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

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

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

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2010-05-30

00:01:48funmanwodz: fuze would do it ? 220x176
00:02:24***Saving seen data "./dancer.seen"
00:05:57*jae wishes he still had his Fuze... but he's happy seeing that v2 is coming along nicely (right?)
00:06:18wodzfunman: to speedup testing You can increase log rate in powermgmt.c
00:08:47funmani changed 60*HZ to just HZ line 690, is that enough?
00:09:07wodz722 also
00:09:38 Quit pamaury (Quit: exit(rand());)
00:10:41 Quit slck (Ping timeout: 260 seconds)
00:12:06funmanlooks fine
00:14:39wodzfunman: isn't such line to thin for You?
00:15:03funmanhm it's visible enough i think
00:16:26wodzI was thinking of adding some grid also but honestly I have no idea how to setup portably light gray lines
00:17:02 Join slck [0] (Venci@Slackware.SlackPix.Com)
00:17:11 Quit Soap_Hotel (Quit: CGI:IRC (Ping timeout))
00:18:09funmana grid?
00:18:52wodzyes like gnuplot can do for example
00:19:18wodzit is easier to see abolute values than
00:19:40funmanare values really important? i use this screen to see if it's increasing or decreasing
00:20:52wodzbecause it is autoscaled I would like to see at first glance if values changes much or only slightly
00:21:14 Join Soap_Hotel [0] (
00:21:37funmanperhaps just make the grid 1pixel and the line wider
00:22:33wodzmaking line wider is not so easy I am afraid
00:24:12wodzbut maybe ploting copy of the curve shifted 1px down will be enough
00:24:27 Quit bmbl (Quit: Bye!)
00:29:26 Join TillW [0] (
00:33:43 Quit TillW (Client Quit)
00:35:41funman% grep -E 'ldm.*pc' .|wc -l 102
00:36:01funmansome of them are not run on armv4 so they are ok
00:38:21 Quit kramer3d (Ping timeout: 260 seconds)
00:40:13 Quit orzech (Ping timeout: 248 seconds)
00:41:32 Join n17ikh [0] (
00:42:07 Join orzech [0] (
00:42:25TorneJdGordon: the patch needs more work, it goes wrong if the system boots with sysfont and then the user selects a font later (it will try to bufalloc then which is bad)
00:42:33TorneJdGordon: also the min/max values should be tweaked
00:42:56CIA-5New commit by 03funman (r26399): fuze*/e200v2 YUV lcd code: remove a useless instruction forgotten in r21795
00:43:20tmztis there some reason to update the bootloader on fuzev1 if it hasn't been built recently (like last year)
00:43:59 Join kramer3d [0] (
00:44:01 Quit kramer3d (Changing host)
00:44:01 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
00:44:54r0b-does anyone know how much ram the Clip/Clip + has?
00:46:40tmztwhere did tools/configure go?
00:46:45 Join robin0800 [0] (~quassel@
00:47:22funmanr0b-: 2MB+384kB iram for the v1, 8MB+1MB iram for others
00:47:26funmantmzt: never moved
00:48:04r0b-i know the Clip's cannot play Doom :P
00:48:09r0b-damn 3 color LCD
00:48:13r0b-2 Color
00:48:14tmztit failed to build, I ran make clean and now the tools directory is gone
00:48:25gevaertsr0b-: while you're correcting, it's not an LCD
00:49:29TorneJdGordon: i'll look at it soon maybe, i have some beast stuff to fix as well tho.
00:50:06 Quit wodz (Quit: Leaving)
00:53:34 Quit robin0800 (Remote host closed the connection)
00:53:47 Quit aevin (Ping timeout: 240 seconds)
00:54:03 Join robin0800 [0] (~quassel@
00:58:35 Join evilnick__ [0] (
00:59:56funmanThe AAPCS requires that all sub-routine call and return sequences support inter-working between ARM and
00:59:59funmanThumb states.
01:02:02 Quit evilnick_ (Ping timeout: 265 seconds)
01:04:41 Quit saratoga (Changing host)
01:04:41 Join saratoga [0] (~9803c6dd@rockbox/developer/saratoga)
01:10:30 Quit funman (Quit: free(random());)
01:10:48r0b-wb saratoga
01:12:06 Quit ender` (Quit: In the beginning, there was nothing. And God said, 'Let there be Light.' And there was still nothing, but you could see a bit better.)
01:16:02saratogaTuesday is the 8 year anniversary of Rockbox 1.0
01:17:29gevaertsWe'll have to celebrate!
01:21:01 Join phanboy4 [0] (
01:22:24 Quit whydoubt (Read error: Operation timed out)
01:22:29 Join whydoubt [0] (
01:28:38 Quit orzech (Ping timeout: 260 seconds)
01:30:12 Join orzech [0] (
01:32:59Battousaiby porting rockbox 1.0 to all current targets?
01:33:47 Quit petur (Quit: Zzzzz)
01:35:02 Quit halmi (Read error: Connection reset by peer)
01:42:48 Quit slck (Ping timeout: 276 seconds)
01:42:50 Join slck [0] (Venci@Slackware.SlackPix.Com)
01:45:58 Quit Soap_Hotel (Quit: CGI:IRC (Ping timeout))
01:51:03 Quit pyther (Quit: Lost terminal)
01:51:23 Quit bertrik (Ping timeout: 245 seconds)
01:53:20 Quit DataGhost (Ping timeout: 240 seconds)
01:55:01 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
02:01:26 Join S_a_i_n_t [0] (S_a_i_n_t@
02:02:28***Saving seen data "./dancer.seen"
02:12:00Unhelpfulfunman: you'd mentioned being able to adjust the display clock on clip... is this the actual rate at which the display refreshes? and could we maybe temporarily make it *faster*?
02:16:58 Quit shodan45 (Remote host closed the connection)
02:17:21 Join shodan45 [0] (
02:26:40 Quit efyx_ (Remote host closed the connection)
02:28:46 Nick keanu is now known as [keanu] (~keanu@unaffiliated/keanu)
02:36:46 Quit wincent (Ping timeout: 260 seconds)
02:44:24 Quit robin0800 (Remote host closed the connection)
02:50:15 Nick kramer3d is now known as kramnap (~kramer@unaffiliated/kramer3d)
03:05:56 Quit domonoky1 (Read error: Connection reset by peer)
03:13:53Rob2223OK, as it seems we could have found a difference between the FuzeV2s that allow RB firmware.
03:15:13 Join steve|m1 [0] (
03:17:28saratogaRob2223: since the model number doesn't even tell you if you get a v1 or v2 player, I doubt it tells you about some minor change to the frimware update
03:18:03 Quit steve|m (Ping timeout: 276 seconds)
03:18:32 Quit steve|m1 (Client Quit)
03:18:38Rob2223hmm, I just wonder how much different model numbers are out there!?
03:18:53 Join steve|m [0] (
03:18:54saratogaprobably a lot
03:19:54Rob2223well i asked a third person with the problem
03:20:00Rob2223hope i get an answer
03:20:08CIA-5New commit by 03bieber (r26400): Themeditor: Got the ParseTreeNode class in good shape, preparing to start on ParseTreeModel
03:21:07saratogayour best bet is probably to figure out what changed in the firmware files
03:21:28Rob2223thats just not easy
03:21:32Rob2223a already tried
03:21:36saratogayes but its actually useful
03:22:05saratogaeven if the serial numbers do mean something its not clear to me how that helps at all
03:22:36Rob2223when i return my brand new player i would like to get a player that accepts RB in the 2nd try
03:22:51Rob2223i only bought it for RB so i will retuen it in 10 days
03:23:01Rob2223as long i have the chance to
03:23:01JdGordonbieber: hey, is there an easyish way to get a nice debug output for a skin with your parser?
03:24:06 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
03:24:13bieberIf you stick a skin_debug_tree(test) in main.cpp it'll give you the ASCII debug info
03:24:13bieberHopefully in a few minutes here I'll have a Qt treeview working
03:26:38JdGordonalso did you see my qeustion last night?
03:27:22bieberThe one about using skin instead of WPS?
03:27:43bieberYeah, I'm pretty sure that's taken care of now
03:27:53bieberIt was just in ParseTreeModel.h/cpp, right?
03:27:55JdGordongreat :)
03:28:11JdGordonpossibly, I saw it in one of the .h's
03:28:22bieberOkay, it should be clear
03:29:02JdGordonstill WPS in the comment
03:29:12 Quit merbanan (Read error: Operation timed out)
03:30:52Rob2223saratoga: I fear that the new check has something to do with the hardware or some code that is already in the players. Cause the same .33 files working for some players and for some they dont :(
03:31:37bieberThanks, got it
03:33:14saratogaRob2223: maybe the update depends on some code on the AMS ROM, and that got updated in newer players
03:34:44saratogathey might have run out of the original run of AMS chips and had to order another batch or something from the fab
03:35:59Rob2223and they tooks the chance for a rom upgrade?!
03:36:19Rob2223do you know if the AMS rom is used by the player?
03:37:51saratogai assume its used to load the firmware file off the flash memory
03:38:12saratogatheres no other memory on the device, so i don't know what else could load the firmware
03:38:35Rob2223are there not files on a hidden partition?
03:39:14 Join moparx [0] (~moparx@unaffiliated/moparx)
03:39:18Rob2223i mean is the original firmware from a firmware upgrade file the only data in the hidden space?!
03:39:38Rob2223i didnt found any information about that
03:40:37saratoganone of this is in a partition
03:40:47saratogathe ROM is memory mapped
03:41:15saratogathe as3525 datasheet explains how it works on the fuze/clipv1, probably not much different on the clip+
03:41:19Rob2223well ok, maybe bad word coice from me. the area to where the file is flashed at the firmware upgrade process
03:41:50Rob2223couldnt there be other files/data?
03:42:21Rob2223yeah i think i know the memory mapped procedure
03:43:04Rob2223i thought the firmware is flashed onto a hidden part of the 2GB memory
03:43:18Rob2223is it flashed into the SOC rom?
03:43:41saratogaas I said above, its loaded from the NAND
03:44:35saratogaunless the NAND can be memory mapped, its probably copied into RAM by whatever code is already on the ROM
03:45:09Rob2223so the nand is the 2gb/4gb/8gb chip, right?
03:45:24saratogayes, NAND is a type of flash memory
03:46:02Rob2223so the firmware upgrade copies the fusp.bin forom the fat partinioned part of the NAD to another area of the nand thats not in the FAT partition
03:46:24Rob2223(sorry for my grammar, im already damn tired)
03:46:44Rob2223to at the hidden place of the nand could be some more information
03:47:09 Join Blue_Dude [0] (~chatzilla@rockbox/developer/Blue-Dude)
03:47:17Rob2223or maybe even the firmware upgrade routine could be thare and it did never get touched by the firmware upgrade process
03:47:23saratogaAFAIK the hidden part is a verbatim copy of the firmware bin, at least on the V1 players
03:47:32saratoganot sure if anyone ever dumped it on the newer devices
03:47:37CIA-5New commit by 03bieber (r26401): Theme Editor: Got a barely functional treeview in place
03:47:53JdGordonbieber: at a high level, how does the parse tree get new branches?
03:48:12JdGordonor, whats the breakdown?
