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 2010-06-21

00:00:23AzureSkyk, just seemed weird
00:00:52stripwaxyep, it's normal
00:04:07stripwaxok, database seems sane again after reinitialising. pretty odd bug
00:07:04***Saving seen data "./dancer.seen"
00:07:54AzureSkyok i have a problem compiling
00:08:18funmanmake clean
00:08:48AzureSkysorry thanks
00:09:53pixelmastripwax: great chance to investigate ;)
00:13:03 Quit GeekShadow (Quit: The cake is a lie !)
00:14:09funmanwhat do you think of adding a (possibly #ifdef DEBUG) per-thread backtrace buffer?
00:14:51funmanwould be useful when an exception happens in memset/memcpy & friends
00:21:01TheSevenfunman: I already wondered if a perror()-like error buffer would make sense, that would be displayed on the panic screen, to catch problems that cause things like data aborts because of missing error handling
00:21:16TheSevenregarding a backtrace buffer: who should fill it?
00:21:52funmanTheSeven: so write this buffer each time we modify errno?
00:22:16funmanabout the buffer, i don't know but we could look at what glibc does, i suppose it's a gcc feature
00:22:21TheSevenbasically keep the most recent error message around in case the return code was just ignored
00:22:32TheSevenwe might just remove -fomit-frame-pointer and do stack unwinding
00:23:02funmanhow does unwinding work?
00:23:12funmani've seen unwind.o files in libgcc
00:23:51TheSevenif you have frame pointers, you can reconstruct the return addresses and call arguments from the stack without needing to know anything about the functions
00:24:21TheSevenhm, arguments only if they aren't passed in registers :-/
00:24:33amiconnThat comes at a cost which isn't adequate, imo
00:24:40funmanamiconn: which cost?
00:24:40TheSevenbut still, knowing the code path that lead to some crash might be helpful
00:24:56TheSevenamiconn: I don't really care about cost for debug builds :-)
00:24:58funmanhaving to remove -fomit-frame-pointers from my builds is adequate enough for me
00:25:19amiconnMore binsize and more register pressure
00:25:35amiconnFor debug builds it should be okay - if you get it to build
00:25:40TheSevenand a tiny bit more stack usage
00:26:18funmanlibgcc/unwind-c.o just needs abort(), but perhaps we can write a simple unwind
00:26:24amiconnSome asm blocks are already difficult to get to compile without frame pointer, because of the amount of registers they use
00:27:07funmanframe pointer reserves a register, right? is that ip ?
00:27:19amiconnNo, fp
00:27:36TheSeven12 iirc
00:27:49*TheSeven wonders what ip is a synonym for
00:27:54funmanip is r12 on arm
00:28:25TheSevenoh, frame pointer is r11
00:29:06 Quit ender` (Quit: A computer program will always do what you tell it to, and seldom what you want it to.)
00:29:21TheSevenr12 seems to be just a scratch register
00:29:45funmanIntra Procedure call scratch register
00:31:54 Part domonoky
00:32:33 Join stripwax_ [0] (
00:34:18AzureSkysaratoga's volume patch works quite well, but something strange, i updated svn then did a build and the build on my fuze didnt update(says same revision number) rather then r26998 it says r26979
00:34:31 Quit stripwax (Ping timeout: 260 seconds)
00:35:42funmanif you want the revision number to be changed, run 'rm version.*' befoer make
00:36:18 Quit DataGhost (Ping timeout: 240 seconds)
00:36:33 Quit pamaury (Remote host closed the connection)
00:36:33pixelmaI never do and my revision number usually changes correctly
00:36:41AzureSkyah thanks
00:38:06funmanpixelma: then there's a bug, version.h shouldn't be rebuilt
00:38:36pixelmawhat? After I update?
00:38:49funmanit should never be rebuilt, it has no dependancy
00:39:28pixelmathat doesn't make sense to me at all at the moment
00:39:43funmanwhat doesn't make sense?
00:40:51pixelmathat the revision number *shouldn't* change after I update and compile again
00:41:39funmanwell it's a known limitation
00:42:03gevaertsfunman: does that mean that the absence of "M" in the revision number no longer implies an unmodified build?
00:42:56pixelmahow should I know if I test the build I want to test (if I just need to compile, no removing version.h, make clean or reconfigure)?
00:43:12funmansvn info
00:43:59pixelmadoesn't help me at all on target if e.g. the installed and the rolo'ed build show me the same revision
00:44:19*gevaerts thinks that we need a mechanism to force version.h to be recreated when needed
00:44:31funmanmake it depend on every file in svn
00:44:34gevaertsOtherwise I don't really see what problem the file solves
00:44:41funmanit doesn't solve anything
00:44:43pixelma(e.g. a flashed build on Archos vs. a rolo'ed)
00:45:09funmanit just makes it easy to have the version number be correct: use "rm version.*; make"
00:45:26pixelmaeven if it works this way currently, I don't believe that is the intended way
00:46:45funmanpixelma: if you know an easy way to have version.h be rebuilt each time a file in svn is modified / svn up is run, please tell
00:47:13funmani proposed an unconditional Makefile rule but then make wouldn't output 'nothing to be rebuilt'
00:47:30gevaertsIs that a problem?
00:47:34 Quit stripwax_ (Quit:
00:47:43funmanfor the devs who objected, it seems
00:48:19gevaertsCurrently we risk incorrect version strings. I prefer make not printing "nothing to be rebuilt" to that...
00:48:49pixelmadoes it still only builds what needs to be rebuild (things that changed, in this case maybe only the version things)
00:49:03funmanif someone does it correctly, yes
00:49:31gevaertsfrom what I can see already handles this correctly, so the only thing that happens is that runs and does nothing
00:49:55pixelmasounds better than incorrect version numbers at a first glance
00:50:00gevaertsIf the version changes, of course there's a bit more that gets rebuilt
00:50:13 Join BHSPitMini [0] (
00:50:22funmangevaerts: hm right
00:50:34AzureSkyok tried to run "rm version.* and it says cannot remove 'version.*' : no such file or directory, so i need to run it from inside a specific folder?
00:50:38funmandidn't remember that
00:51:03funmangevaerts: s/a bit more/exactly what needs to be rebuilt/
00:51:12gevaertswell yes
00:51:26funmanAzureSky: in the build folder root
00:51:48AzureSkythats where i ran it from
00:53:04funmanTheSeven: you got any doc on unwinding handy?
00:53:14TheSevennope, sorry
00:53:49TheSevenbut basically you just follow the stack frames, which point to each other by their frame pointer
00:54:09amiconngevaerts: Imo it would be best if version.h is only actually rebuilt if the version or date changes
00:54:16funmani wonder how that tells where lr is in this frame
00:54:17amiconnI think that should be possible
00:54:35funmanamiconn: obviously
00:54:57funmanproblem is just Makefile deps
00:55:31funman: when do you check if it version needs change
00:55:58amiconnCan't you just run unconditionally, and then make it only actually rewrite version.h if the content will change?
00:56:14gevaertsyes, it should already do that
00:56:27gevaertsonly that bit seems buggy
00:56:51gevaertsApart from that firmware.make just needs a .PHONY line I think
00:57:02funmangevaerts: no, it's not run unconditionally
00:57:15gevaertsfunman: that's what you need the .PHONY for
00:57:17TheSevenIIRC LR is always immediately above FP in the stack frame
00:57:24funmanonly when a file depending on version.h has changed
00:57:33 Quit bertrik (Quit: De groeten)
00:57:51TheSeven(because FP is basically the parent's SP)
00:58:04funmanA phony target is one that is not really the name of a file. It is just a name for some commands to be executed when you make an explicit request.
00:58:28funmanTheSeven: but you can have any number of registers before lr ?
00:58:57gevaertssure, but the result should be the same
00:59:33funmanso each explicit rule + the implicit one should depend on this one?
00:59:35TheSevenIIRC a stack frame on ARM looks like this: func2scratch func2FP func2LR func2args func1scratch func1FP func1LR func1args ...
00:59:57funmangevaerts: IIRC i tried that but dependencies are calculated before running the commands so make doesn't see version.h has changed, but feel free to try
01:00:30amiconnA leaf function will not (always) put lr on stack
01:00:47TheSevenamiconn: it will, if it calls subfunctions
01:01:03amiconnThat's why I said leaf function
01:01:17amiconnIf a function calls another one, it isn't a leaf function
01:01:19gevaertshm, right
01:01:30TheSevendidn't spot that word :-)
01:01:30funmanwell it's possible to do it anyway
01:01:58TheSevencan one determine if LR is on the stack or still in the reg in a leaf function?
01:02:21TheSevenwill it always be on the stack if there are scratch words on the stack?
01:02:29amiconnSame applies to SH1 and its 'pr' register btw
01:03:02amiconn(actually in a more strict way, as pr is separate from the general purpose registers)
01:03:49funmanwhat's "pr" ?
01:04:35amiconnThe SH equivalent of arm's lr
01:04:49amiconnNot identical though, as it's not usable as a general register
01:05:03amiconn"procedure return"
01:05:32funmanah fp points to saved "fp, lr" it seems
01:06:26amiconnColdfire doesn't have an equivalent - it always stacks return addresses. Don't know about mips in that respect
01:07:03funmanit's r11 because you shouldn't push r12 (ip), r13 (sp), r15 (pc); so it's always followed by lr or nothing
01:08:11funmanperhaps you could try to pop the next word from the stack and if it's different from lr, assume that the real lr comes from the stack
01:08:52funmanbut that doesn't give you scratch regs content?
01:10:39funmanah i see argument registers are pushed too
01:13:46funman <-
01:14:25 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
01:16:01gevaertsfunman: I suspect I know a way, but I'm not sure yet
01:16:50funmanwe'd need to modify asm functions to save fp :/
01:17:17gevaertsThe trick I'm thinking of is to put the logic directly in a makefile, and have it write out a makefile fragment that directly echoes the version number to version.h, and then include that makefile fragment
01:18:11 Quit bluebroth3r (Ping timeout: 265 seconds)
01:20:09funmanlooks messy but let's see how that works
01:21:00gevaertsyes indeed :)
01:22:02 Join binaryhermit [0] (
01:22:29 Join Buckwheat [0] (~chatzilla@
01:22:30 Quit Buckwheat (Client Quit)
01:22:54 Join Buckwheat [0] (~chatzilla@
01:23:16 Quit Buckwheat (Client Quit)
01:24:36 Quit DerPapst (Read error: Connection reset by peer)
01:27:30funmanclipv2 builds fine without -fomit-frame-pointer
01:28:56 Join DerPapst [0] (
01:40:24funmanadding -O0 doesn't build
01:41:56funman*_irq() macros want mask to be a reg, not an immediate
01:43:00funmanmdct doesn't build ..
