come from features.txt only
22:59:31pixelmafml: but this new option is currently one you add manually to the specific platform file. Once all remote targets' button tables are filled out we might be able to use something from features.txt
23:04:13fmlpixelma: the problem with the current approach (IMO) is that you have to do it in one shot. Couldn't it be done in a way that if you don't specify a remote key then some default values is inserted?
23:05:18AlexPfml: Feel free to improve
23:05:29pixelmaif you are a tex magician and can make that work...
23:06:25pixelmait's already nice that you only have to do it for one target in one go not all at once
23:09:30fmlAlexP: if I knew how... :-)
23:09:39AlexPfml: Exactly :)
23:10:03 Join Thundercloud [0] (
23:10:56CIA-70New commit by bagder (r21413): specify −−type=n for "normal" builds
23:12:29 Quit n1s (Read error: 104 (Connection reset by peer))
23:12:43 Join n1s [0] (n=n1s@rockbox/developer/n1s)
23:13:13bluebrotherAlexP: why are you using \newcommand{}? You have an environment in that, why don't make it a new environment by itself?
23:13:23 Join jkl [0] (
23:13:27CIA-70New commit by alle (r21414): Merge the "Replaygain Off" option into the replaygain type; eliminate the "On/Off" setting (FS #10356). All WPS tags should work as before.
23:13:39AlexPbluebrother: As I didn't know better?
23:14:27AlexPbluebrother: I started off making a new environement (e.g. btnremotemap) to do it, but \end{} inside an opt doesn't work
23:14:33bluebrotherI'm a bit out of the loop with LaTeX, but what's the need for this \beginbuttonmap?
23:15:16AlexPbluebrother: One mo
23:15:38 Join mc2739_ [0] (
23:15:52 Quit mc2739 (Nick collision from services.)
23:15:53 Nick mc2739_ is now known as mc2739 (
23:16:27bluebrotherhmm, as far as I can tell you are defining the btnmap environment depending on the remove_keymap. I don't see why there is something special here −− it's just a different environment
23:16:47AlexPbluebrother: It saves having to have \opt{remote_keymap}{\begin{btnmap}{}{}} \nopt{remote_keymap}{\begin{btnmap}{}{}{}}
23:16:58AlexPAt the start of every btn map table
23:17:21bluebrotherwell, I wouldn't do that.
23:17:26AlexPIt replaces the \begin{btnmap}{}{} and afds the correct amount of {}{} to it
23:17:37bluebrotherfor one, why are there only 2 options in the remote keymap table?
23:17:50AlexPWhich remotekeymap table?
23:18:30bluebrotherthe btnmap when remote_keymap is defined. Why two captions? Are you sure this works at all?
23:18:30 Join Ubuntuxer [0] (
23:18:37AlexPIt works
23:18:52AlexPbluebrother: As we have targets with and without remotes
23:19:01bluebrother(I've never tried adding two captions, the situations I needed something like that was only with the subfigure package)
23:19:32bluebrotherbut what is this caption for? It's not the heading.
23:19:33AlexPAh, that bit - sorry misunderstood
23:19:40AlexPIt is the heading of the column
23:19:52AlexPBut that may be wrong
23:19:57bluebrothercan't be.
23:20:08AlexPcan't be what?
23:20:22AlexPYou get a table with three columns, each with a heading
23:20:29bluebrother\caption{} is the caption of the environment. Usually it's typeset below the element, and something like "Table 23: foo"
23:20:53AlexPSo should it be label?
23:21:14bluebrotherthe heading is \btnhead in our case.
23:21:20bluebrotherno, label is the (internal) label for cross-references
23:21:36AlexPSo what should that be instead?
23:21:47kugelfml: yes
23:22:08bluebrotherwhich means that both tables are different only in the number of columns (i.e. the 2nd {} on \tabularx) and the heading, i.e. the \btnhead macro
23:22:15kugelnothing with id2p though, the settings values aren't localised
23:22:34bluebrotherwell, both environments should have two arguments: caption and label. AFAIK both are currently unused.
23:22:35AlexPThe two versions of btnmap are different only in twerms of columns, yes
23:23:19bluebrotherso you don't need a different number of arguments (the number of columns isn't defined by number of arguments ;-)
23:23:29fmlkugel: I left it as it is because I'm not sure how I would have to change WPS tag handling if I placed the OFF option in the beginning
23:23:41bluebrotherplus you don't need those \beginbuttonmap macros.
23:23:42AlexPbluebrother: To \begin{btnmap} ?
23:23:47kugelit's fine :)
23:24:04AlexPYes, if I don't need different numbers of {} then I don't need the macro :)
23:24:04bluebrotherputting an environment start in a macro is something I remember to continuously make problems :)
23:24:13AlexPWhich makes it easier :)
23:24:21bluebrotheryou definitely don't need it.
23:24:24fmlkugel: now to adjust the manual and deprecate one string in the lang
23:24:44AlexPbluebrother: OK, cool - I'll get rid of that
23:25:40bluebrotheryou might even be able to use a macro for the column layout (i.e. the 2nd {} in the \tabularx), so you might be able to go away with simply two macros defined differently, leaving the btnmap macro completely untouched
23:25:55bluebrotherthough I have no idea if tabularx likes that. Never tried.
23:26:46AlexPI'll get rid of the existing macro first, then have a play
23:28:42CIA-70New commit by alle (r21415): Deprecate the string made unnecessary in r21414
23:34:30CIA-70New commit by bertrik (r21416): Fix e200v2 radio problem (missing Si4702 initialisation)
23:35:59CIA-70New commit by bagder (r21417): cleanup the round properly when the build round is done
23:36:03BagderTOWER TIME!
23:36:12*Bagder positions cam-2
23:37:00CIA-70New commit by zagor (r21418): Build and upload zip
23:37:15 Join __lifeless [0] (n=lifeless@
23:38:44 Quit webguest03 ("CGI:IRC (Ping timeout)")
23:39:05CIA-70New commit by alle (r21419): Update manual (include changes from r21414)
23:39:32Bagderthat's a fine set of tower material
23:39:40Mikachuare you going to play jenga?
23:39:53Bagderlast year's tower is hard to beat
23:42:10ender`how do you unmake that tower without having half of the players fall down? :)
23:42:25CIA-70New commit by alle (r21420): Remove the setting from the list as well
23:43:13 Join Cory` [0] (
23:44:10 Quit Ac-town (Remote closed the connection)
23:44:30pixelmabluebrother, AlexP: for the manuals there are also still buttons mentioned in some textual descriptions, mostly the core \ActionSomething. I first thought of putting both keys into the \Action but that will break if that action macro is used in button tables (would be quickest though). The only thing I can currently come up with is have both \ActionSomething and \ActionRemoteSomething and then add the latter to the text inside an \opt{remote}...
23:44:31bluebrotherender`: it worked fine last year :)
23:44:57ender`will it still work if there's more players? :)
23:45:00pixelma...downside of it is though that you have to every text you need it and change it there.
23:45:09CIA-70New commit by bertrik (r21421): Fix delays in generic i2c driver, reduce delays in ams sansa fmradio i2c driver (tested on m200v4, clip, e200v2, fuze).
23:51:21CIA-70New commit by Ubuntuxer (r21422): Remove viewport menu from star and remove useless button macros in code.(FS #10281)
23:53:19 Quit _lifeless (Read error: 113 (No route to host))
23:53:33bertriktoday is a good commit day

Previous day | Next day