03:48:17bieberJdGordon: Okay, now it displays the file in a treeview. The formatting sucks atm, but that'll get better when I have some more time to work on it
03:48:48Blue_DudeS_a_i_n_t: Sorry, I haven't even tried to do a eabi toolchain on Cygwin. I guess it's possible but I just have a "stock" install. Well, I also added Python for analysis scripts, but the rest is standard.
03:48:48bieberAt the highest level, everything gets broken down into either a SUBLINES branch or a LINE branch
03:48:58Rob2223if that were for the V2, too, then the code that rejects newer firmware must be in the .33 firmware upgrade anywhere.
03:49:42Blue_DudeS_a_i_n_t: for that matter I'm not really sure what eabi is, so I'm probably not the guy to ask. :)
03:49:51bieberAlso, while looking over the stuff from the treeview, I just noticed another bug in my parser. Apparently when scanning for sublines, it's stopping any time it encounters a newline, even if it's at the end of a comment
03:49:53Rob2223saratoga: my only ida atm would be disassemble the FW and trying to analyze it but even if i am able to read assembler i dont think that i will find that part :(
03:50:06JdGordonbieber: ok, viewports actually need to be the hieghest level
03:50:41saratogai think comparing the firmware files is probably easier
03:50:42JdGordonis that going to be difficult without giving the parser too much knowledge about what a viewport is?
03:50:56saratogaif you figure out whats different it doesn't matter how the firmware update is implemented
03:51:38bieberThat would really be beyond the realm of the parser, since at that point you're dividing things up according to the way the renderer will interpret it, rather than just the semantic structure of the document
03:51:40JdGordonbieber: also (a request only) can #( and #) or )# be added for multi line comments?
03:52:01bieberMulti-line comments will be easy
03:52:18Rob2223saratoga: the problem is from .28/.31 to .33 are many differences in the firmware file.
03:52:45 Quit Blue_Dude (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
03:52:52saratogayes but you only care about changes to the header
03:53:13JdGordonbieber: I'd still argue that the document structure goes viewport > line > subline
03:53:17saratogaremove the firmware bits and compare whats left
03:53:26JdGordonmaybe it is not obvious because the defualt viewport isnt put into the document
03:53:54bieberSo %V tags can never be conditional, and they can always be viewed as marking a new section of the document?
03:54:31JdGordonyes, but that include %Vi and %Vl... maybe add a char to the param list saying "start a new top level node"?
03:54:41Rob2223saratoga: theres no change in the header to 0x400 but the checksums :(
03:55:05bieberJdGordon: If it's just those three cases, I can hardcode it easy enough
03:55:27JdGordonthat would work also :) I was just thinking more for keeping the parser very generic
03:55:44Rob2223saratoga: have .31 to 33. open on right screen and .28 to .33 open on left screen. i already compared the header today
03:56:52CIA-5New commit by 03bieber (r26402): Theme Editor: Fixed parsing bug that allowed comments to form a new logical line in a skin document
03:57:15saratogaRob2223: did you try dumping firmware libraries and looking at whatever was left?
03:57:29saratogaperhaps theres more to the header then just the first 0x400 bytes
03:58:00saratogasince older players can still take this upgrade files, i assume the header didn't change much, but that doesn't mean they didn't add another header somewhere further back
03:58:05Rob2223is there a tool for?
03:58:14Rob2223i just compared the .bin binary
03:58:23saratogayou haven't actually dumped one of the bin files?
03:59:10bieberJdGordon: I'll have it divide documents up by viewports whenever I have a chance to code some more, it shouldn't take me long
03:59:11Rob2223no i compared the fuzpa.bins
03:59:30 Quit DerPapst (Quit: Leaving.)
04:00:08Rob2223saratoga: you didnt have a compiled win26/64 version at hand, or?
04:00:11saratogayou'll probably want to look at the amsinfo tool's parser to get a feel for how the file works
04:00:57Rob2223yeah, good idea. i read all i could found until now
04:01:36Rob2223saratoga: have you a compiler chain set up?
04:01:53saratogai don't know how to compile stuff for windows, and since you'll have to change the code anyway i'm not sure how much use a binary would be
04:02:12JdGordonbieber: sweet, and please remember to add the default node so if the first non comment isnt %V it will still work as expected
04:02:32***Saving seen data "./dancer.seen"
04:02:58bieberGot it
04:10:49 Quit GeekShadow (Ping timeout: 264 seconds)
04:15:05 Join JdGord [0] (~jd@
04:18:27 Quit orzech (Ping timeout: 260 seconds)
04:19:10Rob2223saratoga: i just compiled it with a gcc windows version i had, but i just get "ERROR: reading firmware: No error" when im trying to dump .33. may i ask why you sayed i _have_ to modify the source? hope i didnt bother you
04:19:52saratogaRob2223: well if the tool already knew what the difference between the firmware files was, you probably wouldn't be having this problem
04:21:04Rob2223it doesnt dump the older 02.01.17 either. but i see what you say.
04:22:51saratogaRob2223: it works for me
04:23:47Rob2223then my compilation is bad :(
04:24:17Rob2223or a windows issue
04:25:16 Quit moparx (Ping timeout: 276 seconds)
04:29:48 Join Thetorminator [0] (
04:31:19 Quit Thetorminator (Client Quit)
04:32:26CIA-5New commit by 03jdgordon (r26403): zip up the output after doing all the skins
04:32:45CIA-5New commit by 03jdgordon (r26404): and woops
04:36:19Rob2223saratoga: ill give up for now. i dont see why it compiles without error but just doesnt work. but im over-tired. too. will try tomorrow under linux. but thx for your help
04:36:21saratogaskimming the amsinfo output, theres an extra 100KB (exactly) of "unknown" blocks
04:36:35saratogait may not actually run on Windows
04:37:10Rob2223could you zip/rar me a dup of .28 .31 .33 taht i could check the diff?
04:37:52Rob2223ah well, dont want to bother, i hope it runs under kubuntu
04:38:54saratogaheh if I edit the tool to dump the unknowns i get about 5000 512 byte files
04:39:13saratogai guess figuring out how the blocks fit together into files would be helpful
04:39:22 Quit Barahir_ (Ping timeout: 265 seconds)
04:40:35 Join Barahir [0] (
04:40:46Rob2223well, assuming these blocks have the same sequence and the same filename i could compare them with total commander directory sync.
04:40:55Rob2223ok i bet linux has something like that, too
04:41:40 Quit amiconn (Disconnected by services)
04:41:42 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:42:04 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:42:34 Quit pixelma (Disconnected by services)
04:42:36 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:42:55 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:44:23Rob2223saratoga: are you still on this? i have to go to bed.
04:45:04saratogai'm working on something else now
04:45:18Rob2223ok. then thx for help :)
05:00:26JdGordonbieber: is tag_table.[ch] technically outside of the parser? the reason I ask is because I want to add a bunch of info to that struct which is only really relevant to the actual build (not the theme editor or parser in general)
05:02:19JdGordonthe extra info will make tag_table.c more annoying for you, but if a tag is ever added or changed it means only fiddling with one place to make the changed tag acpeted everywhere
05:07:22JdGordonthis is even ignoring the fact that some tags are target dependant...which I guess makes perfect sense to keep ignorin
05:32:30 Nick kramnap is now known as kramer3d (~kramer@unaffiliated/kramer3d)
05:37:11 Quit [CGL] (Ping timeout: 245 seconds)
05:41:12 Join [CGL] [0] (~CGL@
05:45:35 Quit Battousai (Remote host closed the connection)
05:45:48 Join Battousai [0] (~bryan@gentoo/developer/battousai)
05:47:40JdGordonbieber: FS #11333
05:52:31 Join SliMM2 [0] (
05:55:04SliMM2What are the key mappings of the Spectrum emulator on an iPod?
05:57:18 Quit Battousai (Quit: No Ping reply in 180 seconds.)
05:58:12 Join Battousai [0] (~bryan@gentoo/developer/battousai)
06:02:36***Saving seen data "./dancer.seen"
06:06:51 Join Leanne [0] (
06:10:59 Quit SliMM2 (Quit: SliMM2)
06:56:48 Quit JdGord (Quit: Bye)
07:05:01 Quit Horscht (Ping timeout: 240 seconds)
07:06:25saratogafunman: (for the logs) someone else reported their clip dying when they installed rockbox (using rbutil), do you have a theory about that?
07:06:31saratogathis one was also a clipv2
07:09:03 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
07:09:35Leannehey so even if I use rockbox utility for mac....the ipod has to be formatted for windows to make it work?
07:16:08 Quit shodan45 (Remote host closed the connection)
07:24:58 Part toffe82
07:25:59topikLeanne: the ipod has to be formatted as FAT. it's the only filesystem rockbox knows.
07:38:58 Quit Leanne (Ping timeout: 258 seconds)
07:46:48bieberJdGordon: It shouldn't be any problem at all, since I'm just using a function to look up info from the table and pick out the field I need for the parser
07:47:19JdGordoncool :)
07:51:53CIA-5New commit by 03bieber (r26405): Applied JdGordon's patch to add tokens to the Theme Editor tag table
07:56:11 Quit JdGordon (Quit: Leaving.)
08:00:04 Join Horschti [0] (~Horscht2@xbmc/user/horscht)
08:02:40***Saving seen data "./dancer.seen"
08:02:55 Quit Horscht (Ping timeout: 260 seconds)
08:07:52 Quit panni_ (Quit: ( :: NoNameScript 3.81 :: ))
08:12:15 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
08:27:05 Join ucchan [0] (
08:29:37ucchanS_a_i_n_t: both binutils and gcc (arm-elf-ebmi) build success on my cygwin !
08:31:29 Join Leanne [0] (
08:32:30ucchanThe reason for the problem of INCLUDE_PATH (not find <alloca.h>) is that the setting of my window's native PATH was broken.
08:32:34 Quit kramer3d (Quit: Leaving)
08:34:37ucchanThen binutils and arm-elf-gcc-ebmi were normally created without changing
08:37:03Leannehow do I tell what gen my ipod classic is?
08:37:19Leanneactually....can any ipod classic use rockbox? or all of them are no dice?
08:45:22 Quit r0b- (Ping timeout: 264 seconds)
08:45:26 Join vaguerant [0] (
08:45:31 Quit vaguerant (Changing host)
08:45:32 Join vaguerant [0] (~vaguerant@wikipedia/vague-rant)
08:46:26 Quit Leanne (Ping timeout: 258 seconds)
08:46:48 Join r0b- [0] (
08:49:16vaguerantKind of a dumb question re: RoLo; does it just boot original firmwares renamed with an appropriate extension?
08:49:57vaguerantSo like clppa.bin from SanDisk (the Clip+ firmware) renamed to ANYTHING.sansa for booting on Clip+?
08:50:28saratogai don't think thats going to work
08:51:02saratogaAFAIK we don't actually have code to load the OF images into RAM, we always let the sandisk software do that
08:53:14JdGordonit might work.. unlikely though
08:54:46saratogayou'd probably have to setup memory correctly for it too
08:54:53 Join JdGordon1 [0] (
09:01:50 Quit JdGordon (Ping timeout: 240 seconds)
09:04:51 Join Rob2222 [0] (
09:08:01 Quit Rob2223 (Ping timeout: 245 seconds)
09:14:10 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
09:19:35S_a_i_n_tucchan: I am glad you managed to get it working, my windows native PATH is as it should be, I tried to build arm-elf-eabi last night...and fell asleep, when I woke up it had failed. :'(
09:20:27S_a_i_n_tOn kugel's suggestion I'm trying to build eabi by modifyine line 29 of to "make" instead of "make -j4"
09:20:45S_a_i_n_tseems promising so far, it got further than it did last night ;)
09:23:57S_a_i_n_tIt just failed to build again...