01:46:13funmanbut then, text_viewer does work fine Oo
01:50:04CIA-8New commit by 03gevaerts (r26999): Fix file replacement logic
01:51:41CIA-8r26999 build result: All green
01:53:51funmangevaerts: i think it is wrong
01:54:03gevaertsfunman: it's right, but not all to fix this yet
01:54:10funman% if diff /dev/null /dev/null; then echo 1; else echo 2; fi
01:54:23CIA-8New commit by 03gevaerts (r27000): Add some magic to force make to regenerate version.h properly when the revision changes
01:54:26funmanhmm oops i'm reading the other side
01:54:41gevaerts/dev/null is not readable, so diff will fail
01:54:51funmanit sure is
01:56:09CIA-8r27000 build result: 266 errors, 0 warnings (gevaerts committed)
01:56:18 Join thefirstM [0] (
01:57:21gevaertsHow did I do that?
01:57:29TheSevenwhat the hell
01:58:17funmanif you think rockbox is slow, try building with -O0 !
01:59:05funmangevaerts: version.h needs an argument
02:00:01gevaertsfunman: where?*
02:02:38CIA-8New commit by 03gevaerts (r27001): Tell where to look for the svn repository
02:03:17CIA-8New commit by 03funman (r27002): debug menu: avoid using constant variables in array initialization ...
02:04:05 Quit Dark_Rak3r (Ping timeout: 272 seconds)
02:04:25CIA-8r27001 build result: All green
02:06:17CIA-8r27002 build result: All green
02:07:06***Saving seen data "./dancer.seen"
02:11:20gevaertsfunman: is there actually a reason to autogenerate version.c? In my tests, generating version.h seems to be enough
02:11:53funmanthe symbol must be somewhere
02:12:12gevaertssure, but version.c can be in firmware/common
02:12:26gevaertsit doesn't matter much of course
02:13:37funmanranma: ping
02:13:38*TheSeven is fighting the last on-stack sector buffer
02:13:52gevaertsexcept that if version.h is newer than version.c, make keeps trying to regenerate version.c, which is a bit useless
02:14:15funmanfirst script created them both at the same time
02:15:26 Quit S_a_i_n_t (Read error: Connection reset by peer)
02:18:01 Join S_a_i_n_t [0] (S_a_i_n_t@
02:18:32CIA-8New commit by 03gevaerts (r27003): Don't autogenerate version.c. Just version.h is enough
02:18:44funmani still hear glitches when playing mp3 on clip+ and there's some failure reports on ABI
02:19:59CIA-8r27003 build result: All green
02:20:37gevaertsIf that last bit causes problems I didn't see, just revert it...
02:20:47funmanlooks ok
02:21:12funmani haven't used AMSv1 for some time so please tell if current builds still work fine or cause glitches
02:22:49 Quit CGL_party (Read error: Connection reset by peer)
02:23:11funmanranma: can you make text_viewer plugin crash on c200v2 and share your text_viewer.elf+rockbox.elf and memory dump? i couldn't see where the bug comes from except that the plugin works when building rockbox with non-eabi compiler or with -O0
02:23:20 Quit funman (Quit: free(random());)
02:23:55Unhelpfulperhaps we could report build results only when there are yellows/reds?
02:39:40 Quit Dhraakellian (Ping timeout: 264 seconds)
02:56:14 Join Dhraakellian [0] (
02:59:08TheSevenhm, that sectorbuffer-on-stack-free build seems to work, but it boots up dog slow
02:59:49 Quit kadoban (Read error: Connection reset by peer)
02:59:53TheSevenand the database still seems to have bad buffers somewhere
03:05:44 Join halmi [0] (
03:05:50 Quit halmi (Read error: Connection reset by peer)
03:09:09 Quit thefirstM (Remote host closed the connection)
03:09:58 Join kadoban [0] (
03:25:47 Quit efyx (Remote host closed the connection)
03:36:21S_a_i_n_tfunman, pixelma: S_a_i_n_t would like to point out that his revision number changes correctly after updates.
03:36:42S_a_i_n_treading that this is a bug I find quite odd, that it is intentional, plain silly.
03:42:30 Quit BHSPitMini (Quit: Ex-Chat)
03:47:33 Quit steve|m (Ping timeout: 276 seconds)
03:54:47 Quit detaos (Ping timeout: 265 seconds)
03:57:08S_a_i_n_ts/is/was/ s/find/found/ backlog reveals I'm now talking inpast tense. Great, thanks gevaerts.
04:01:07 Join steve|m [0] (
04:02:49 Join fdinel [0] (
04:07:09***Saving seen data "./dancer.seen"
04:09:50 Join thefirstM [0] (
04:13:39 Join Barahir_ [0] (
04:15:03 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
04:16:12 Join retroj [0] (~retroj@unaffiliated/retroj)
04:16:28thefirstMI am trying to build Rockbox with GCC 4.5 on ipod nano2g, but for some reason playing music at the WPS doesn't work. I get an "Undefined instruction at E5958718" error. How might I figure out what source code corresponds to that memory address?
04:16:54retrojmorning. just popping by to report broken links for ipodvideo manuals at
04:17:23 Quit Barahir (Ping timeout: 276 seconds)
04:24:54S_a_i_n_tthefirstM: "" should tell you what that address is IIUC
04:26:00S_a_i_n_thave a look in the .dir you built in, it'll be in there unless you cleaned up afterward.
04:26:15thefirstMIts still there. I am checking now.
04:28:55thefirstMThat's weird. I can't find it in there. There is a bunch of stuff starting with 0x08 and few things with 0x00, but nothing in the range I am looking for.
04:29:21 Join saratoga [0] (9803c6dd@gateway/web/freenode/ip.
04:29:32thefirstMWhat base is the number from the error message in?
04:29:35S_a_i_n_tI was just looking at mine, it doesn't resemble any value in the .map file.
04:29:49saratogathats pretty common actually, since one of the normal ways to get an undefined instruction is to load from something thats not memory :)
04:30:04thefirstMOh. Duh!
04:30:32saratogaprobably means the binary contained nonsense
04:30:48saratogaunfortunately, thats going to make tracking down your issue fairly hard
04:31:40thefirstMMost of it (everything except for that problem) seems to work just fine, including the plugins.
04:32:41 Join DerPapst1 [0] (
04:33:20saratogabut audio playback doesn't work at all?
04:33:43thefirstMNo, outside of the WPS, audio works just fine. (I tried mpegplayer and Doom)
04:34:22saratogawell that narrows the problem down to about 2/3 of all rockbox :)
04:34:39 Quit amiconn (Disconnected by services)
04:34:41 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:34:45 Quit DerPapst (Ping timeout: 272 seconds)
04:34:59 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:35:06JdGordonthefirstM: which theme are you using?
04:35:20thefirstMCabbiev2 (the default)
04:35:22JdGordonvery possibly it failed to load the theme and couldnt fallback on the default
04:35:30 Quit pixelma (Disconnected by services)
04:35:33 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:35:38thefirstMShould I try a different theme?
04:35:48 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:36:24saratogatry wav decoding
04:37:04thefirstMTrying a different theme does not work, I still get the same error at the same address. .wav files also produce the same error.
04:40:49thefirstMI do see just a flash of the WPS before I get the error. However, no audio is played before the crash.
04:41:24S_a_i_n_tsounds like it is failing to load the default WPS and the user selected one, then quitting.
04:41:57thefirstMIs there any way I can try playing the music without the WPS?
04:42:06 Quit Strife89 (Quit: Leaving)
04:42:17S_a_i_n_t...though, if that were the case you shouldn't even see a "flash" of it.
04:42:17thefirstMI tried loading it (an MP3 file) into mpegplayer, but that didn't work.
04:42:31S_a_i_n_tand, no, there isn't IIUC
04:43:42 Join Strife89 [0] (
04:45:34thefirstMAre you guys planning to switch to GCC 4.5 at any point?
04:46:26Unhelpfuli think some folks are starting to experiment with it for link-time optimization. we only *just* switched ARM to 4.4 so we could use EABI ;)
04:47:02Strife89Unhelpful: Oh? What revision?
04:47:20thefirstMYeah, I saw a conversation on here about that the other day. I guess they aren't having undefined instruction problems.
04:47:21Strife89Sounds like I've been away from my compiler for too long.
04:49:34JdGordonthefirstM: oh, you''re not using the recommended compiler? that could very well explain it
04:49:46*Unhelpful points to r26769: "Make EABI gcc the default for ARM platforms"
04:50:25thefirstMJdGordon: Sorry, I thought you had read that part.
04:50:27Strife89Unhelpful: Yeah, it's definitely been a while.
04:51:23thefirstMIs it possible to create a memory dump from rockbox?
04:51:33Strife89Unhelpful: Call me a noob, but I suppose I need to install a new cross-compiler to use it?
04:51:52thefirstMStrife89: Yes
04:51:56 Quit panni_ (Read error: Connection reset by peer)
04:52:23thefirstMJust press "e" after executing
04:53:06S_a_i_n_tand clear out existing cross-compilers first.
04:53:12S_a_i_n_tOr expect weird errors.
04:53:29thefirstMOh, yes. I shot myself in the foot with that. Several times.
04:53:42UnhelpfulS_a_i_n_t: um, why should he need to do that? the eabi toolchain has a different executable prefix...
04:54:04S_a_i_n_tUnhelpful: I had to, several others had to also.
04:54:10S_a_i_n_tWhy? I can't say.
04:54:17S_a_i_n_tbut it cuases problems.
04:54:34 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
04:55:11Unhelpfulworst case you should only have to clear them from your path :P
04:56:04S_a_i_n_tthings don't always seem to make sense though ;)
04:56:07 Quit TheSeven (Ping timeout: 240 seconds)
04:56:24S_a_i_n_tMy eabi install just wouldn;t work with the existing toolchain in place.
04:56:45S_a_i_n_tcleared it, bingo.
04:56:46*Unhelpful will simply assume this to mean that you screwed something up ;)
04:57:03S_a_i_n_tI've seen similar comments here also.