09:23:59ucchancygwin 1.7's packages are downloding now. maybe tomorrow, I can try build on cygwin 1.7.
09:25:15S_a_i_n_tucchan: My eabi build just failed again...I really don;t know much about this, other than it's breaking and I don't know how to fix it.
09:27:56 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
09:28:21 Join BHSPitMini [0] (
09:29:01S_a_i_n_tIf anyone wants to look at my arm-elf-eabi build errors:
09:29:02ucchanCygwin 1.5 and 1.7 can be used at the same time. Should install cygwin 1.5 ?
09:29:13S_a_i_n_tAny insight would be *greatly* appreciated.
09:30:10S_a_i_n_tucchan: Perhaps I should try that. I'd really like to know why it doesn't work though.
09:30:22S_a_i_n_tI don't like things to get the better of me like this.
09:32:03ucchanwrong cygwin's gcc-4 ?
09:32:16 Join stoffel [0] (
09:34:23S_a_i_n_tI'm thinking about just installing ALL of the CygWin-Devel packages...then if it works, narrowing down what I actually need.
09:34:47S_a_i_n_tI suspect that the base packages on the Rockbox Cygwin wiki don;t accomodate for eabi
09:35:35S_a_i_n_tThe packages on are from, that may well be correct.
09:39:22S_a_i_n_tucchan: Hmmm, pastebin cut the end off my cygwin output
09:39:34S_a_i_n_tthis is the part where it actually errors
09:39:44S_a_i_n_tdoes that mean anything to you/anyone?
09:41:51saratogaucchan: did you see my email?
09:42:09ucchannot install libiconv?
09:42:31S_a_i_n_tucchan: checking the cygwin ow.
09:42:55S_a_i_n_t"checking my cygwin packages now" rather
09:43:37JdGordon1i finally got my beast back together... not the origional hard disk, im onyl getting a white lcd.. is that normal?
09:46:04 Quit vaguerant (Ping timeout: 260 seconds)
09:51:02 Quit JdGordon (Quit: Leaving.)
09:51:06ucchanI installed cygwin 1.7 packages. I try build binutils and gcc-arm-elf-eabi on cygwin 1.7.
09:51:32 Join DerPapst [0] (
09:51:58S_a_i_n_tucchan: Could you just try running and see if that fails for you also?
09:52:04S_a_i_n_tit would be really good to know
09:52:13S_a_i_n_tit's OK if you'd rather not though.
09:52:23 Join flydutch [0] (
09:53:08S_a_i_n_tucchan: By the way, turns out the libiconv wasn't installed. So, I'm trying again.
09:53:44S_a_i_n_tI'm making a list of packages the rockbox wiki doesn't include so that if I manage to build arm-elf-eabi toolchain I can update the wiki
09:56:48 Join petur [0] (~petur@rockbox/developer/petur)
09:58:37 Join Akranis [0] (
10:00:55ucchanBecause a necessary package did not suffice, the build has not been done yet.
10:02:44***Saving seen data "./dancer.seen"
10:05:50S_a_i_n_tI installed libiconv (and it's dependencies)...trying again.
10:06:18amiconnBuilding the eabi toolchain on cygwin 1.7 does work
10:06:19S_a_i_n_tI'd like to do this (and taking notes as I go) so I can update the wiki at the end of it if I finally get it right.
10:07:00 Join DataGhost [0] (
10:07:00 Quit DataGhost (Changing host)
10:07:00 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
10:07:23S_a_i_n_tamiconn: Oh, I'm sure it does...but obviously not if you follow the rockbox wiki dorections. There are obviously packages needed that are not noted.
10:07:38amiconnYou need libmpfr-devel for gcc 4.4 and higher
10:07:46S_a_i_n_tOn cygwin 1.7 I can even get arm-elf to build using :/
10:08:38*amiconn installed cygwin 1.7 on this box back when it was beta
10:09:11amiconnEven then building the rb toolchains was no problem
10:09:36*amiconn isn't using it much these days, because it's so slow
10:13:20ucchanmpfr can build gcc-3.4.4 (I build yesterday)
10:14:03 Quit Barahir (Ping timeout: 260 seconds)
10:14:07ucchansorry noise my comment
10:15:22 Join Barahir [0] (
10:21:42S_a_i_n_tOK...right. Installed libiconv, libmpfr, gmp, and a bizallion dependencies for them...*fingers crossed*
10:22:12ucchanI start now...
10:22:43amiconnYou don't need a bazillion packages... just libmpfr-devel (+ dependencies) like I said
10:24:30S_a_i_n_tthe "bazillion" is the dependencies.
10:24:47S_a_i_n_tI also needed libiconv according to a previous error it spat at me.
10:25:17 Quit CaptainKwel (Quit: Ex-Chat)
10:25:34S_a_i_n_tIt'd probably be safer to just instal ALL the devel packages....but I' trying to nut this out to update the wiki.
10:26:15amiconnI don't have libiconv explicitly on my list
10:26:42amiconnMaybe it's a dependency of some other package I do install anyway for a rockbox dev envrionment
10:27:00S_a_i_n_t(error from trying to build eabi)
10:27:00 Join ender` [0] (
10:27:02*amiconn has a list of just 20 packages
10:27:28amiconnOf course those will install some more as dependencies
10:28:08S_a_i_n_tI've been trying to note everything I've installed so far...but I fear I've missed a few dependencies.
10:28:20S_a_i_n_tWhich, should be ok as long as I got the main packages.
10:34:36S_a_i_n_tamiconn: Thanks...I'll see if this build works. Then I'll compare it with my installed packages.
10:35:09S_a_i_n_tIt may seem a stupid question, but what did you do to output your installed packages like thst?
10:35:20S_a_i_n_tI'm guessing you didn't type it out by hand.
10:36:40S_a_i_n_tthough...not listing dependencies, I'm not sure it'll help much.
10:37:56 Join mischasworld [0] (
10:44:55amiconnI did type that list by hand
10:45:43amiconnIt's not the list of installed packages, it's just the list what packages I selected in addition to the default. It also doesn't include dependencies
10:47:33 Join merbanan [0] (
10:48:33ucchannow binutils build success.
10:58:01S_a_i_n_tsurely there's a way to get cygwin to output installed packages...but, I can't seem to find it. :/
11:00:02 Quit merbanan (Ping timeout: 240 seconds)
11:04:49 Join bluebrother [0] (
11:04:49 Quit bluebrother (Changing host)
11:04:49 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
11:08:01 Quit bluebroth3r (Ping timeout: 252 seconds)
11:19:22 Quit JdGordon1 (Quit: Leaving.)
11:26:10 Join ender1 [0] (
11:26:47 Quit ender` (Disconnected by services)
11:26:52 Nick ender1 is now known as ender` (
11:29:02 Join halmi [0] (
11:31:42 Join funman [0] (~fun@rockbox/developer/funman)
11:37:17funmansaratoga: no idea what went wrong for his clip
11:40:54 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
11:41:33funman <- building with -mthumb -Os makes no difference for battery life
11:42:29 Quit halmi (Quit: halmi)
11:44:10funmanamiconn: what do you think of losing 4 bytes per ASM function and 1 cycle per return on arm7tdmi (no difference on others) to make armv4t able to return to thumb code (change ldm sp!, {...,pc } to ldm sp!, {....,lr} + bx lr ? is that a big deal ?
11:46:58funmanI think a #if !defined(THUMB) || ARM_ARCH == 4 would be just too much
11:50:14 Join Buschel [0] (
12:01:12 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
12:02:46***Saving seen data "./dancer.seen"
12:04:40 Join kugel [0] (~kugel@rockbox/developer/kugel)
12:08:34funmankugel: using bx in system-arm.h macros reduces ram usage by 2.75kB (if we do that and don't make the macros real functions we can leave them inlined on ARM builds)
12:09:32Buschelanyone with a Cowon D2 willing to test a patch?
12:10:05kugelfunman: nice
12:11:02funmani think making them functions will make the code smaller though because there will be just one bl each time they are called
12:11:58funmanbtw i was expecting some change for battery life (for good or for bad) since memory use is modified but nothing :/
12:15:02 Quit BHSPitMini (Ping timeout: 240 seconds)
12:15:07 Join JdGord [0] (~jd@
12:21:32funmanmaking them functions makes things bigger :?
12:26:07 Quit JdGord (Ping timeout: 252 seconds)
12:27:13 Quit Barahir (Ping timeout: 252 seconds)
12:27:21 Join efyx [0] (
12:27:41funmanah no, saves 1344 bytes
12:28:53 Join Highlander [0] (
12:31:08 Quit ucchan (Ping timeout: 248 seconds)
12:32:18CIA-5New commit by 03bertrik (r26406): Slovak language update ...
12:41:11kugelbertrik: are you backporting it to 3.6?
12:43:01bertrikIs there something wrong with the commit I just did?
12:44:04funmanbertrik: language updates should be applied to the branch, too
12:45:09JdGordonkugel: fs#10030.. whats the story?
12:47:03bertrikI don't know how to do that
12:49:46 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
12:50:54JdGordonbertrik: svn co the branch, then commit...
12:51:12JdGordonI'm bored so I can do it for you
12:52:40CIA-5New commit by 03jdgordon (r26407): Slovak language update ...
13:01:39amiconnfunman: Does thumb actually improve performance?
13:02:09funmanAC3 and AAC are a bit faster on armv5
13:02:35funmanAAC is among the slowest codecs but AC3 is quite fast
13:02:40amiconnI am thinking of PP especially
13:03:05funmani have no PP to test
13:03:18 Join ucchan [0] (
13:03:45 Quit stoffel (Remote host closed the connection)
13:03:59funmanyou just need the diff to target/arm/*.S I pasted yesterday and the script on the FS# task to test
13:04:24funmanmpegplayer won't work and perhaps some codecs but every test_file which loaded into 8MB of memory decoded fine
13:04:59funmanerr, I didn't run test_codec on clipv1, so on armv4t at least mp3 will work, dunno for others
13:05:31ucchanI'm leaving for a while, binutils and gcc (arm-elf-eabi) already built. no error, install succeed ! built on cygwin 1.7.
13:08:18funmanamiconn: i'll run test_codec on fuzev1
13:08:33 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
13:09:00ucchanI try build other targets cross comilers. maybe no problem on cygwin 1.5/1.7.
13:10:10JdGordondomonoky: I've got the updater script "working" (I tihnk)... can i get the zips from the theme site?
13:10:57JdGordonI've got it making a diff of the changes which I was hoping could be auto emailed to the theme authors
13:13:44domonokyJdGordon: i can get you all the zips, but what should the theme authors do wit hthe diff ?
13:14:07JdGordonjust have a look I guess... more to make sure it didnt do anything stupid
13:15:11JdGordonalso with the zip, I need to know if it is for a mono display or not
13:15:24 Join Barahir [0] (
13:16:33domonokyif a zip is mono or not, is only noted in the DB, so i cant tell you that if i download all the zips somewhere.
13:17:23domonokyalso there are all those old zips with outdated themes, you cant know which is the newest without info from the db.
13:17:37S_a_i_n_tyou could perhaps do some check on the bitmaps depth with the script?
13:17:47*S_a_i_n_t has no idea if that's actually possible or not.
13:18:17JdGordondomonoky: you need to do a db query to get the zips anyway right? can you do 2 then? one for mono and one for others?
13:18:30JdGordon%V needs different handling for mono which is why i ask
13:19:02domonokyJdGordon: i dont need a query to get the zips, i could just download the zips from the server...
13:19:26JdGordon"(21:17:22) domonoky: also there are all those old zips with outdated themes, you cant know which is the newest without info from the db." <?
13:19:48domonokyye, that needs the db too...
13:20:29JdGordonright, but if we only want to work on the latest zips?
13:20:31JdGordonor too hard?
13:22:53domonokyits not really hard, but its a few hours work. I would have to write a php script with the correct querys etc, i probably wont have enough time for that till devcon.
13:23:43JdGordonprobably a good time to leave it to then :) you and scorche can decide how to handle the theme site split
13:25:28domonokyjup, if we have a working skinupdater, we should be able todo that a devcon.
13:26:07JdGordonfaster to run it locally ont he server instead of here
13:29:05*S_a_i_n_t notes that FS #11270 is getting rather fancy anyone considering this for a commit?
13:33:07 Join Kitr88 [0] (
13:33:21 Quit Kitar|st (Read error: Connection reset by peer)
13:33:38*kugel wonders why saratoga is always refering to me as the pictureflow guy
13:34:30 Join Kitar|st [0] (
13:35:13AlexPkugel: cos you like shiny :)
13:37:22 Quit Kitr88 (Ping timeout: 240 seconds)
13:38:54JdGordonAlexP: hey! i finally got my beast back together... when i turn it on the LCD is white.... how do i get i up and running?
13:39:01 Quit AlexP (Remote host closed the connection)
13:39:16 Join AlexP [0] (~ap@rockbox/staff/AlexP)
13:39:29JdGordonnot using the origional hard disk if that changes anything
13:40:32 Quit yosafbridge (Ping timeout: 245 seconds)
13:41:16funman < test_codec arm/thumb results on fuzev1
13:42:34 Join teru [0] (
13:42:53funmanwma/mp3 are faster, aac/mpc/cook/vorbis are slower
13:44:57kugelI'm surprised that it has so little influence
13:45:22funmani thought it would be much slower
13:46:07JdGordon11270 looks ok.. is it actually ready to be checked in?
13:46:20gevaertsDoesn't this just prove that the critical bits are asm anyway?
13:46:43funmanspeeds vary between 98.77% and 100.18%
13:47:39 Quit ucchan (Quit: Leaving...)
13:47:56funmangevaerts: yeah probably
13:48:39funmanon clipv1: 97.87% <-> 104.64%
13:49:31*kugel wonders how thumb can be faster
13:49:43 Join yosafbridge [0] (
13:50:08funmanless code to load from ram
13:50:23funman-> more things in cache
13:50:36gevaertsthumb could be a win on low-iram systems
13:51:14funmandepends what you put there, if it's data (tables) then it doesn't change
13:51:28funmanneither if it's 32bits asm routines
13:51:46funmanwell one must try, i think this includes the gigabeats? no other dap?
13:53:53 Quit AlexP (Quit: Please insert girder)
13:54:54CIA-5New commit by 03jdgordon (r26408): add support for the possible viewport colour tags (%Vf and %Vb). use -c to disable them
13:55:56funmanhm i misread the tests
13:56:07funmanon clipv2 everything is faster, except aac
13:57:18funmanwell I had read correctly, I misread the test_codec parser output
13:59:59 Quit teru (Quit: Quit)
14:01:35funmanI misread fuzev1 results though, only mp3 and wma are faster
14:01:39 Nick fxb__ is now known as fxb (
14:02:46funmanfuzev1 has 8kB + 8kB (I+D) caches
14:02:47***Saving seen data "./dancer.seen"
14:06:14 Join teru [0] (
14:13:17funmanclipv2 has 16kB + 16kB, perhaps thumb has less influence
14:13:45funmanI have put codec numbers on FS #6734
14:15:11funmansomeone wants to test on PP ?
14:17:29 Join wincent [0] (
14:19:23 Quit mischasworld (Ping timeout: 276 seconds)
14:21:22kugelfunman: if you do me a build, sure
14:21:35funmanwhich target?
14:22:52Rob2222may i ask that tumb is you talking about every day?
14:24:59Rob2222ty funman
14:26:05Rob2222ah, smaller opcode to sace code data size, nice
14:29:40funmankugel: (i only tested wma/mp3/mpc/vorbis/cook/aac, others might not work)
14:31:02kugelfunman: doesn't boot
14:31:17kugelfreeze at bootlogo of the bootloader
14:32:03funmanlet me change lcd code
14:33:45funmankugel: try again? (same url)
14:34:00kugelgive me some minutes, I need to unbrick it first :)
14:34:22funmanthere's no hardware power off ? :/
14:34:36funmanit should be in UIE() undefined instruction
14:34:56funmanhmm well UIE() use lcd ops, which cause another undefined instruction :p
14:36:07kugelfunman: nevermind, I forgot that I can boot the OF :p
14:36:10Rob2222do you use opened devices where you can reconnect the battery to unbrick?
14:37:32kugelfunman: still no boot
14:37:48 Quit teru (Quit: Quit)
14:39:06 Join halmi [0] (
14:39:34gevaertsRob2222: if you can revive a player by just reconnecting the battery, it's stuck, but it's not anywhere near bricked
14:39:59funmani wonder about the ldm.*pc in thread.c -> it seems to work on clipv1/fuzev1
14:40:31kugelfunman: because it works on arm9
14:40:44funmannope, on armv5
14:40:52Rob2222gevaerts: ok. good point. so you use these special usb mode or a jtag, i assume. ok
14:41:00kugelonly arm7 has the ldm pc constraints AFAIK
14:42:59funmanand armv7+ can use mov pc,
14:43:00gevaertsRob2222: it will depend on the player. Actually, us developers don't generally count players that can be revived using a special USB mode as bricked either :)
14:43:26 Join aevin [0] (eivindsy@unaffiliated/aevin)
14:43:30gevaertsIf you need soldering, it's bricked. If not, it's merely being annoying
14:43:31kugelfunman: arm7tdmi, not armv7
14:43:43funmanno, i said armv7
14:43:51kugelbut I said arm7
14:43:57funmanarmv4t must use 'bx'
14:44:08funmanarmv5t+ can use ldm ... { , pc }
14:44:12funmanarmv7+ can use mov pc, xx
14:44:58kugelI'm not sure about that
14:45:27Rob2222gevaerts: ok
14:45:37funmankugel: well prove me wrong
14:45:55kugelI'm trying to, but google fails on me :)
14:46:09funmanthen perhaps it's right?
14:48:30funmanbuild updated with thread.c:load_context() change
14:48:51funmanranma: VIC_INT_EN_CLEAR = X; while (VIC_INT_ENABLE & X) ; => still crashes
14:49:43funmanif i disable DMAC interrupt just after enabling the channel nothing happens (progress bar doesn't move) so I guess there is some race condition
14:52:26kugelfunman: same url?
14:53:32 Join stoffel [0] (
14:53:49kugelfunman: no boot :(
14:56:00funmanah i must check for ldr pc, [X]
14:58:26 Join mischasworld [0] (
15:02:53funmannothing harmful :/
15:05:54funman"(((ldr|mov) *)|(ldm.*))pc" should cover all armv4 returns which don't preserve T bit ?
15:06:35 Join einhirn [0] (
15:11:10funmannrv2e_d8.S use 'mov pc, lr' and comment says "stay in current (THUMB) mode", but according to this can't work
15:11:57 Quit mischasworld (Remote host closed the connection)
15:16:00 Quit funman (Quit: tiuq\)
15:20:42 Join fdinel [0] (
15:27:05r0b-Rob2222 didnt you try to rockbox a Fuze V2?
15:33:56 Quit bmbl (Ping timeout: 258 seconds)
15:38:13 Join MethoS- [0] (~clemens@
15:38:46 Join anewuser [0] (anewuser@unaffiliated/anewuser)
15:38:46 Quit MethoS- (Remote host closed the connection)
15:40:01 Quit Zarggg (Quit: Zarggg)
15:43:18 Join selectohh [0] (
15:44:17selectohhdoes anyone know if there is guidelines / a website with guidelines for submitting a plugin to be accepted in the build? like how it needs to be commented or some stuff like that?
15:44:33S_a_i_n_tcheck the wiki
15:44:49kugelselectohh: only docs/CONTRIBUTING
15:44:49S_a_i_n_tone sec, I'll find the link.
15:44:55selectohhi saw something kind of general for patches.. nothing for plugins
15:45:23 Join Zarggg [0] (~zarggg@2001:0:4137:9e74:0:fbf0:beb1:ba3d)
15:46:12selectohhoh ok, i see the docs/CONTRIBUTING. thanks i think that's what i was looking for. i did google the heck out of this but for some reason google seems to overlook svn for me.
15:47:22 Join liar [0] (
15:48:13r0b-have any of you tried rockbox on the Fuze v2?
15:50:31JdGordonr0b-: do you have an actual question?
15:51:02r0b-not at this moment
15:51:04 Part r0b-
15:56:04 Quit einhirn (Read error: Connection reset by peer)
16:01:02 Quit esperegu (Ping timeout: 240 seconds)
16:02:47 Join lpereira [0] (
16:02:49***Saving seen data "./dancer.seen"
16:02:53 Join AlexP [0] (~ap@rockbox/staff/AlexP)
16:05:27 Join robin0800 [0] (
16:05:43 Join funman [0] (~fun@rockbox/developer/funman)
16:07:22 Quit AlexP (Ping timeout: 240 seconds)
16:07:48 Join Abaoji [0] (
16:08:09S_a_i_n_tJdGordon: Do you remember at all what I was going to add to skinwishlist the other day?
16:08:15 Join AlexP [0] (~ap@rockbox/staff/AlexP)
16:08:21JdGordonhaha... no :p
16:08:31JdGordongrep the irc logs maybe?
16:08:32S_a_i_n_tneither :/
16:08:44JdGordonI need stuff outside the skins to play with...
16:08:54JdGordonunless you want to talk about tag changes...
16:09:01S_a_i_n_tyes, yes you do...or you'll go insane.
16:09:08 Quit AlexP (Remote host closed the connection)
16:09:08JdGordontoo late!
16:10:19AbaojiHi everyone, I have a Sansa E200 series running most current version of Rockbox. I know Rockbox has dual-boot function. I wondering if it's possible to permanently disable manufacturer's firmware[i.e. Sansa] and use Rockbox as default firmware.
16:12:04JdGordonyou can, but it is not advisable...
16:12:10JdGordonI accidently did it to mine :p
16:12:15JdGordonjust ignore the OF
16:12:46AbaojiI want to disable the manufacturer's firmware because I bought a new FM transmitter and everytime when I when turn off the player, Rockbox won't be able to load.
16:13:12AbaojiJdGrodon: How to disable?
16:14:37kugelAbaoji: you mean the OF boots when you turn on the e200 when attached to that thing?
16:14:55Abaojisorry what's OF?