04:57:18S_a_i_n_tsomething regarding "as" I didn't really parse too well.
04:57:28 Quit MethoS- (Read error: Connection reset by peer)
04:59:09 Join fdinel [0] (
04:59:33saratogafor what its worth i did not have to clear out the cross compilers before building EABI just worked for me on ubuntu
05:01:44 Quit kadoban (Remote host closed the connection)
05:01:47Unhelpfulsee! ;)
05:02:10 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
05:03:36 Join JdGordon| [0] (~jonno@rockbox/developer/JdGordon)
05:04:56 Join kadoban [0] (
05:05:34 Quit Rob2222 (Quit: Rob2222)
05:08:14saratogawe should really do a new rbutil release so that theme installing works again
05:08:39*S_a_i_n_t suggests that missing currupted thems get fixed up too.
05:08:53S_a_i_n_tbut, tha doesn't seem to be much a priority
05:16:28 Quit JdGordon| (Quit: leaving)
05:17:06 Join JdGordon| [0] (~jonno@rockbox/developer/JdGordon)
05:17:29 Quit JdGordon| (Client Quit)
05:25:23saratogadon't you spend a lot of time looking at wps code? maybe figuring out what the converter couldn't handle would help
05:26:57S_a_i_n_tIt seems to have skipped a few purely at random, and I've noticed it doesn't like coordinates after the backdrop %Xd line...but other than that I've noticed no real "pattern" to what themes were missed.
05:27:37S_a_i_n_tand a few seem to have has a % added to the front of every character in the WPS. :/
05:28:37S_a_i_n_tI managed to contactthe author of one particularly affected theme as contacting themesite admins was beginning to get painful.
05:29:59S_a_i_n_tOh, sublines barf on some also, the thing is I know of at least one error that escapes checkwps, so running the conversion globally from the old themesite would perhaps be a good thing to try.
05:30:18S_a_i_n_tThe skinupdater recieved at least one fix since it was run initially.
05:31:20S_a_i_n_t"%Xd(backdropimage) 0| 0|" seems to pass checkwps :/ but fails on device.
05:32:35saratogado the ones it skips update if you manually run the update script on them?
05:34:19S_a_i_n_tMine did at least. It skipped two of my themes, then when I ran the skinupdater it outputted perfect syntax.
05:34:32S_a_i_n_tI need to make a list of all that failed, but it's a daunting task.
05:34:48S_a_i_n_tthere's also no way of knowing which ones "passed", but are actually broken.
05:35:00S_a_i_n_t(other than trial and error)
05:37:11 Quit Horscht (Quit: Verlassend)
05:54:41 Quit kramer3d_ (Quit: Leaving)
05:55:57S_a_i_n_tn, o, p?
05:56:17AzureSkysorry was trying to type make clean into cygwin
05:56:29 Join jennifur [0] (
05:59:51 Quit binaryhermit (Ping timeout: 260 seconds)
06:07:11***Saving seen data "./dancer.seen"
06:13:16 Quit whydoubt (Ping timeout: 265 seconds)
06:14:03 Quit BHSPitMonkey (Quit: Ex-Chat)
06:20:28 Join will [0] (~chatzilla@
06:20:51willdoes the mpeg player typically cause a lot more battery drain than your average audio file?
06:20:59 Join whydoubt [0] (
06:24:22JdGordonwill: presumably yes
06:24:35JdGordonbieber: OK, I'm making a new theme in the editor and loving it so far :)
06:24:46JdGordoneven just for the viewport positioning helper
06:24:52 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
06:25:10bieberOh, awesome :)
06:25:22bieberIt should be getting a lot more helpful this week
06:26:46JdGordonIt would be nice if the UI area was drawn if you are in a sbs
06:27:00JdGordonusing a flat colour like 255,0,255 would be good even
06:27:33 Quit will (Read error: Connection reset by peer)
06:30:00JdGordonarg, not having the screen magically resize is annoying :p
06:31:29JdGordonwhat are the magic settings? #screenwidth?
06:33:33JdGordonA, it doesnt like a space between the # and setting name
06:37:56JdGordonbieber: is it possible to add tag parameter hints? i.e if you start typeing %T( it will tell you the expected params?
06:38:17bieberOoh, that would be nice
06:38:27bieberI'll have to give it a go once I've gotten some more of the rendering done
06:42:51JdGordonalso drawing touch regions maybe
06:43:01*JdGordon loves adding to other peoples todo lists :)
06:45:33JdGordonbuttons in the sbs work :)
06:45:40JdGordonDAMN that was simple :)
06:46:35JdGordon my uber awesome sbs
06:59:11 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
07:08:09 Quit shai (Quit: Leaving)
07:11:58 Quit whydoubt (Ping timeout: 240 seconds)
07:13:34*JdGordon finds a really odd bug in the settings list screen
07:13:41 Join shai [0] (
07:14:13JdGordonno, all lists when using touchscreen regions in the sbs
07:15:10S_a_i_n_tbug == ?
07:16:18S_a_i_n_tthose aren't the arrows I made, surely.
07:17:16JdGordonno, these are easier to test with :)
07:17:21JdGordonI dont need good looking to test
07:17:23JdGordonmspaint ftw
07:17:24S_a_i_n_tno, no they're not. S_a_i_n_t breathes a sigh of relief.
07:17:43S_a_i_n_tI made mine with mspaint I'll have you know!
07:17:49S_a_i_n_tand mine are pretty! :P
07:18:41 Quit storm` (Ping timeout: 265 seconds)
07:18:51JdGordonit appears the current list handling touchscreen code is braindead!
07:19:16JdGordonand by that I mean I'm using it wrong :p
07:19:24 Join hebz0rl [0] (
07:19:27JdGordonbut I cant say that because IIRC it is mostly my code anyway :D
07:19:41 Join whydoubt [0] (
07:19:53S_a_i_n_tots possible to use your own implementation wrongly, I do it all the time :P
07:19:59S_a_i_n_t*its even
07:23:30 Join storm` [0] (
07:28:09Mode"#rockbox +o pixelma" by ChanServ (ChanServ@services.)
07:29:15Topic"Daily manuals currently available at | Rockbox 3.6 now released! | Please read before speaking: | Please direct offtopic/social chat to #rockbox-community | This channel is logged at" by pixelma (quassel@rockbox/staff/pixelma)
07:29:39Mode"#rockbox -o pixelma" by ChanServ (ChanServ@services.)
07:30:16AzureSkygot a compile error, any advice to fix it would be helpfull
07:38:35*S_a_i_n_t wonders if JdGordon or bieber or anyone really would be kind enough to compile an SVN themeeditor for Win.
07:38:48S_a_i_n_tQmake and I are having some issues still :/
07:39:09JdGordonyou probably need to ping bluebrother.... I havnt got the cross compiling stuff giong
07:39:38S_a_i_n_taha, I do, but QT has paked a massive sad one me :/
07:39:46S_a_i_n_t*packed even
07:40:29AzureSkya nice theme editor would be great, and any clue how to fix that compile error i have?
07:40:41S_a_i_n_tmake clean?
07:41:00S_a_i_n_tif it spits again, then I'll have a serious think about it.
07:41:00AzureSkytried that, i can try it again(3rd time)
07:41:10S_a_i_n_t"make veryclean"
07:41:48AzureSkystrange, even redownloaded the svn clean >.<
07:42:09S_a_i_n_tthat won't matter if the build directory is still "dirty"
07:42:31JdGordonS_a_i_n_t: if you find cross compiler instructions I might be able to get you going
07:42:46S_a_i_n_tand, fwiw, if you want to return the checkout to an SVN state just use "svn revert"
07:42:53S_a_i_n_tsave yourself some bandwidth.
07:43:09S_a_i_n_tAzureSky: ^
07:43:25AzureSkythx :D
07:43:38S_a_i_n_tAll these commands should be available with "svn −−help" which I'm fairly sure I've mentioned.
07:44:34S_a_i_n_tI also have a line that will delete all non-svn entries...but it's pretty brutal.
07:45:07S_a_i_n_tIt really should move them to a dir, but instead it just uses "svn stat", and deletes anything not in that list :P
07:45:12S_a_i_n_tie. brutal ;)
07:45:46S_a_i_n_t(but good for cleaning up after lots of patches etc.)
07:45:53AzureSkythe only patch i want to use is saratoga's volume patch
07:46:14S_a_i_n_tyou'll find others when you get more confident, trust me.
07:46:28S_a_i_n_tI have about 20 or so in my builds.
07:46:41AzureSkydidnt see any for the fuze that looked worth my time honestly
07:46:58S_a_i_n_tMost aren't specific to a target.
07:47:19saratogamost useful patches end up in SVN pretty quickly
07:47:36*S_a_i_n_t awaits the pong AI inclusion then :P
07:47:41AzureSkyyour volume patch helps with some of my phones :D
07:48:05S_a_i_n_tFS #5855 == pure awesome.
07:48:37S_a_i_n_tFS #6697 should also replace the current "dict" IMO.
07:48:48S_a_i_n_tthat dictionary is braindead.
07:51:39S_a_i_n_tFS #6697 actually has scrolling, and fuzzy searching, and is just still needs a little work (which I'm getting through slowly) but it is vastly superior to dict.rock
07:52:02pixelmalast things I heard about stardict is that it has problems with scrolling etc. and is a bit overkill with typesetting. I admit I haven't tried myself but couldn't find motivation to do it, reading about the former and having rather "small screen targets"
07:53:13S_a_i_n_tpixelma: If you care to try it, I'd appreciate it. I have "small screen" playrs also. (well, 176x132 is fairly small to me) and it wraps the text fine.
07:53:18S_a_i_n_tscrolling is good too.
07:53:51pixelmaahem... the target with the biggest screen I have is 160x128...
07:54:24S_a_i_n_tstill, I suspect that shouldn't be a problem.
07:54:28pixelmaor better: its screen size is 160x128 greyscale ;)
07:54:58pixelmaand how does the thing work on greyscale and monochromme, have you per chance tested in a sim?
07:55:11S_a_i_n_tThe desc's in the dict I use are formatted quite sanely, and text wrapping works very well in this plugin.
07:55:34S_a_i_n_tI haven't tested for greyscale I'll admit, though I can do so tonight.
07:55:38S_a_i_n_t(a few hours)
07:56:38S_a_i_n_tThe thing I find awesome about this is that the search makes it very easy to use premade dictionaries for language translation.