16:15:08S_a_i_n_toriginal firmware
16:15:14kugelthe manufacturer's firmware
16:15:28Abaojiyes OF boot when I attached to my FM transmitter
16:16:09 Join AlexP [0] (~ap@rockbox/staff/AlexP)
16:16:26kugelAbaoji: e200v1 or v2?
16:16:53funmanso dock works with e200v2? nice
16:17:01kugelwell, it's possible but strongly discouraging
16:17:25kugelgevaerts: ping
16:17:39 Join esperegu [0] (~quassel@
16:18:05AbaojiI understand the risk. but I already invested so much in the FM transmitter already, I want it to work perfectly. So can you pm me the instruction?
16:19:17kugeldisable the usb check in rbutil/mkamsboot/dualboot/dualboot.S, recompile mkamsboot and use that for installing rockbox
16:19:29AlexPJdGordon: Sorry, I was disconnected for a bit - you asked something?
16:19:30 Join orzech [0] (
16:19:50gevaertsAbaoji: are you running the latest bootloader? Does your e200 boot to the OF or to rockbox if you connect it to a PC when it's switched off?
16:19:52JdGordonwhite LCD on my beast... new hard disk.. what do I do?
16:19:53gevaertskugel: gnip
16:20:15*bertrik plans to fix the samsung yp-s3 bootloader to he can demo it at devcon
16:20:28kugelgevaerts: I started #ifdef SIMULATOR conversion here:
16:20:35kugelnot all need to be converted
16:20:38Abaojigevaerts: it boot to rockbox first, but it didn't work so it revert to OF.
16:20:52AlexPJdGordon: er, don't know - I've not seen that
16:21:08AlexPWhen I put the new disk in it just went through that partitioning/reformatting bit
16:21:10JdGordonmaybe i didnt put it together properly then.. what does it do when no disk is connected?
16:21:17gevaertsAbaoji: how long ago did you install the bootloader?
16:21:19AlexPJdGordon: Could you have semi disconnected the lcd?
16:21:53Abaojigevaerts: half year ago, lemme check the version..
16:22:48gevaertskugel: the very first changed line in that diff is wrong :)
16:22:49Abaojikugel: how do you recomplile mkamsboot?
16:22:52 Join panni_ [0] (
16:23:27gevaertsAbaoji: ignore me. I didn't see that you said it's a v2...
16:23:37JdGordonAlexP: hehe the ribbon came completly out
16:23:50AlexPJdGordon: That'd do it :)
16:23:58funmanAbaoji: e200v2 shouldn't "boot to rockbox first", please check the bootloader's version
16:24:03kugelgevaerts: oops, in fact I haven't even test compiled
16:24:11Abaojiok ok I will check again
16:24:23kugelI just wanted to show where things are going to
16:25:36 Quit halmi (Read error: Connection reset by peer)
16:31:13gevaertskugel: are you just converting the #ifdefs mostly blindly, or are you also checking if the ifdefs actually make sense, and if they e.g. should be handled differently for RaaA and the sim?
16:31:16bertrikI'm getting compiler warnings trying to get the samsung yp-s3 bootloader to build, it complains about format expecting 'unsigned int' but argument having type 'long unsigned int'. The argument is a uint32_t defined in stdint.h
16:31:52bertrikWhy is uint32_t defined as 'unsigned long' instead of simply 'unsigned int'?
16:33:08kugelgevaerts: the latter
16:33:26funmanbertrik: both are 32 bits, so any will do
16:33:30bertrik(this is for sprintf)
16:33:33kugelI left some #ifdef SIMULATOR as is, for instance those which are for debugging purpose only
16:34:35bertrikfunman, yes I know they are for our targets, but the compiler thinks they are different. Somehow a decision was made to make it explicitly 'long'
16:35:00funmanthe sizes are the same, the types are different
16:35:03 Quit orzech (Quit: leaving)
16:35:23funmanprinting a uint32_t requires special format strings usually, like PRId32
16:35:40kugelgevaerts: basically, those that are relevant for RaaA are converted, the others not
16:36:00funmanbut we don't have these defines in rockbox, so you can just cast it to anything, that's what other parts of rockbox do already
16:36:05gevaertskugel: probably out of scope for RaaA, but I think some of the ifdefs shouldn't be there for the sim at all, like e.g. battery_voltage in plugin.c. It's probably useful to ask MrSomeone to review the things again later on for that :)
16:36:57kugelyes, it would be a lot of work I think
16:37:09 Quit bertrik (Quit: De groeten)
16:37:12 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
16:37:40gevaertsIn some cases the ifdef is only there because people didn't feel like adding a stub I think
16:37:57kugelthat too
16:37:58gevaertsBut anyway, your job is to not make the sim worse, not to improve it :)
16:39:08JdGordonAlexP: "contact manufacturer for repair"?
16:40:46AlexPJdGordon: I can't rememeber the exact messages - can you send it a firmware with sendfirm?
16:40:57AlexPIf not, maybe the disk cable isn't fully in?
16:41:04 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
16:43:40JdGordonusb connects :)
16:44:10bertrikfunman, really? uint32_t would still fit in an unsigned int on most (if not all) targets. I don't see the need for a long, I notice only disadvantages
16:45:40funmanthere's no difference between long and int on our 32 bits targets, except that the C language treats them differently, and so *printf need different format specifiers
16:45:46funmanint isn't better, or worse, than long
16:46:07funmanit's just a different type, but since they have the same size you are free to cast them to each other
16:47:28 Quit solexx (Ping timeout: 260 seconds)
16:48:38 Quit Akranis (Ping timeout: 264 seconds)
16:49:02funmanif you want to print uint32_t properly, we should have a #define PRIu32 "lu" (for unsigned long) and you would use printf("XXX%"PRIu32"blabla", (uint32_t)x)
16:49:04 Join solexx [0] (
16:49:19 Quit togetic (Quit: WeeChat 0.3.0)
16:49:20kugelwell, int is more likely to be 32bit across platforms
16:49:28funmanor we could do PRIu32 "u" if we define uint32_t to unsgiend int
16:50:45JdGordonAlexP: where do i find sendfirm?
16:51:08gevaertskugel: I've just looked at some random occurences of SIMULATOR in apps/, and I can only wish you good luck...
16:51:26AlexPJdGordon: utils/MTP
16:52:37 Join Akranis [0] (
16:53:04 Quit Akranis (Client Quit)
16:53:44kugelgevaerts: hehe
16:53:58*gevaerts adds an idea to the DevCon2010 page :)
16:54:39funmanwe can change memory voltage from 1.8V to 1.7V on as3525v1, it seems OF does that
16:56:48gevaertskugel: have a look at e.g. apps/plugins/zxbox/spmain.c. Why on earth do those depend on !SIMULATOR?
16:57:50kugelexcellent question
16:58:17gevaertsMy guess is "possibly historical reasons"
16:59:25kugelshould I fix those most useless ones, ignore them or convert them?
17:01:03gevaertsI'd say ignore them. Anything else will make the eventual commit less clean
17:01:57kugelok, ignoring them possibly also shows how useless they are since RaaA doesn't define SIMULATOR
17:03:56funmanbertrik: gcc has a limits.h itself, are you sure it's not that one which is included?
17:06:31bertrikI'm not sure I understand any of this any longer, is there someone here who knows and can explain in simple words why we use longs for 32-bit units?
17:07:18 Quit Abaoji ()
17:08:57gevaertsbertrik: my guess is "because the logic to define them was borrowed from somewhere that did it like that"
17:09:21funmani'd say "why not?"
17:09:33kugelsome of the most strange things in firmware actually come from newlib :)
17:09:34bertrikfunman, casting hell in printf-like functions
17:10:06*kugel thinks using int instead wouldn't hurt
17:12:14gevaertsJust inverting the order in stdint.h should achieve that
17:13:07funmanit's an OK workaround, but i guess existing casts will need to be changed
17:14:28funmaninttypes.h say /* could possibly have (f)printf format specifies here */
17:15:06JdGordongevaerts: re your devcon idea.. is that cracking kugels whip? or doing his work for him? :D
17:15:31 Join AdB3 [0] (
17:15:44 Quit AdB3 (Client Quit)
17:15:55gevaertsJdGordon: neither I think. The fact that e.g. cpu_frequency is not just stubbed but #ifdeffed all over the place is not really related to RaaA
17:15:58 Join adb3 [0] (
17:16:49JdGordonyes, getting rid of all of them would be great
17:18:25 Join simonrvn [0] (
17:18:57kugelI'm always in favor of stubbing, for target builds too
17:19:21adb3I want to build a RB plugin / skin
17:19:46AlexPThose two are quite different :)
17:20:00adb3* I Know
17:20:36adb3the issue is that the skin need display images generated by the plugin
17:20:50funmanbertrik: , can you use "%"PRId32" xx" instead of "%d xx" ?
17:21:30adb3I think that would technically be called a viewer
17:21:30*kugel finds these PRIdX ugly
17:22:13kugelfunman: "%h" is a format specifier?
17:22:25kugelI always thought short/char just use %d
17:22:29funmani think it's actually hd
17:22:41ranmafunman: Is that on Clip+? Crash == "Unhandled masked IRQ 04: INT_DMAC[...]"?
17:22:42funmanjust learned about h and hh in man 2 printf
17:22:43bertrikfunman, I'm totally unfamiliar with the PRI stuff, to be honest
17:22:47funmanranma: clipv2
17:23:00JdGordonadb3: what do you want to do?
17:23:01funmanbertrik: well that's just how you should print (u)intXX_t
17:23:25funmansince the header can use int, long, or whatever fits in the desired width
17:24:17adb3I would like my MP3 player to feel like Serato's ScratchLive...
17:25:00adb31st, I would like to have a similar look
17:25:07adb3the skin part
17:25:17CIA-5New commit by 03jdgordon (r26409): Accept FS #11313 by Chris Savery. Add composer to the track info screen and some general cleen up.
17:25:19funmanhowever our format.c doesn't know 'h' so we could use d (my linux header does that)
17:26:29adb3then I would like to be able to see the 'overviews' (waveform-ish image of the track which is encoded in the ID3 tag)
17:26:40kugelfunman: I think char/short is not valid in va_arg() so there would be no advantage of %h
17:26:58funman (edited) <- ok for commit?
17:27:26funmankugel: yep, i just said that ;)
17:27:37JdGordonadb3: well, right now we have no way for a plugin to draw onto the wps. There is a WIP patch that would enable that, but nothing in svn
17:27:50kugelfunman: I mean generally not our format.c
17:27:54adb3I have the already done that last part in python
17:28:10kugelva_arg() has some constraints IIRC
17:28:31adb3well, my target is a WIP... (Cowon D2+)
17:28:32bertrikfunman, wait, please let me try it first
17:29:18ranmaMaybe the problem only happens on AMSv2?
17:29:46funmanranma: fuzev1 too
17:30:10adb3Gordan: what are thinking???
17:30:30JdGordonnot very much
17:31:04S_a_i_n_twell...I'm wondering how python will help you with an RB plugin.
17:31:20adb3just prototyping....
17:31:31adb3but havin access to all those libraries was nice
17:31:52S_a_i_n_teven so...better to do it in code that will actually runon the device.
17:32:03adb3i know
17:32:33funmankugel: where would this restriction come from ?
17:32:52kugelI don't know and I can't find anymore where I read about it
17:32:56ranmaHmm, so then I wonder why I haven't been able to reproduce it here :(
17:33:15funmanif our format.c works it just means that short/char/int have the same size as parameters
17:33:29funmanranma: it just takes 5 minutes to crash
17:33:39CIA-5New commit by 03jdgordon (r26410): fix red. no replaygain on hwcodec
17:33:55ranmaI left it running for at least 20 minutes IIRC
17:35:27*JdGordon has a topic that should be discussed at devcon, but doesnt know how to bring it up in email as he wont be there :(
17:35:39bertrikfunman, I don't really like it much to have a PRId32 macro in printf statements
17:36:05S_a_i_n_tIRC protocol/antiflaming?
17:36:34kugelbertrik: that's probably why virtually nobody uses it. but it's in fact iso c99
17:36:44adb3JdGordan: wps - refresh my memory... its the 'display', what gets rendered
17:37:06funmanit's needed for portable code which uses intXX_t
17:37:52ranmaDo you just play an mp3? Because AFAICS the pcm_play_dma_get_peak_buffer() is only called from plugins and the pcmbuf_beep (keyclick I suppose)
17:38:05adb3and this patch???
17:38:09funmanranma: yep, just playing
17:38:30bertrikfunman, ok thanks for the explanation.
17:38:35funman <- i added *intptr_t
17:39:02bertrikThe alternatives are either casting the arguments, or adding an 'l' to the % formatter, right?
17:39:21funmans/alternatives/workarounds/ yes
17:39:41funmanor just using an int in the first place ;)
17:40:14kugelfunman: shouldn PTR be %(l)X?
17:41:11funmaninstead for x/X we should have PRIxN (N= 8, 32, PTR)
17:41:29kugelah I see
17:41:35funmanand PRIo for octal
17:42:00funmani skipped also MAX, SCN (scanf), LEAST, FAST
17:42:42funmanwe don't use int_least* or int_fast*, dunno about scanf
17:42:57funmanwell it's used
17:43:02ranmafunman: I just put a breakpoint on pcm_play_dma_get_peak_buffer, it's not called during normal playback so far, only if I seek or pause/unpause.
17:43:24funmanranma: perhaps it doesn't come from this function
17:43:47funmani just reverted 26316 from current svn and it still crashed in 5 mins
17:46:38ranmaBut you blamed it on 26311 in 26316 and the former only changes pcm_play_dma_get_peak_buffer :)
17:47:00ranmaMaybe I should update to current and see if I can reproduce it then...
17:47:11funmanperhaps i blamed wrongly, i don't know
17:47:44gevaertsJdGordon, adb3: FS #11027 isn't really a work in progress. It's more of a failed experiment...
17:48:10JdGordonyeah, I saw that last comment.. must have forgotten that was how it ended :/
17:48:26adb3maybe skipping the skin would be easier, just make a plugin which draws everything. Plugins must be able to control playback
17:48:37gevaertsadb3: yes, they are
17:49:10ranmaAnyway, with INT_AUDIO the fact that it didn't lock up when you were getting the interrupt storms seems a strong indication to me that the VIC is working correctly. :)
17:49:57 Quit DerPapst (Quit: Leaving.)
17:50:07funman <- adding scanf macros will duplicate each d,i,x,u entry, just not the X
17:50:41funmanranma: i just had panicf unhandled interrupt with INT_AUDIO in vic raw status several times
17:50:56funmanwhen plugging usb cable
17:51:01funmanperhaps plugging/unplugging
17:52:19 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
17:52:30 Quit linuxstb (Ping timeout: 240 seconds)
17:53:34bertrikWe don't have a cabbie radio screen right? Maybe a nice idea to discuss / design one at devcon?
17:53:59funmanand at the beginning of the port we had this message several times (unhandled interrupt with no unmasked source) but i don't remember the cause (if we ever foudn it)
17:55:17JdGordonbertrik: we dont, I've been tring to get the forum to do one, but they are useless
17:55:21adb3JdGordan: the seccond part of the plugin would be to utilize Serato's music library format, so that users could view their tracks in the order they are use to. Somebody else has made a Java tool for working with Serato's library & crates
17:58:04 Quit GeekShadow (Ping timeout: 258 seconds)
17:59:51kugelfunman: do we have scanf?
17:59:54adb3anyways, thank you for your help. I have to go DJ now. It seems like this will be a rather large project, but by no means imposible... have to start coding up concepts
18:00:10CIA-5New commit by 03funman (r26411): inttypes.h: add (some) iso c99 fprintf format specifiers ...
18:00:23funmankugel: yep i see it in firmware/libc (sscanf)
18:00:58funmanafaict only imported plugins use it
18:00:59adb3ooh one more question: whats the ID3 library / functionality???
18:01:15funmanrockboy/zxbox/doom/pdbox and lua exports it too
18:02:52***Saving seen data "./dancer.seen"
18:03:02JdGordonadb3: ?
18:05:05adb3I need to read obscure ID3 tags (a GEOB tag....)
18:05:43ranmafunman: weird, I'm pretty sure I've never hit the unhandled interrupt panic so far... :(
18:07:39JdGordonadb3: you'll probably need to hack up the metadata parser then also
18:07:58CIA-5New commit by 03funman (r26412): inttypes.h: remove excessive PRI*PTR declaration when long isn't 64 bits
18:07:59adb3sometimes the wiki really pisses me off
18:08:10S_a_i_n_tit does?
18:08:13adb3theres no real order...
18:08:17JdGordononly sometimes?
18:08:47S_a_i_n_tadb3: how *should* it be ordered?
18:08:54S_a_i_n_tI'm genunely curious.
18:09:31adb3its seems comprehensive, but trying to navigate to a topic (sometimes even a direction) is hard
18:10:02adb3its always: wiki > main > page
18:10:10ranmafunman: Hmm, you said you were using a newer gcc, right? I sure hope it's not compiler related :)
18:10:23funmanwhat's with the player sim red? i hit that sometimes
18:10:30*ranma is still using arm-elf-gcc (GCC) 4.0.3
18:10:31adb3not wiki > field > topic > page
18:10:32funmanranma: i hit it with 4.0.3
18:10:41ranmaHmm, ok.
18:10:47S_a_i_n_tadb3: Well, pretty much everything you'll be interested in is in DocsIndex
18:10:49funmani use default gcc, too
18:11:16JdGordonfunman: minor bug in the dependancy stuff probably
18:11:53funmanranma: i'll try on fuzev1, gcc 4.0.3, and reverting between those 2 revisions to check again (but later)
18:12:00S_a_i_n_tadb3: There is also a search, so the ordering doesn't really matter.
18:12:56*ranma is compiling 26412 with 26316 reverted for testing
18:13:19amiconnHrmm, no wodz around
18:14:10ranmaConfigure 'normal' and -g added manually to Makefile (if I use (D)ebug it's crashing)
18:18:03kugelhm, only 130 ifdefs to go
18:19:21kugelah no, 200
18:20:21 Quit adb3 (Quit: CGI:IRC (EOF))
18:20:45 Join newnick [0] (
18:20:59 Nick newnick is now known as adb3 (
18:26:07 Quit efyx (Remote host closed the connection)
18:31:07 Quit stoffel (Remote host closed the connection)
18:32:55 Join efyx [0] (~efyx@
18:37:33 Join domonoky1 [0] (
18:38:41 Quit domonoky (Ping timeout: 260 seconds)
18:40:52 Quit adb3 (Quit: CGI:IRC (EOF))
18:42:57 Quit Rob2222 (Quit: Rob2222)
18:43:21 Join linuxstb [0] (~linuxstb@rockbox/developer/linuxstb)
18:51:10 Quit TheSeven (Ping timeout: 240 seconds)
18:52:46*kugel is confused by the define in spc_codec.h
18:54:21funmanranma: why -g ? do you use gdb?
18:57:48 Join Rob2222 [0] (
19:01:22 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
19:10:46 Quit efyx (Quit: Quitte)
19:12:09 Quit n17ikh (Ping timeout: 260 seconds)
19:13:30 Join kugel_ [0] (
19:13:46 Quit kugel (Disconnected by services)
19:13:50 Nick kugel_ is now known as kugel (
19:14:01 Quit kugel (Changing host)
19:14:01 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:18:14 Join n17ikh [0] (
19:22:35 Join DerPapst [0] (
19:23:34funmanranma: fuzev1, 15 minutes mp3 OK with r26412, now trying r26316 (should be OK too)
19:31:24 Join kugel_ [0] (
19:31:44 Quit kugel (Disconnected by services)
19:31:48 Nick kugel_ is now known as kugel (
19:31:52 Quit kugel (Changing host)
19:31:52 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:32:33 Quit robin0800 (Remote host closed the connection)
19:33:29funmanin case you missed it, the initial report was on
19:38:09 Join stoffel [0] (
19:38:14 Join epicfailguy [0] (
19:39:31 Join MethoS- [0] (~clemens@
19:40:53 Quit stoffel (Remote host closed the connection)
19:40:57 Quit wincent (Read error: Connection reset by peer)
19:42:59 Join stoffel [0] (
19:43:31 Join wincent [0] (
19:44:17 Join CaptainKwel [0] (
19:46:12S_a_i_n_tIs the patch that allows Doom to build under eabi in SVN?
19:48:46kugelalso, it did build, it just didn't run very well
19:49:03S_a_i_n_tIt didn't to begin with ;)
19:53:30kugelgevaerts: some plugins use #ifdef SIMULATOR because it's more powerful :(
19:54:00gevaertsin what way?
19:54:27 Quit wincent (Changing host)
19:54:27 Join wincent [0] (~wincent@rockbox/developer/wincent)
19:55:12kugelgevaerts: e.g. beatbox.c, although that one seems it isn't compiled anymore
19:55:22S_a_i_n_tIS the Nano2g *really* "incomplete, less usable, has problems that limit it to advanced users"?
19:55:35 Quit stoffel (Remote host closed the connection)
19:55:58S_a_i_n_tI know it isn't "stable", but it doesn't really fit the "unstable" definition.
19:56:27funmanis complete in the sense 'with the features AND the bugs' ?
19:56:40S_a_i_n_t(nitpicking I know)
19:56:48gevaertsS_a_i_n_t: that's not meant as *and*
19:57:14S_a_i_n_tgevaerts: but, it isn;t really *any* of those is it?
19:57:32S_a_i_n_tisn't it just the wonky FTL?
19:57:40funmanS_a_i_n_t: what are the problems with FTL ?
19:57:52kugelS_a_i_n_t: haven't you just said yourself it wouldn't be stable?
19:57:59S_a_i_n_tfunman: Nothing far as I know.
19:58:21kugelthat means it's "less stable" and fits very good under unstable
19:58:28S_a_i_n_tIt used to WSOD daily, but...mine are what I would consider to be more stable than my Nano1gs
19:58:33gevaertsS_a_i_n_t: in that case the developers who know about it should propose promoting it to stable
19:59:09 Quit n17ikh (Ping timeout: 248 seconds)
19:59:13S_a_i_n_tTheSeven has AFAIK
20:00:49 Join ssorgatem [0] (~ssorgatem@
20:01:36funmanranma: r26315 & r25316 played fine 15 minutes on fuzev1 ..