07:57:04S_a_i_n_tdict.rock not scrolling is a real PITA
07:57:19S_a_i_n_tespecially for large definitions
07:58:03AzureSkysame error, giving it one more go after another make veryclean
07:58:28S_a_i_n_tIf it keeps doing it, I suspect all the make clean/verycleans in the world won't fix it.
07:58:35S_a_i_n_tthe *exact* same error?
08:01:05 Join stoffel [0] (
08:02:47AzureSkyyeah it was
08:03:10S_a_i_n_tand, it has compiled properly before, with the same exact setup?
08:03:17S_a_i_n_tor, was this a first time compile?
08:04:09 Join kaylinsigswort [0] (
08:04:14AzureSkyyeah same one, now im seeing others for some reasion....bah, guess i try your hardcore command
08:04:22CIA-8New commit by 03jdgordon (r27004): Touchregion support for the Base Skin and FM Skins. display obviously needs to be in stylus mode for this to work. Just about all screens should be ...
08:04:37JdGordon^ that should make the D2 far more usable :)
08:04:53S_a_i_n_tAzureSky: the thing I was talking about will only delete files not normally in SVN.
08:04:58S_a_i_n_tnot much use here.
08:05:11S_a_i_n_tyou said it is a "clean" checkout though right?
08:05:28S_a_i_n_twhat does "svn stat" report?
08:05:49AzureSkyit removed the volume patch ad some other files
08:06:17CIA-8r27004 build result: All green
08:06:21AzureSkyuclpack.exe, codepages.exe, convbdf.exe and others
08:06:51S_a_i_n_t"svn revert" doesn't "remove" anything, it just returns the source to an SVN state
08:07:12JdGordonwhich is damn annoying when new files are added
08:07:15***Saving seen data "./dancer.seen"
08:07:31S_a_i_n_tconvbdf shouldn't have been effected by saratoga's patch.
08:07:39AzureSkyah, well it found some files it changed them from what im seeing
08:07:58S_a_i_n_tyou've installed something else I take it, some other patch?
08:08:03AzureSkywill give it a test again now
08:08:09AzureSkynope one patch
08:08:13S_a_i_n_tpastebin the output of "svn stat"
08:08:19AzureSkyjust the volume pne
08:08:37S_a_i_n_twhat is the FS# of saratoga's patch.
08:08:44AzureSkyhow am i to copy/paste the content of this window
08:08:57S_a_i_n_tright click the cygwin icon.
08:09:02bluebrotherS_a_i_n_t: just did one yesterday:
08:09:04S_a_i_n_tit'll give option
08:09:06AzureSkyFS #11364
08:09:39bluebrotherand I'm building native, wasn't too motivated to figure how to cross compile Qt.
08:09:49bluebrotherVMs are nice for that.
08:10:52bluebrotherbieber: there are a few new warnings in the theme editor you might want to give a look. Especially that "no return in non-void function". Such things caused crashes for me in C++ code in the past.
08:11:00AzureSkybest i can offer i cant select the text to copy
08:11:07bluebrother(no idea if that would occur in this case)
08:11:26S_a_i_n_tAzureSky: You've definitely not had a clean checkout...
08:11:34bluebrotherAzureSky: click the window title icon, then Edit / Mark
08:11:40S_a_i_n_tthat patch touches none of the files you mentioned.
08:12:17AzureSkyyeah, so im going to test now after that svn stat see if it helped
08:12:30S_a_i_n_tsvn stat changes nothing.
08:12:35S_a_i_n_tit's just an output.
08:12:49S_a_i_n_tif you used svn revert, it should be in a "clean" state
08:13:24AzureSkyBain2k9@Bain2k9-PC ~/rockbox
08:13:25AzureSky$ svn stat
08:13:25AzureSky? vol1.patch
08:13:25DBUGEnqueued KICK AzureSky
08:13:25AzureSky? Build
08:13:25AzureSky? trunk
08:13:25***Alert Mode level 1
08:13:25AzureSky? tools/uclpack.exe
08:13:25***Alert Mode level 2
08:13:25AzureSky? tools/codepages.exe
08:13:26***Alert Mode level 3
08:13:26AzureSky? tools/convbdf.exe
08:13:26***Alert Mode level 4
08:13:26AzureSky? tools/bmp2rb.exe
08:13:27***Alert Mode level 5
08:13:27AzureSky? tools/scramble.exe
08:13:27***Alert Mode level 6
08:13:27AzureSky? tools/rdf2binary.exe
08:13:28***Alert Mode level 7
08:13:28AzureSky? tools/mkboot.exe
08:13:28***Alert Mode level 8
08:13:28AzureSkyM firmware/drivers/audio/as3514.c
08:14:01AzureSkyBain2k9@Bain2k9-PC ~/rockbox
08:14:01AzureSky$ svn revert
08:14:01AzureSkysvn: Try 'svn help' for more info
08:14:01***Alert Mode level 9
08:14:01AzureSkysvn: Not enough arguments provided
08:14:09bluebrotherAzureSky: please use a pastebin for that.
08:14:25AzureSkyhelp doesnt give me arguements that i see for revert.....
08:14:48S_a_i_n_t"svn revert"
08:18:50S_a_i_n_tuse "svn rever -R"
08:19:00CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
08:19:00*S_a_i_n_t forgot he uses aliases for this.
08:19:14S_a_i_n_t"svn revert -R"
08:19:55S_a_i_n_tIf it just suddenly "broke", my first thought is a local change.
08:20:42S_a_i_n_tif it compiled before, and nothing changed, it should compile again. this is simple logic.
08:21:30pixelmabluebrother: just in case you didn't notice and for your info - I put the link to your manuals into the topic so that some are available (that's how I understood you yesterday)
08:22:37CIA-8New commit by 03ranma (r27005): mps depends on endpoint, I wonder why gcc didn't warn about i being used uninitialized...
08:22:57S_a_i_n_tAzureSky: Sorry man, I'm on a roll it seems.
08:23:03S_a_i_n_tI forgot the "."
08:23:17S_a_i_n_t"svn revert -R ."
08:23:44S_a_i_n_tthough, the svn stat output looked perfectly normal.
08:24:02***Alert Mode OFF
08:24:12S_a_i_n_tI just don't see why it would suddenly not work, if nothing in the environment changed.
08:24:14CIA-8r27005 build result: All green
08:25:15bluebrotherpixelma: yeah, that was the idea. And I've seen the topic change :)
08:26:04S_a_i_n_tbluebrother: Thanks btw.
08:26:12S_a_i_n_tI'll have a play with it very soon.
08:26:16bluebrotherAzureSky: is "usage: revert PATH..." not descriptive enought? ;-)
08:26:40bluebrother(that line might be a bit hard to spot)
08:27:46 Join ender` [0] (
08:30:53 Join kugel [0] (~kugel@rockbox/developer/kugel)
08:36:50S_a_i_n_tAzureSky: I also noticed you're using the first edition of saratoga's volume patch.
08:37:06S_a_i_n_tIt's not really important, but there are 3 more revisons after that.
08:37:14S_a_i_n_t is the final installment.
08:38:05S_a_i_n_tJdGordon: bieber: IS there any documentaion about the themeeditor? (other than the wiki page)
08:38:16S_a_i_n_tsimple instructions, howto's etc.?
08:38:30bieberOther than the README, no
08:38:51S_a_i_n_tbieber: Thanks.
08:38:55bieberI'm a little hesitant to put too much into documentation atm, as the ui is changing pretty frequently
08:39:02bieberAny specific questions I can answer for you?
08:39:29S_a_i_n_tTrial and error is important I guess, I'll try going in blind (as most users do) and see how daunting it is/isn't.
08:39:46S_a_i_n_tbieber: Not yet, just trying to look at it like a first time user.
08:39:57S_a_i_n_t(which I actually am ;))
08:41:27bieberThe one thing that's pretty much impossible to find out from poking around atm is that if you set the keys #screenwidth and #screenheight in your config file, it will set the screen size in previews to those dimensions. Of course, if you have a backdrop defined then it will just grab the dimensions from there and you don't need to worry about it
08:41:46bieberAlthough all the preview does atm is show the backdrop and the viewports blocked out in red
08:42:00JdGordonwhich is enough :)
08:42:16 Quit stoffel (Remote host closed the connection)
08:42:21bieberJdGordon: While you're here, a couple basic questions about rendering
08:42:28JdGordonbieber: I've been thinking and I would prefer if only the actual theme settings were in your config drop list... only a set list are acceptable to the theme site
08:42:59bieberI'm assuming that when I'm rendering a viewport, it's transparent unless the user specifies a backdrop or a bg color?
08:43:48JdGordonumm, yeah, mostly
08:44:02S_a_i_n_tIIUC it will grab FG/BG colours defined from the config in this case
08:44:05JdGordona hidden viewport clears the region when it becomes visible
08:45:32 Quit hebz0rl (Quit: Ex-Chat)
08:45:45bieberI see a couple of ways to define a "Custom UI Viewport", is that where the SBS goes?
08:46:16S_a_i_n_tUI viewport is where the lists are drawn UI viewport takes preference over the config UI viewport.
08:47:00bieberLists as in playlists?
08:47:09S_a_i_n_tas in menu
08:47:41S_a_i_n_t"Bookmarks, Files, Playlists, Now Playing" etc.
08:48:55 Join hebz0rl [0] (
08:49:12bieberOhhh, SBS is for the menus?
08:49:50 Quit Zarggg (Read error: Connection reset by peer)
08:49:56bieberAlright, now things make a lot more sense
08:50:00 Join mitk [0] (~mitk@
08:50:20 Join Zagor [0] (bjst@rockbox/developer/Zagor)
08:50:24bieberFor some reason I was under the impression that sbs was for the status bar on the while playing screen
08:50:27*bieber facepalms
08:51:03S_a_i_n_tthe important thing about the UI viewport is that although it can be defined in two places, only one is ever used.
08:51:20bieberMakes sense
08:51:20S_a_i_n_tIf its defined in the config file and the .sbs, the sbs values will be used.
08:53:47 Join flydutch [0] (
08:54:03JdGordonit can be in the sbs multiple times also :)
08:54:09JdGordonbut only one is ever used
08:54:22bieberJust the last one loaded?