20:02:47gevaertskugel: those beatbox #ifdefs look interesting indeed
20:02:53***Saving seen data "./dancer.seen"
20:03:15kugelthe exact same is in midi
20:03:47S_a_i_n_tkugel: err, sorry. I should have said "wasn't it the FTL", instead of "isn't". That and the HID, but HID has been removed from the Nano2g, so is no longer a problem.
20:04:00kugelI know the sim is limited to 44.1 (for no real reason IIUC) but being a sim it shouldn't use more of X just because it can
20:04:25kugelS_a_i_n_t: without HID it's "incomplete" :p
20:04:29gevaertskugel: I fully agree.
20:04:52S_a_i_n_tdo *all* the stable targets have HID ?
20:05:03S_a_i_n_t(legitimate question)
20:05:06gevaertsS_a_i_n_t: if he thinks it's stable, and nobody opposes that, he should promote it (or ask for help with promoting it)
20:05:12kugelS_a_i_n_t: nope
20:05:57funmanideally, before thursday ;)
20:06:10S_a_i_n_tWell, I don't think it's "rock solid", but it's no longer "unstable"...I'm a little afraid to cause a flamewar :/
20:06:17gevaertsfunman: that's different again :)
20:06:42S_a_i_n_tTheSeven seems to have fallen off the radar a little lately.
20:06:53S_a_i_n_tUni I think.
20:07:56gevaertsS_a_i_n_t: Is it "stable" with the rockbox bootloader? IIRC there was some discussion a few days ago about differences
20:08:14S_a_i_n_tI use the rockbox bootloader in mine.
20:08:19S_a_i_n_tAnd it's ANCIENT
20:08:55S_a_i_n_tthe GF uses iLoader in hers. I personally wouldn;t say that eaither is more stable than the other, iLoader just does more.
20:09:05gevaertsI (personally) don't actually care which bootloader people should use, but the rockbox one is the one that's documented in the manual
20:09:45S_a_i_n_tYes, if I was pushed...I'd probably have to lean toward iLoader, but only because of the boot-time options.
20:10:59kugelgevaerts: it turns out I can ignore most ifdef SIMULATOR actually
20:11:09 Quit flydutch (Quit: /* empty */)
20:11:46kugelthe majority is a) debugging, b) within HAVE_RECORDING (which RaaA won't have (initially)) or c) within #if CONFIG_CODEC == HWCODEC
20:12:06S_a_i_n_tbut, there is the benefit of "non-scary installation" withthe Rockbox Nano2G bootloader with it being NAND based and not NOR.
20:12:52S_a_i_n_tThere was *some* bricking risk in the early days, but not anymore as I understand it.
20:13:05gevaertskugel: right, that was my impression too. (a) is the only one that should still be there in the long term I think
20:13:52kugelstill quite a job to check all
20:15:28gevaertsS_a_i_n_t: I'd say find TheSeven the next time he's online. I really don't want to declare a target stable based only on hearsay (I don't have one, and since I've not been following it closely I don't know what issues there have been for various people, how serious or general they were, and if they were resolved), and I suspect that other non-owner devs will feel the same
20:15:50S_a_i_n_tOh, I agree totally.
20:19:53funmanranma: ah on the forum the crash report for fuzev1 says '3 crashes in 2 hours'
20:21:02funmana difference between as3525v1 & v2 is the CPU: v2 run boosted
20:22:38 Join saratoga_ [0] (~463f90ed@gateway/web/freenode/x-lvhzqjdwsmmexgit)
20:23:31saratoga_does the thumb patch simply ignore files with ASM blocks in them?
20:24:52funmansaratoga_: i posted 2 gcc frontends, the first runs preprocessor, remove unused static inline functions, and then just grep for 'asm' -> if there is no match the file is built with -mthumb, else with default (no thumb)
20:25:14funmanthe second one (both are .py scripts) just tries to build with -mthumb, and falls back to default only if it fails
20:25:34saratoga_funman: then thats why the codecs don't get slower
20:25:38funmanso if the asm is valid thumb assembler it will be built with thumb
20:25:40saratoga_they'll all have an ASM block
20:25:54funmanbut they use rockbox functions anyway?
20:25:55saratoga_for the fixed multiply code
20:26:13funmanand perhaps there is some c files in codecs which don't have asm
20:26:34saratoga_things like bitsream.c/h on the ffmpeg based codecs probably get smaller
20:26:50 Quit solexx (Quit: leaving)
20:27:10saratoga_but anything that does multiplication will have an ASM block, so all the core of each codec, if not all the helper functions
20:27:16funmanin the 2nd script youcan uncomment the secodn to last line so it shows files not built with thumb on stderr
20:27:31saratoga_do any codecs get much smaller?
20:27:49funmandidn't check any .rock or .codec
20:28:39saratoga_might be worth comparing to
20:28:47saratoga_but i doubt its a big deal
20:29:02saratoga_most codecs are pretty small anyway, and the space is largely data tables
20:29:06CIA-5New commit by 03Buschel (r26413): Submit FS #11240 by Raphael Jakse. Allows to reduce volume on WM8985 to -89 dB (e.g. used for Cowon D2). Below -57 dB the line out is affected. The ...
20:33:38S_a_i_n_tso, who can tell me the benefit of using arm-elf-eabi as opposed to arm-elf?
20:34:00S_a_i_n_tso far, all I see is a binsize decrease. build time seems to be extended.
20:34:12CIA-5New commit by 03Buschel (r26414): Fix FS #9193. Remove recording source <Mic> for iPod Video and iPod nano 1G. The manual already describes the correct behaviour.
20:34:59saratoga_S_a_i_n_t: bin size
20:35:46S_a_i_n_tOh, is that it? I guess with the drama I went through to build the toolchain I was expecting it to wash my dishes and do the vaccuuming ;)
20:36:24kugelS_a_i_n_t: some codecs run faster as well it makes it possible to remove a hack which adresses gcc's inability to make proper calls of static functions across memory sections
20:38:17S_a_i_n_tAwesome, if binsize was the only advantage I'd have a hard time seeing the point.
20:38:32S_a_i_n_tI'm glad there's more to it.
20:42:06funmanif you know how to alter shell $(()) precision please tell ^^
20:42:43saratoga_looks about right
20:43:34saratoga_maybe its worth just disabling thumb for the codecs so the compile goes faster
20:47:24funmansaratoga_: (edited)
20:47:56funmanbig improvements are only for small codecs
20:53:05funmanit would be nice if there was a #pragma -mno-thumb or something like that
20:53:16 Quit leavittx (Ping timeout: 258 seconds)
21:00:43 Join leavittx [0] (~leavittx@
21:02:37saratoga_the codecs that improve are mostly the PCM ones, since they don't really do any processing, they're just bitstream parsing and a memcpy
21:04:12 Quit leavittx (Read error: Connection reset by peer)
21:05:52 Join n17ikh [0] (
21:10:15funmanperhaps we can get better results if we separate asm functions from C functions
21:11:54funmanbuildtime for clipv1 using eabi: 220s 1st run, then 50s with ccache. thumb: 275s 1st run, then 110s with ccache
21:12:08kugeldoesn't the codeclib also have inline asm in .h files?
21:12:48saratoga_you're thinking in the core or in codecs?
21:13:02 Join efyx [0] (
21:13:04funmanyeah all the mul
21:13:49saratoga_i wouldn't worry about codecs anyway
21:14:25funmanso we can reduce binsize but it's only an improvement on targets with small buffer
21:14:35funmane.g. clipv1/c200v2/m200v4
21:14:43saratoga_you mean the codec buffer size?
21:14:44 Quit bieber (Ping timeout: 258 seconds)
21:14:47funmanno just rockbox
21:14:57 Join bieber [0] (
21:15:08saratoga_yeah its worthwhile for the main binary, but for codecs I think the possible gains are quite small
21:15:14funmanape is one of the biggest and it's only down by 1%
21:15:22ssorgatemin, what does the "changed source" column mean?
21:15:44ssorgatemand how can that be fixed?
21:15:53funmansaratoga_: it reduces of ~100kB but over 8MB that's useless
21:16:29funmanssorgatem: the string has been translated in the past, but the original english has changed since
21:16:59kugelssorgatem: the source string of string in a .lang file doesn't match with the corresponding source string in english.lang
21:17:03funmanwell perhaps not useless: you can load 100kB bigger files in test_codec ;)
21:17:14saratoga_the biggest codecs are wma and aac, and they're so big mostly because of enormous lookup tables
21:17:35saratoga_WMA has a 64KB random number table because I was too lazy to figure out how to shrink it :)
21:18:03ssorgatemmm, also, if I try the automatic code cleanup tool
21:18:07ssorgatemit gives my 404
21:18:08funman:p which one?
21:18:29ssorgatemare there others?
21:19:06funmanssorgatem: this was for saratoga_, not for you
21:19:27saratoga_but the size of the codec buffer is determined by AAC and Vorbis due to their malloc
21:19:39saratoga_so shrinking it doesn't accomplish anything
21:20:02funmanthat reminds me, still no mail on tremor ML
21:21:27saratoga_yeah i need to bug them
21:21:38saratoga_also to compute the RMS error between the new and old versions
21:21:40 Quit esperegu (Read error: Connection reset by peer)
21:21:44saratoga_and check that 96khz files still work
21:22:30CIA-5New commit by 03funman (r26415): Fix LANG_TAGNAVI_UNTAGGED source string
21:22:30 Join wodz [0] (
21:24:09 Join Casainho [0] (
21:24:30wodzamiconn: ?
21:24:49Casainhohello :-)
21:25:13ssorgatemso, how can I fix the "changed source" thing?
21:25:27Casainhodoes anyone here knows the "sb" format for image files of SigmaTel SoC? there are a few targets where Rockbox run with this SoC...
21:26:47 Quit efyx (Remote host closed the connection)
21:27:17CIA-5New commit by 03funman (r26416): french translation update
21:27:46Casainhobecause we at Lyre, we started using i.MX233 ARM9 from FreeScale that seems to be an reincarnation of Sigmatel SoC, and that uses encrypted firmware images
21:29:11 Join efyx [0] (
21:30:59funmanCasainho: if you have a question just ask it
21:32:12saratoga_i don't htink much is known
21:32:18saratoga_we've never had a port to one of their devices
21:32:39Casainhomy question is: does anyone knows if on Rockbox sources are code to build a 'sb' type image, from an elf file?
21:32:57funmanno there's not
21:33:23funmanthere's only code for TCC and PP i think
21:33:28Casainhoimx233 is of type STMP36xx
21:33:38saratoga_are you sure?
21:33:41saratoga_i don't think it is
21:35:06saratoga_i think freescale has reference manuals for all the MX series, maybe they're a better place to start
21:35:52Casainhowhen I run the 'elftosb2 -v' on Linux, I get: elftosb 2.2.1
21:35:54CasainhoCopyright (c) 2004-2007 SigmaTel, Inc. All rights reserved.
21:36:32Casainhothis forum message on FreeScale, talks about this:
21:36:44CasainhoFreeScale bought SigmaTel
21:37:47Casainhofor imx233, the firmware image must be encrypted! and we have now a blink LED code... but from factory, keys are 0
21:38:31Casainhoand we have the Linux 'elftosb2' executable that encrypts the elf files, but we would like to have the sources for 'elftosb'
21:39:02Casainhoso, maybe Rockbox had some code to build that encrypted firmware images, for STMP36xx
21:39:37Casainhoanyone knows how is the Rockbox bootloader installed on STMP36xx devices?