08:54:33S_a_i_n_tit can be conditional
08:57:21 Join Zarggg [0] (~zarggg@2001:0:4137:9e76:0:fbf8:beb1:ba3d)
08:58:54 Quit bieber (Remote host closed the connection)
08:59:46 Join bieber [0] (
09:00:08S_a_i_n_tbieber: Where do I set the screen dimensions? o_0
09:00:49bieberYou have to have a project open (File->Open Project and select the .cfg file)
09:01:17bieberThen set #screenwidth and #screenheight values in the .cfg files, save, and your screen sizes should be correct
09:01:33S_a_i_n_tthis will change though, right?
09:01:55bieberJdGordon: You mentioned only allowing valid keys in the dropdown: do we have a list of those somewhere? I grabbed the current list out of an appendix in the manual
09:01:56S_a_i_n_tit seems a little odd to not be able to set screen dimensions immediately ina new project
09:02:19bieberS_a_i_n_t: Yes, soon I'll be adding a whole dialog full of options for skin preview
09:02:53 Quit jennifur (Remote host closed the connection)
09:04:13S_a_i_n_t*whew*, is there an easier way currently than, new document, save, open project? I assume loading a backdrop will auto adjust the values but I don't have a backdrop set.
09:05:19S_a_i_n_tHmmm...that doesn't even work how I inteded.
09:05:30 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
09:05:38S_a_i_n_tI have to manually create a project .cfg atm?
09:06:06 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
09:06:15 Quit bieber (Read error: Connection reset by peer)
09:06:39 Nick fxb__ is now known as fxb (
09:08:07S_a_i_n_tIt seems quite useful for working on an existing theme, but unless I'm missing something creating one from scratch seems a bit of a PITA
09:08:36 Join bieber [0] (
09:09:07bieberS_a_i_n_t: So far, yes. I haven't decided just how to handle generating project files/directories yet
09:09:29S_a_i_n_tI see.
09:09:43S_a_i_n_tIf I can help with suggestions, just shoot.
09:10:11 Join einhirn [0] (
09:10:36S_a_i_n_t"New Project" with a window to choose available targets would be great.
09:11:01S_a_i_n_tWith predefined attributes for known targets, and a way to add new targets also.
09:11:27bieberPredefined targets are definitely coming
09:11:39bieberI'm more wondering, how would you prefer to have it set up a new project, filewise?
09:11:41S_a_i_n_tthat would make creating a project from scratch relatively painless.
09:12:03bieberJust create all the likely directories on disk? Ask which directories you intend to use? Should it be left loose or all packed into an archive?
09:12:04S_a_i_n_thow do you mean, second question.
09:12:40S_a_i_n_tLoose would be best, as then you could access it with more ease outside of the editor.
09:12:51bieberWell, you have a lot of directories that could potentially be used... /fonts, /backdrop, etc., but only some will actually be needed
09:13:08JdGordonbieber: we should be able to get a list pretty easily
09:13:29JdGordonactually it would be best if it could be shared with the themesite so maybe a text file on the web would be good?
09:13:29S_a_i_n_tI would say create them all, them only archive the ones that aren't empty for the finished product.
09:13:50bieberS_a_i_n_t: That makes sense
09:13:57bieberJdGordon: Is the web stuff also in the SVN tree?
09:14:00S_a_i_n_tHappy to help.
09:14:22JdGordonits in the branch.. but I tihnk its list is stored in the db
09:14:31S_a_i_n_tIt would be simple to check if the dir has content, if not, don't include it in the final archive of the finiched theme
09:15:52S_a_i_n_tI *love* being able to pull things out of the mainwindow, resize them, reposition them etc.
09:15:56S_a_i_n_tgood work on that.
09:16:55S_a_i_n_tI think it is great for now, but new project handling definitely needs some work.
09:17:19AzureSkystill get 1 compile error, guess i can give it a try again tomarrow.....
09:17:55S_a_i_n_tAzureSky: If it fails today, it'll probabl;y fail tomorrow.
09:18:03S_a_i_n_tI'm looking at your pastebin now.
09:18:24bieberJdGordon: The DB on the themesite? Would we just manually sync it with the Theme Editor source from time to time?
09:18:37JdGordonno, the list is pretty static
09:18:40AzureSkyi fully deleted my rockbox folder and remade it and redownloaded the svn and its still erroring
09:19:15S_a_i_n_tAzureSky: What target is this?
09:19:30S_a_i_n_tI can try on my environment.
09:21:06S_a_i_n_tstick around for a while and I'll let you know.
09:21:53AzureSkyk, can stick it out for like 30min, its 12:22 here(am) im gonna need some sleep for wurk tomarow :/
09:23:12AzureSkyif the compiler made better use of my hexa core compiles would go faster, as it stands, its not super slow but its far from blazing fast
09:23:47 Quit bieber (Read error: Connection reset by peer)
09:23:54S_a_i_n_tthere's an option to use multicore.
09:24:02 Join bieber [0] (
09:24:10 Join webguest49 [0] (
09:24:12S_a_i_n_t"make -j(number of cores)" IIRC
09:24:33S_a_i_n_tbut I don't think it supports hex-core
09:24:48 Join webguest18 [0] (
09:24:48 Quit webguest49 (Client Quit)
09:26:50S_a_i_n_tbieber: How do I save a project file?
09:26:58S_a_i_n_tIs this done automatically?
09:27:10bieberJust Ctrl-S, same as with a wps file
09:27:18bieberBe warned, though
09:27:37bieberIt trashes any formatting and any comments that aren't in #value:key form when saving
09:27:51S_a_i_n_tHmmm. :/
09:28:24S_a_i_n_tIt's a great idea, but for now I think I'll only be using it to propose advancements to its development.
09:28:40bieberIt's definitely not production yet ;)
09:28:44S_a_i_n_tI believe I can code faster by hand at the moment.
09:28:49 Join mt [0] (~mtee@rockbox/developer/mt)
09:29:05S_a_i_n_tI understand its a WIP though, not ragging on it.
09:29:22bieberFor now it's treating .cfg files just as machine readable. I need to stick preserving comments/order of options to my to-do list
09:34:05 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
09:37:22 Quit hebz0rl (Quit: Ex-Chat)
09:37:29 Join JdGordon| [0] (
09:37:30 Quit JdGordon| (Changing host)
09:37:30 Join JdGordon| [0] (~jonno@rockbox/developer/JdGordon)
09:39:54 Join Kitr88 [0] (~Kitar_st@
09:41:03TheSevenpamaury: damn, readwrite() from file.c directly passes buffer pointers on large writes, which are likely to be misaligned :-/
09:41:38 Quit kugel (Ping timeout: 240 seconds)
09:41:56pixelmaZagor: you answered to the English voices mail and said you wanted to fix, seems the daily manuals are affected too (which is then why they are gone since today and bluebrother was so kind as to provide some)
09:42:21pixelmajust a heads-up ;)
09:42:26 Quit webguest18 (Quit: CGI:IRC (Ping timeout))
09:42:31pamauryTheSeven: then perhaps we should enforce alignment on all big buffers used with file I/O
09:42:37Zagorpixelma: right, I hope to find time to fix that today
09:43:02 Quit Kitar|st (Ping timeout: 252 seconds)
09:43:10TheSevenpamaury: that won't help if the file offset they're written to is not on a 16byte boundary
09:43:17pixelmaZagor: nice, thanks
09:43:36TheSeveni don't think we can avoid copying-around to a sectorcache buffer in that case
09:44:04pixelmahaving a little more than the last 3 or 4 dailies available would have given a bit more time :)
09:44:47 Quit Kitr88 (Ping timeout: 264 seconds)
09:46:18 Join hebz0rl [0] (
09:48:39 Quit kaylinsigswort (Remote host closed the connection)
09:49:30 Join Kitar|st [0] (Kitar_st@
09:51:29S_a_i_n_tAzureSky: Compiles fine here.
09:51:50S_a_i_n_tI suggest your CygWin is sick.
09:51:51AzureSkyi get rockbox.elf error, whatever that is
09:52:39AzureSkyoh joy u mean i need to reinstall it......what a pain in the arse
09:52:49S_a_i_n_tbah, that's easy.
09:53:25AzureSkythen wait forever for the cross compiler to compile :P
09:53:55S_a_i_n_tInstall according to the wiki, don't install the precompiled toolchains though, download the source (you can just keep the checkout you already have), run, done.
09:54:16S_a_i_n_tyou can also just backup the toolchain.
09:54:22S_a_i_n_tNo need to recompile it.
09:55:08S_a_i_n_tJust backup the user/local dir
09:57:02AzureSkyok will work on that tomarow after work, maybe at work if im really board :)
09:57:07 Join JdGordon1 [0] (
09:57:08S_a_i_n_tHmmm...I just checked no also, and it compiles fine with the patch you've been adding also.
09:57:21S_a_i_n_tyour environment is definitely a bit ill.
09:57:35AzureSkywana zshare that compile for me ;)
09:57:58S_a_i_n_tleave your IRC open and I'll PM you the link shortly.
09:58:07 Join petur [0] (d408b802@rockbox/developer/petur)
09:58:13AzureSkyok thanks :)
09:58:18S_a_i_n_tYou can check it when you wake up or whatever
09:58:23S_a_i_n_tNo worries.
09:59:25AzureSkyim sure you know what its like to bang your head against a keyboard when something just wont work :)
09:59:47TheSevenpamaury: which function is responsible for closing directories? (freeing the structs)
10:02:46 Quit petur (Client Quit)
10:04:14TheSevenwhat the hell. i replaced "unsigned char cache [SECTOR_SIZE]" in the file struct with "struct sectorcache_entry* cache" and that compiles without warnings!
10:04:30TheSeven(without changing the accesses to it)
10:06:37 Quit DerPapst1 (Quit: Leaving.)
10:07:17***Saving seen data "./dancer.seen"
10:10:01 Quit bmbl (Quit: Bye!)
10:10:35S_a_i_n_tTheSeven: I blame black magic.
10:11:07*TheSeven blames the fact that there is not even a single type-safe access to it
10:11:36pamauryTheSeven: closedir you mean ?
10:11:48pamauryfor fat dir, there is no such a function
10:12:08TheSevenso those are just forgotten about?
10:12:33TheSeventhat's bad, as I will need to unlock a buffer when the dir is released
10:13:15pamauryI guess you can add one and call it from closedir
10:13:36pamauryI always thought it was strange that is wasn't done this way
10:13:53pamauryBut check my statement, perhaps I'm wrong
10:18:37pamauryTheSeven: iirc, there is no file close for fat, only a "closewrite"
10:19:09TheSevenbut the fd must be marked free somewhere
10:19:51pamauryyou want to have a look at file.c and dir_uncached.c
10:20:08pamaurythis is were the fat driver is used
10:20:19pamauryand the file handles are hanled
10:20:40TheSevenwhat I also don't get is all the seeking stuff in file.c
10:20:57TheSeven/* seek back one sector to put file position right */ arount line 600
10:21:14TheSevenis that just to make a subsequent writeback end up in the right sector?
10:21:40TheSevenso the fat file pointer will be advanced only during writebacks?
10:22:02pamauryFrom what I remember, if you seek at the position where the current sector is not the same, this code write back the sector basically
10:22:58pamauryI don't have the code at hand, only the online browser so it's no great so if you could figure out yourself, that would be better :)
10:24:06pamauryHum, the comment you state is not in seek code, it's in readwrite
10:25:30pamauryThere are some strange pieces of code:
10:25:50 Join petur [0] (d408b802@rockbox/developer/petur)
10:25:56pamaurywithin the "if ( file->fileoffset + nread < file->size )" condition, it checks for "if ( file->fileoffset > file->size )"
10:27:06*TheSeven considers the whole fat code a bit strange :-P
10:27:27pamaurywell, it
10:27:29pamaury's fat
10:27:48pamauryfs code is always a bit strange
10:28:13pamauryAnd the one sector caching stuff makes it tricky
10:28:57TheSeven...and the LFN hell
10:29:12TheSevenyes, FAT is a bit strange by itself
10:29:26pamauryYou are not allowed to make any comment on LFN ! I rewrote it recently :)
10:32:53*amiconn wonders what TheSeven and pamaury are trying to achieve
10:33:19TheSevenamiconn: getting rid of all sector buffers on the stack and all unaligned accesses to the storage api
10:33:20pamauryWe trying to get rid of the sector buffers hanging there and there
10:33:47amiconnYou cannot get rid of unaligned accesses
10:33:54pamauryBeacuse that's ugly and a waste of space and it does some poor caching
10:34:06amiconnLarge reads and writes pretty much have to go through
10:34:27TheSeventhose can be copied-around sector by sector if they're misaligned
10:34:30pamauryYes but there are so big reads that go to stack buffers we could easily be aligned
10:34:35amiconnBetter caching might be useful, if it's done in a way that doesn't waste ram or binsize
10:34:36TheSeventhat's what the storage driver would do anyway a few levels deeper
10:35:01amiconnCopying around costs extra time.
10:35:21TheSeventhe storage layer will have to do that anyway if it's misaligned, because it can't do DMA otherwise, at least on some targets
10:35:21amiconnNot all drivers require the data to be aligned, so forcing alignment would be a waste
10:35:46TheSeventhat's why it will only be copied around if it's misaligned and STORAGE_WANTS_ALIGN is defined
10:36:07TheSevenactually, the more interesting part is getting rid of the on-stack buffers that lead to dozens of stack overflows on large-sector targets like nano2g
10:37:25pamauryAnd anyway, sector buffer on stack are evil
10:37:50TheSeventhey're a source of unneccessarily misaligned accesses
10:38:55amiconnImo they are not evil in general
10:39:07amiconnBuffers on the stack don't take memory permanently
10:39:19TheSevenbuffers from a sector cache won't either
10:39:39TheSeventhey'll be used as cache memory as long as they aren't locked by someone who needs them
10:40:58amiconnHmm. If you can get this to work efficiently...
10:41:22pamauryAnd that will avoid the big waste of space due to buffers in file.c, dir_uncached.c, ... That might be a way to factor this out and even provide some proper caching
10:41:26amiconnRegarding the misaligned accesses - it may or may not help to copy them to/from a buffer
10:41:57TheSevenamiconn: hm?
10:42:00amiconnSure, you can use dma then - but at the cost of extra copying. Doing the transfer without dma may be more efficient overall.
10:42:16amiconnThat's target dependent, of course
10:42:18pamauryWe don't want to copy anything
10:42:37pamauryWe want the cache to responsible for read/write and buffer "allocation"
10:42:46pamaury* to be
10:42:46TheSevenamiconn: maybe we should add another define then
10:43:35amiconnImo something that depends on the driver should be handled on the driver level
10:43:37TheSevenfor example on nano2g, there is (currently) no way to do unaligned accesses, so they're caught by copying-around code in the nand driver
10:44:02amiconnYes, and that's where it belongs imo
10:44:08TheSeventhat code needs some otherwise unneccessary buffers at the driver level
10:44:24TheSeveni'd like to have the ability to align those at the sector cache level
10:44:25pamauryamiconn: so you would sacrifiece performance because the rest of the code is of poor design ?
10:45:22amiconnNo, why?
10:46:39pamauryThat's waht you are saying, "something that depends on the driver should be handled on the driver level"
10:50:48amiconnYes, but I don't sacrifice performance that way
10:51:11amiconnI have to do the alignment somewhere, but only for some targets
10:51:28TheSevenamiconn: the cache layer can do it without the need for additional memory
10:51:28amiconnSo I prefer doing that where the code is target dependent anyway - and that's the driver
10:52:02amiconnThe cache can be aligned, sure. But large accesses shouldn't go through the cache, as that *reduces* performance
10:52:26TheSeventhey need to go through the cache for consistency anyway
10:53:26TheSeventhe cache will check for collisions and pass them (without copying) to the storage layer, as long as they're aligned
10:53:28JdGordonthe rockbox UI is so not touch freidnly :p
10:53:39pamauryThat is not only for alignment. We want to solve the issue of sector buffers being handled in a non-uniform way. If we do that correctly, there will be no penalty.
10:53:42amiconnI don't see why
10:54:02TheSevenpamaury: actually there will be a small penalty for some checks
10:54:03amiconnWhy they have to go through the cache, that is
10:54:16pamauryamiconn: because that's a cache !
10:54:34 Join swilde [0] (
10:54:41pamauryIf something is in the cache, you want to enforce consistency or take advantage of it
10:54:55amiconnReading a large chunk of data from a file (e.g. an audio file) will almost for sure happen only once
10:55:06pamauryTheSeven: yes but the penalty is small considered to actual reading/writing
10:55:38TheSevenamiconn: that's why those will just be passed through by the cache after checking that there are no dirty pages in that area
10:55:57amiconnIf it goes through the cache, it will throw out previously cached data (e.g. parts of the fat or directory blocks), which are likely to be needed again before reading another file
10:56:15amiconn-> performance loss
10:56:25TheSevenno, it won't
10:57:04TheSeventhe cache will just pass very large accesses which are aligned and don't touch dirty pages
10:57:18amiconnDirty pages? You want to do write caching?
10:57:37TheSevenonly in a limited way
10:58:23TheSeventhe cache would allow for arbitrary write caching, but the file system code will usually trigger page cleaning as soon as it has finished modifying the sector
10:59:08pamauryamiconn: we don't invest anything, all this caching stuff is already done
10:59:13pamaury* invent
10:59:26pamauryin more or less clever ways
10:59:32TheSevenit's just done in ugly on-stack or in-struct buffers right now
11:01:04TheSeventhe only difference on the caching side right now is more clever read caching, as the buffer space for currently-unused files or directories will be used as an MRU cache for small accesses
11:03:52amiconnWell, let's see
11:07:08 Join efyx [0] (
11:07:43TheSevenpamaury: I'm about to make the file.c code always lock sectors exclusively, even when only reading. Any objections?
11:08:24pamauryWhy exclusive ?
11:08:43TheSevenbecause I can't easily distinguish between reads and writes at the place where I need to lock it
11:08:51TheSevenand nobody else should be needing it anyway
11:09:06pamauryA file can have several read handles
11:09:09pamauryAt least in theory
11:09:37TheSevenhow can i find out if the handle may have write access?
11:09:59TheSeventhe write flag in filedesc?
11:10:08pamauryEach knows if it needs write access or not I think
11:10:18pamauryCheck in open
11:10:46TheSevenas far as i can tell, you can just issue writing request to read-only file handles
11:10:52TheSeventhat doesn't seem to be checked anywhere
11:13:50pamauryhuh ?
11:13:54pamauryI doubt it
11:14:02TheSevenat least not in readwrite()
11:14:06pamauryThat's wrong then
11:14:17pamauryWe don't want to enforce such a bad practice
11:15:10pamauryPerhaps thr fat code checks that ?
11:15:29pamauryIn fact, I'm pretty sure it does
11:15:29TheSeventhat would be evil
11:16:00TheSevenand might cause delayed failures, because you could still write to the sector cache
11:16:39TheSevenanyway, I'm gonna check it in readwrite, can't hurt if it's checked twice
11:17:10pamauryThis way, you don't need exclusive lock for read
11:23:45TheSevenalso, the current implementation seems to only detect an out of disk space situation after the first non-fitting sector was written, while the cached implementation will already complain when trying to lock the sector
11:25:22TheSevenhm, and it looks like reading past EOF in the tail will cause EIO
11:30:37pamauryreading past eof should truncate read, that's the expected behaviour but it should not trigger an error iirc
11:42:21Zagorpixelma: do you know if we can cut down on the output from the manual builds? it's awfully noisy.
11:43:04pamaurycan't you redirect everything to /dev/null ?
11:43:16Zagorwe still want to see any errors
11:44:23pixelmaZagor: yes, bluebrother provided a patch, (there must be some pastebin links from yesterday). Unfortunately it also filters errors on older tex installations
11:45:18pamaurylatex is a pain about filtering errors
11:45:35pixelmaalthough we adjusted it a bit later (no complete patch for that but it works even in older installations
11:46:51pixelmaZagor: all in the logs of yesterday, a bit sprinkled across the hours though. While you are at it, do you know which latex package the daily build server runs, how old it is?
11:50:41Zagorit's running texlive 2009-9
11:55:10TheSevenhm. "Misaligned data buffer at 09090909 (sector 151587081, count 1)"
11:55:17TheSeventhat address is a *bit* suspicious
11:55:29 Join evildork [0] (
11:56:03TheSevengiven the fact that 0x09090909 = 151587081, it's even more weitd
11:58:39pamaurythat's *really* suspicious
12:01:18evildorkIf I use Rockbox, will I be able to organize my music by custom tags such as Composer>Mood>Opus, etc.?