21:40:09saratoga_can you ask them for the sb format description?
21:40:14kugelCasainho: we don't have any code for this gsoc
21:40:25saratoga_they seem fairly open about this, maybe they have a document
21:40:34 Quit saratoga_ (Quit: Page closed)
21:42:12Casainhokugel: you mean, for STMP36xx or i.MX233?
21:42:46kugelCasainho: both. you know where to look for what we actually have
21:42:54ssorgatemI updated catalan language translation:
21:43:09Casainhokugel: ok, thanks.
21:44:19kugelfunman: don't forget to update the branch :)
21:44:38funmanoops right, another reason to clone rockbox with branches!
21:47:40 Join merbanan [0] (
21:52:36Casainhohere some pictures and details of current state of our Open Hardware board, now blinking a LED:
21:52:50*S_a_i_n_t notices something new at compile time "" and ""...have I gone insane?
21:52:58S_a_i_n_tthese *are* new, yes? is rather old.
21:53:38bluebrotherit's been around since quite a while (earlier called
21:55:46funmani added a few days ago but i didn't add anything to hide it from make output
21:56:40S_a_i_n_taha...I'm quite accustomed to spotting irregularities at compile time, it just looked odd to me seeing such a long line.
22:02:55***Saving seen data "./dancer.seen"
22:04:29 Quit Buschel (Ping timeout: 276 seconds)
22:11:47funmansaratoga: +586333.33% realtime
22:11:47funman+0.04MHz needed for realtime
22:12:34funmani used a normal build and only replaced rockbox.sansa with thumb built, but some codecs didn't play well
22:12:56 Join anewuser [0] (anewuser@unaffiliated/anewuser)
22:16:34funmanaac is between 0.92% and 1.88% slower, cook bugs, mpc 0.09(!)% slower, wma 0.37(!)% faster, mp3 is the same, mpc 0.09(!)% slower
22:16:54 Nick YPSY is now known as Ypsy (
22:18:46 Join DataGhost_ [0] (
22:18:46 Quit DataGhost (Disconnected by services)
22:18:47 Nick DataGhost_ is now known as DataGhost (
22:21:34funmanis it worth profiling to see which codecs api functions are slowed/enhanced ?
22:21:52kugelnot for 0.09% I'd say
22:22:25funmanalso to see where the 2% slower comes from
22:22:29funmanand what makes wma faster
22:24:42kugelgevaerts: I'm done with apps/ I think
22:26:28saratogamaybe huffman decoding is faster on WMA in thumb mode
22:26:53saratogathats basically just load, shift, store, etc
22:27:03 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
22:27:21 Quit elinenbe (Read error: Connection reset by peer)
22:27:27funmansaratoga: but huffman is in the codec, i only changed rockbox.sansa
22:27:45funmaneven codeclib is built in the .codec
22:28:00 Join elinenbe [0] (
22:28:35saratogaah sorry misunderstood
22:29:36gevaertskugel: Nice!. Now firmware?
22:30:09gevaertsThat shouldn't be too much either I think
22:30:20gevaertsmainly dircache and other file related things?
22:31:41kugelhaven't grepped yet
22:33:00kugelsome ifdefs are in files I'm sure we won't compile in RaaA
22:34:31kugelbacklight.c for instance
22:35:46 Quit lpereira (Quit: Leaving.)
22:37:44 Quit guymann (Ping timeout: 252 seconds)
22:38:15 Join guymann [0] (~charlie@
22:38:26gevaertskugel: actually, I think backlight.c should be able to lose all of the SIMULATOR ifdefs rather easily
22:39:56amiconnwodz: I'm working on optimising the greylib phase output routines; coldfire first
22:40:20amiconnFirst question - did you run test_fps.rock on the HD200 already?
22:40:37amiconnIf so, please post your results on
22:41:14amiconnSecond thing - I do have a patch for HD200 already... after getting aforementioned baseline, it needs testing (for correctness and speed)
22:41:36*amiconn is fiddling with arm in the meantime
22:41:49amiconnIt's all based on jhMikeS's idea
22:43:26wodzamiconn: HD200 scored ~146fps (greylib) in test_fps.rock (but it runs boosted). About testing - post patch somewhere if I have time I'll test it tomorrow
22:43:41amiconnI need all data, not just plain FPS
22:43:58wodzok - I don't have device at hand now
22:43:59amiconnIn this specific case the cpu load percentage is what's most important
22:44:46amiconn(because that is what the patch should reduce - it will also increase fps a bit if the isr load reduction is significant enough)
22:45:31amiconnOn iriver H1x0, isr load is reduced from 41% to 37% unboosted, and from 19% to 17% boosted
22:48:14 Quit S_a_i_n_t (Read error: No route to host)
22:50:02amiconnHmm, the latter I obviously remember wrong
22:50:19 Join S_a_i_n_t [0] (S_a_i_n_t@
22:53:54soapLlorean, does fit your description?
22:53:58funmanamiconn: which PP don't use lcd-as-memframe.S ?
22:54:34amiconnfunman: Many...
22:54:44funmanwodz: nice, you want me to test on clip?
22:54:50amiconnAfaik e200v1 is the only PP which does use it
22:54:57funmanamiconn: nice, e200v1 didn't boot when built with thumb
22:55:14funmanperhaps on other targets the lcd worsk before the undefined instruction
22:55:16amiconnI would expect such effects
22:55:33wodzfunman: fuze rather
22:55:57amiconnMany .S files return using ldm, which doesn't handle interworking on armv4t
22:56:29funmanamiconn: well supposedly the one in lcd-as-memframe.S was fixed in this build, many parts of the code don't handle interworking
22:56:51funmani'm checking them all and will post a patch on flyspray
22:57:31 Quit bmbl (Quit: Bye!)
23:00:27 Quit wodz (Quit: Leaving)
23:02:05ZambeziI have r18607. Well, it's working, but current stable build is r26416 so I think there's some improvements.
23:04:19bluebrotherwow, that's almost 2 years old
23:04:47 Quit S_a_i_n_t (Ping timeout: 240 seconds)
23:05:47Zambezibluebrother: I'm also consider to buy a Sandisk Sansa Clip+ now cause my iPod can't last forever and I can't go back to non-Rockbox.
23:06:25funmanamiconn: can you explain the 3rd line in thread.c::load_context() for ARM ? it was added in r14879 for dual core
23:06:45funman%0 points to start of thread context struct, which is the saved registers
23:07:12funmanldm %0 { r0, pc } will load saved r0 in r0 and saved r1 in pc <- why r1, and not [%0, #40] ?
23:07:27 Nick Ypsy is now known as YPSY (
23:08:44 Quit GeekShadow (Read error: Connection reset by peer)
23:08:50 Join Highlander_ [0] (
23:09:16funmane.g. why does it load thread_entry->context->r[1] in pc
23:09:39amiconnjhMikeS is the thread.c expert nowadays...
23:10:14funmanit's r4/r5 btw not r0/r1
23:12:03 Quit Highlander (Ping timeout: 240 seconds)
23:17:34funmanjhMikeS: ping ^
23:24:49jhMikeSfunman: yessir?
23:26:28funmancan you explain the 3rd line in thread.c::load_context() for ARM ? it was added in r14879 for dual core
23:28:48jhMikeSit may have been optimized a bit, but I don't think it was changed to support it. It goes to start_thread.
23:29:38jhMikeSstart_thread calls the real entry as defined in create_thread
23:31:31jhMikeSif there were an issue there, I doubt rockbox would run on any ARM...but, what is the concern?
23:32:06jhMikeSsomething with thumb code?
23:32:19funmanit's how i came to this yes
23:32:26funmanthe 3 first lines are equivalent to:
23:32:40funmanthread_entry *t = addr;
23:32:49 Quit petur (Quit: Zzzzz)
23:33:17funmanif (t->context->start) {
23:33:28funmanr0 = t->context->r[0];
23:33:33funmanpc = t->context->r[1];
23:36:06 Quit Casainho (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100423140709])
23:36:36jhMikeSfunman: yes, look at THREAD_STARTUP_INIT
23:36:52funmanaaah ok
23:37:07funmannow it makes sense ;)
23:37:49funman FS #11335 - Make ARM assembly functions thumb-friendly
23:39:48jhMikeSis there a speed gain from it?
23:39:49*funman slaps self - should test patches before posting them
23:40:11funmanjhMikeS: see the other task : FS #6734
23:40:55 Quit Topy44 (Quit: No Ping reply in 180 seconds.)
23:42:00 Join Topy44 [0] (
23:42:39kugelfunman: why "no effect on newer cpus"?
23:43:00funmansame number of cycles
23:43:16CIA-5New commit by 03bertrik (r26417): Update Samsung YP-S3 bootloader demo
23:43:40jhMikeSbut, it can reduce the amount of instruction cache missed, maybe
23:43:46kugelfunman: but there's 1 instruction more, why is that free?
23:44:05funmankugel: ldm doesn't take same number of cycles if you load pc or not
23:44:19funmanjhMikeS: wouldn't it increase it, if functions are bigger?
23:44:22kugelah ok, I didn't know that
23:44:53jhMikeSfunman: that's what I meant, thumb could help a bit, even if the instructions are still they same cycle count
23:45:19funmanjhMikeS: ah for thumb, yes but the effect is small on the codecs
23:45:34jhMikeSloopy code?
23:45:37funmanthe patch works btw
23:45:59kugelfunman: do you think the e200v1 build is fixed with the new patch?
23:46:00jhMikeSlots of repetitive execution in the codecs
23:46:25funmanmore like important code is not built in thumb
23:46:28funmankugel: dunno
23:46:39 Quit Topy44 (Client Quit)
23:47:37 Join Topy44 [0] (
23:48:27jhMikeSfunman: if anything is ARMv6, don't worry about changing the instructions. it always sets T if the pc is loaded
23:48:55funmanjhMikeS: yes, i have put that in the task description
23:49:03funmanARMv5 works too
23:49:28jhMikeSah, even ARMv5 does it right
23:50:03funmani'm not sure about mov pc, lr
23:50:18funmani've read that it works only on armv7
23:50:25jhMikeSprobably in the docs
23:50:35funmanbut we use that in nrv2e code for mkamsboot on armv4t and it works
23:51:09 Quit merbanan (Ping timeout: 248 seconds)
23:52:09 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
23:52:29jhMikeSarmv6 docs say you have to use 'bx lr'
23:52:45amiconnjhMikeS: I've found a way to apply your nice optimisation in lcd-as-gray.S so that it works on mini2g as well
23:52:55funmanmov pc, lr restores the spsr into cpsr
23:53:02amiconnNo speedup there because the serial lcd i/f is the limiting factor
23:53:22jhMikeSamiconn: does it speed up the other at least?
23:53:32amiconnNo slowdown though, and less binsize. Also, smaller ifdefed code blocks in the .S file
23:54:02amiconnYes. On my G2, it reduces ISR load (without GREY_ON_COP) from 33% to 30%
23:54:39 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
23:54:53jhMikeSnot bad I guess esp. if it's at worst not a loss for mini2g
23:56:05amiconnI implemented a similar trick on CF (H1x0, M5 and HD200 first), using mulu.l instead of all this or'ing
23:57:35funmankugel: (e200v1)
23:58:25 Quit bertrik (Quit: De groeten)

Previous day | Next day