12:01:47 Join giovanni [0] (~giovanni@
12:02:21TheSevenpamaury: someone overwrites my sectorcache_entry structs :-/
12:03:03pamauryhow is this possible ?
12:03:21TheSevena bug somewhere
12:03:46TheSevenas I said, the compiler didn't even warn when I changed the type of the cache field in the fd struct
12:03:56TheSeveni probably missed an access to that one
12:04:28 Quit thefirstM (Remote host closed the connection)
12:04:32 Quit slck (Ping timeout: 276 seconds)
12:04:39pamauryI had a similar problem in tagcache. the code was using the wrong structure names ! It worked but that was an incredible mistake
12:05:16pamauryIsn't there a warning about pointer casts ?
12:05:38TheSevenall accesses to that bugger are done in a type-unsafe way
12:07:19***Saving seen data "./dancer.seen"
12:07:28 Join watto [0] (~watto@
12:07:58 Join slck [0] (Venci@Slackware.SlackPix.Com)
12:09:02pamauryTheSeven: try to compile with -Wc++-compat, perhaps it can help
12:09:09JdGordonquick arm question.. does this diff look correct? I want to only boot the OF on hold or usb connected
12:09:19JdGordoncurrently it boots rockbox if hold is on
12:10:29TheSevenJdGordon: USB present = bit 0x4000000 *not* set?
12:10:50JdGordon0x3c when not connected, 0x38 when it is connected
12:11:32TheSevenyou mean 0x3c000000?
12:11:53TheSevenlooks sane then
12:12:12JdGordonthat change shuoldnt be able to brick it right?
12:12:52TheSevenIIUC that's a cowon d2? those aren't brickable anyway IIRC
12:13:40TheSeventhe only way to make that code fail would be a data abort during that read (wrong memory address for the GPIO)
12:13:59TheSevenotherwise, it will always boot one or the other, even if something is the wrong way round
12:14:23TheSevenand you should still be able to use the hold switch to boot it to the OF, even if the code below is broken
12:18:28JdGordonwierd... it seems to work, but wont boot rockbox at all...
12:19:29 Quit AzureSky (Ping timeout: 276 seconds)
12:21:01 Join Rob2222 [0] (
12:21:25TheSevenmaybe the usb detection isn't working yet at that point during boot?
12:21:35TheSevenis the gpio set to input?
12:21:50JdGordondo I need to adjust that second ldreq pc line?
12:22:19JdGordonmake it -44 ?
12:22:20TheSeven*facepalm*, yes, of course
12:23:00JdGordon44 or 48?
12:23:09TheSevenbut that would only make OF boot because USB is plugged fail
12:28:21JdGordongrr... there is another ldr [pc, #-28] a bit further down... can I copy that to a reg to make it more sane?
12:28:56TheSevenif there is a free reg, probably yes
12:29:04 Quit GeekShadow (Ping timeout: 260 seconds)
12:29:14TheSevenwhere did that -28 further down point to originally?
12:30:17JdGordon.word 0 /* Saved entrypoint of original firmware*/ I think
12:30:27TheSeventhe upper one, yes
12:30:37JdGordonthe line after that is the same but points to the rockbox bootloader bin
12:30:46JdGordonso I guess just make sure that second one points correctly?
12:30:49TheSevenbut if there were multiple [pc, -28]s, they can't possibly point to the same value
12:31:06JdGordontwo .word 0 lines together
12:31:23TheSevenis that second one immediately below the #endif?
12:31:27TheSevenif yes, just change it to -44
12:32:54 Join dfkt [0] (dfkt@unaffiliated/dfkt)
12:33:12S_a_i_n_tperhaps a silly question, but what is the difference between a standerd build (N) and a Debug build (D)?
12:34:26JdGordonusb doesnt work though.. its either pointing to the wrong one or logic is bad :)
12:34:35JdGordonthanks TheSeven :)
12:35:03TheSevenare you sure that you want to access 0xf005a080?
12:35:57 Join DerPapst [0] (
12:35:58TheSevenS_a_i_n_t: it enables some additional sanity checks in the kernel, and probably some DEBUGF output
12:36:00JdGordontst doesnt does equality not logical &?
12:36:33*S_a_i_n_t has a parsing error
12:36:47TheSeventst will set the flags on the result of r0 & #whatever
12:37:03TheSeveneq = the result was zero, ne = nonzero
12:37:51JdGordonok, so it is only checking the one bit?
12:38:37TheSevenhowever, it still looks like you ignored the #0x20 offset to the address
12:39:01TheSeven(no idea what it is good for in the first place)
12:39:28JdGordonAH, I see what that is doing
12:39:49JdGordonGPIOA is 0xf005a000 + 0x20
12:39:57JdGordonis that second ldr needed?
12:40:08JdGordonso I'm reading the wrong gio
12:41:08TheSevenif another reg is free, you might want to ldr the base address to r1 instead, use [r1, #0x20] and [r1, #0x60], remove the second ldr r0, =... and adjust the offsets again
12:41:24TheSeventhat would save 8 bytes of binsize :-)
12:42:28JdGordonit works :D
12:43:36JdGordonI'm getting nice scrambled lcd output though when rockbox.d2 eventually loads!
12:48:36CIA-8New commit by 03mt (r27006): Wrote apps/codecs/wmapro.c and modified libwmapro to make the codec work in the sim. Neither libwmapro nor wmapro.c have been added to the main build ...
12:50:37CIA-8r27006 build result: 92 errors, 296 warnings (mt committed)
12:51:42CIA-8New commit by 03jdgordon (r27007): change the D2 bootloader to boot rockbox by default. boot OF is hold is on or usb is connected
12:52:19Zagormt: it certainly looks like they have been added...
12:52:29mtYeah :/
12:53:31 Quit petur (Quit: medic!)
12:53:35CIA-8r27007 build result: 93 errors, 295 warnings (jdgordon committed)
12:54:03JdGordonhow did a warning become an error ? :)
12:55:44mtaah .. the infamous nomsg commit comes back to bite me
12:55:57 Quit mitk (Quit: Leaving)
12:56:06 Quit retroj (Ping timeout: 240 seconds)
12:56:31 Join ucchan [0] (
12:56:55*TheSeven shouts at file.c
12:57:18TheSeventhat thing keeps hiding its memory corruption bug :-/
12:59:10pamauryTheSeven: did you try with the warning I told you ?
12:59:44TheSeveni checked them all manually
13:00:18pamauryTheSeven the warriori does everything himself :)
13:00:32ucchanfunman login?
13:00:43TheSevenit weren't that many
13:01:49 Join AsusFreak [0] (
13:02:01pamauryucchan: he is not online now it seems
13:03:07pamauryIt would probably be a good thing to try to compile rockbox with strict pointer cast rules, just to check if there are more of these incorrect pointer types. But there would be hundreds of false positive
13:04:05TheSeventhere are also some other checks that seem to be disabled, even though most of their hits seem to be actual bugs/quirks
13:04:19pamaurylike ?
13:06:33ucchanpamaury I think so. sigh..
13:07:41pamauryucchan: is the problem related to his last commit on the text viewer ?
13:08:35CIA-8New commit by 03mt (r27008): Fix reds. (Actually removed wmapro from main build, which was added in a previous commit)
13:09:14TheSeven/data/rockbox-trunk/firmware/common/dircache.c:953: warning: identifier new conflicts with C++ keyword
13:09:15ucchanpamaury: yes. your commit is very wrong.
13:09:30JdGordondoes anyone know how the d2 storage works?
13:09:50JdGordontrying to find how the internel/external drives kludge happens (they are swapped)
13:10:18CIA-8r27008 build result: All green
13:10:23pamauryJdGordon: it's basically a hack. If you boot from sd card. The sd card is the main storage and the internal sotrage is seem as a sd card. If you boot from internal storage, it's the contrary I think.
13:10:39JdGordonright, I want to undo that hack :)
13:10:54JdGordonI also cant find where it finds which to boot from
13:11:32pamauryI think the bootloader does the selection but I might ne wrong. I didn't play a lot with the d2 yet
13:12:45pamauryucchan: if you can wait, funman is usually online later on. I would say in 3 or 4 hours, he is likely to be there
13:13:18pamauryJdGordon: ask gevaerts, he probably knows that
13:13:52TheSevenpamaury: damn, -Wc++-compat didn't yield anything related
13:14:01*gevaerts denies everything
13:14:06pamauryThat's a shame
13:14:42JdGordonanyone seen any rogue #ifdef COWOND2 lines in the file drivers?
13:15:33 Quit Battousai (Read error: Operation timed out)
13:15:55TheSevenJdGordon: nope
13:15:55gevaertsJdGordon: the ordering is decided by the order in storage_init() in storage.c. Putting NAND before SD will change that, without impacting any other target (no other target has those two AFAIK).
13:16:34gevaertsJdGordon: I'm not sure if it's a good idea to suddenly change this though. This ordering wasn't really an accident, and people might depend on it
13:16:52AsusFreakHi. Just saw that you're catching Cowon's D2. I just put a task in the bug tracker: absolute point mode still has problems (FS #11424).
13:17:05 Join Battousai [0] (~bryan@gentoo/developer/battousai)
13:17:34TheSevengevaerts: we still don't have R/W NAND access on the D2?
13:17:36gevaertsTheSeven: indeed not
13:18:11JdGordonAsusFreak: I've already replied to it
13:18:14*TheSeven wonders why people are so scared of FTLs
13:18:30JdGordonkugel, get online! :D
13:18:55gevaertsTheSeven: in this case shotofadds did spend lots of time on it. I wouldn't say he's scared of it
13:19:26AsusFreakJdGordon: ok, thanks for information.
13:19:35TheSevenwhat's holding back R/W access then?
13:19:54pamauryPerhaps he doesn't have a full understanding of it
13:20:03gevaertsYou can enable it I think, but it doesn't survive long
13:21:30TheSevendebug it, then :-)
13:21:38*JdGordon dares pamaury to fix it :D
13:22:04gevaertshm, or not
13:22:19gevaertsActually, the write bit isn't there at all yet
13:22:26pamauryI can have a look, if the format is explained.
13:22:30gevaertsIt still goes wrong every now and then even in readonly mode
13:23:19gevaertsshotofadds might have uncommitted write code though
13:23:32TheSevenwell, R/O failures should be rather easy to debug, if you can get hold a flash dump of such a state
13:23:34*gevaerts recommends talking to shotofadds about this :)
13:24:06pamauryyes clearly
13:25:17JdGordongevaerts: so the actual ordering problem is storage doesnt really let a target arbitrarily set the ordering?
13:27:42gevaertsJdGordon: well, yes and no. The sd vs nand ordering thing is a purely D2 issue (by coincidence, sure, but it's the case anyway). I've not seen a case yet where this fixed ordering causes problems, so I at least haven't spent time on it. Changing that shouldn't be too hard though if needed
13:28:56JdGordonisnt it a fairly safe assumption that if a target has sd it is likely to either use it for external? so it should almost certainly be last in the list
13:29:32JdGordonI mean nothing would ever use sd but have a ata port also
13:29:55JdGordonand even if it did, it owuldnt use sd as main storage
13:30:21gevaertsthe Elio has both ATA and sd :)
13:30:32gevaertsbut yes, it will use ATA as main storage
13:30:32JdGordonboth internal?
13:30:39gevaertsno, a card slot
13:31:06JdGordonright, so ATA being before sd is a fairly safe assumption, ditto nand before sd :)
13:31:20gevaertsYou're right of course. The only reason SD *isn't* last is the D2 situation, where at the time it was decided to use SD as main storage to have writable main storage
13:34:08JdGordonand the only situation where writable main storage is required is the .rockbox folder, and that needs to change for RaaA anyway, so maybe we wrap all ROCKBOX_DIR calls in something like open_rbdir(char* filename, int mode) where filename is relative to /.rockbox and it will work out where that actually is?
13:34:41gevaertsthat could work
13:34:45JdGordonwe could probably even do that transparently in the open() call?
13:35:42gevaertsIt's not exactly simple I think. How would you handle it? File not present on NAND, open it on SD?
13:35:44pamauryIf possible, it would be better to avoid the creation of a virtual folder by having a transparent redirection in open
13:36:10JdGordongevaerts: yes
13:36:32JdGordonit would be messy if say config.cfg is in both, but as long as it isnt it would only write/read to the sd
13:36:57gevaertsWhat if e.g. a user does something like generate the database from a PC, and then update it in rockbox?
13:37:27JdGordonis there anything we could do with that? probably not
13:37:40 Quit DerPapst (Quit: Leaving.)
13:37:57gevaertsOr what if a user tries to put a file in .rockbox from the file browser? Should it fail or transparently move it to the SD? What if the SD has a file of the same name?
13:37:59JdGordonor if internal is ro and we are opening for write we reverse the search order
13:38:51JdGordonok, so using a sepeare open function would be safer
13:39:15gevaertsI think so. At least it would have less nonobvious side effects
13:39:30gevaertsPossibly its complexity might offset that of course
13:40:05JdGordonI cant image it being very complex
13:40:19JdGordonI hope kugel hasnt done anything for RaaA on this yet
13:43:08JdGordononly 61 uses of ROCKBOX_DIR in apps... easy :)
13:47:02JdGordonneed to fiddle with open_internal() to make this work I think
13:47:43JdGordonunless there is a good safe way to prefix the disk mount point onto the path?
13:52:40TheSevenah, finally
13:53:04TheSevenafter adding about a hundred debug statements all over the place, one actually triggered
13:53:16TheSevena null pointer is causing the whole mess
13:55:09 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
13:57:43JdGordongevaerts: ok, you were right.. this isnt exactly trivial
14:03:55TheSevenpamaury: ok, the cache seems to mostly work
14:04:18TheSevenhowever, someone isn't unlocking some handles when entering usb mode or shutting down
14:04:19pamaury\o/ mostly ?
14:04:56TheSevenbut having a database initialization succeed through the cache is at least a good sign :-)
14:05:20pamauryTheSeven: iirc, some handles are kept open when entering usb mode. This is not a problem if you are not using ums but when you are, it's embarassing
14:06:51pamauryThat should probably be fixed if it's the case
14:07:22***Saving seen data "./dancer.seen"
14:07:30pamauryYou can try to dump the filenames which are still opened when entering usb screen
14:08:15 Quit pjm0616 (Ping timeout: 260 seconds)
14:08:33pamauryPerhaps the playlist one. I can't remember its name
14:08:43JdGordon.playlist_control ?
14:09:04JdGordonvery likely to be left open
14:09:26JdGordonalso a .m3u if it is too large to fit in the playlist buffer
14:09:27pamauryI had this problelm once, I think it's always open, a bad thing for sure
14:14:18CIA-8New commit by 03jdgordon (r27009): Fix the hotkey touch region in the wps
14:16:10CIA-8r27009 build result: All green
14:17:47TheSevenwhat would happen if that file is modified via usb?
14:17:54TheSevenbad things(tm)?
14:18:09TheSevenone of the opened ones
14:18:23 Join DataGhost [0] (
14:18:23 Quit DataGhost (Changing host)
14:18:23 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
14:18:40pamauryreally bad things(tm)
14:19:12pixelmasomeone else looked around for kugel too?
14:19:39AsusFreakJdGordon: thanks a lot for r27009. Your speed is great.
14:20:41 Join pjm0616 [0] (~user@
14:21:26JdGordonbieber: I just noticed you have bmp displaying also... niiice :)
14:26:40TheSevenpamaury: nvram.bin is the first one that's hit
14:26:41 Join CGL [0] (~CGL@
14:27:09TheSevenwhat is that file good for btw?
14:27:28TheSevenas far as i can tell it's an archos relic
14:27:31pamauryWhat is this fucking file ? Isn't this for fast preferences that are stored in a non volatile ram ?
14:28:09TheSevenstored in an NVRAM on archos, in a file on everything else IIRC
14:28:28pamauryWhat does it contain ? Is it used ?
14:28:51JdGordonit contains a few settings which arnt in config.cfg
14:28:56JdGordonit should be closed when not used
14:28:56Torneit contains the settings that are marked as being in nvram, which i realise isn't a helpful answer :)
14:29:37Torneit is slightly useful, perhaps, because iirc it's used to remember several things that change often, and storing then in nvram.bin is less effort than rewriting config.cfg on every shutdown..
14:29:49CIA-8New commit by 03mt (r27010): Remove the stubbed main() from wmaprodec.c since it's useless now and remove Makefile for the same reason.
14:29:51Torneiirc that's where things like the dircache size and so on go
14:30:23JdGordonI'm looking at the code and it should be being closed on use
14:30:38JdGordonit will only stay open if it is empty on first call
14:30:42JdGordonwhich is a bug
14:31:00JdGordonsettings.c line 125
14:31:35CIA-8r27010 build result: All green
14:31:57TheSeventhat's unlikely to be the cause
14:32:48 Quit hebz0rl (Ping timeout: 264 seconds)
14:34:30CIA-8New commit by 03theseven (r27011): Don't leak a file handle if nvram.bin is <8 bytes in size
14:36:07CIA-8r27011 build result: All green
14:36:18TheSevenbut it's still open
14:38:10CIA-8New commit by 03mt (r27012): Add a patch file in libwmapro to make it easier to add the library and the decoder to the main build and modify README.rockbox accordingly.
14:38:54TheSevenhm, that one was actually a lost buffer handle, not a lost file handle. fixed.
14:39:50CIA-8r27012 build result: All green
14:40:17TheSevenoh well
14:40:25TheSeventhe next one is my UI font :-/
14:42:09 Join AzureSky [0] (
14:42:57 Quit AsusFreak (Quit: CGI:IRC (EOF))
14:43:10CIA-8New commit by 03jdgordon (r27013): Allow the progressbar and volume touch regions to be reverse by prepending them with ! (so full is the left/top instead of right/bottom)
14:45:01CIA-8r27013 build result: All green
14:45:12TheSevenshould we switch to the SYSFIXED font during USB?
14:45:58JdGordonwe dont really need to do we? the text on screen is static
14:45:58TheSevenis there any text besides the HID mode being drawn during USB?
14:47:00gevaertsnot *today*
14:47:15gevaertsThere has been talk about themable USB screens though
14:47:27gevaertshm, right
14:47:31*gevaerts sees the issue :)
14:47:52TheSeveneven the HID mode could potentially cause trouble if there is a glyph cache miss
14:47:57gevaertsyes, indeed
14:48:42*gevaerts now claims that that is actually the problem with HID mode for some people, thereby pushing all responsibility for it to other people
14:48:58TheSevennow how do i get rid of that annoying handle?
14:50:01 Join Ed__ [0] (~chatzilla@
14:51:00Ed__hello, I have an h340 and the WPS doesn't change no matter what theme I pick, and I just used the rockbox utility to install the latest stable release, can anyone tell me why?
14:51:46JdGordonbecause the themes changed recently so rbtuil is installing the wrong themes
14:52:10JdGordondownload the themes manually from
14:52:25Ed__how do I get rid of the themes that are in there?
14:52:49TheSevenJdGordon: any thoughts on how to get rid of that handle?
14:52:55TheSevenhow can I make sure the font is unloaded?
14:53:13JdGordonwhich handle?
14:53:20TheSevento my GUI font
14:53:34TheSeven(and restore the old state again when leaving USB mode)
14:53:55JdGordonforce an unload before going into usb mode, then settings_apply() when you come out
14:54:35TheSevenwill that automatically switch to the sysfixed font?
14:55:04TheSevenmay I do that from the USB thread, or does it need to be done from the GUI thread?
14:55:12JdGordonyou need to wrap that in NB_SCREENS(i) also and unload FONT_UI+i
14:56:06JdGordonumm... I'm not sure what scrolling will do when you unload its font
14:56:16JdGordonotherwise it is probably safe from usb thread
14:56:51TheSevenhm, isn't the gui thread reacting to that USB event anyway?
14:56:57TheSevencan you point me to the handler for this?
14:57:49gevaertsthe gui thread jumps to the usb screen code
15:03:23TheSevenFOR_NB_SCREENS(i) font_unload(FONT_UI + i); doesn't seem to close the handles
15:05:58 Quit antil33t (Read error: Connection reset by peer)
15:06:05 Join antil33t [0] (
15:09:09 Quit mikroflops (Ping timeout: 276 seconds)
15:16:26 Join komputes [0] (~komputes@ubuntu/member/komputes)
15:22:31 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:22:39kugelwho yelled after me?
15:22:47 Quit bieber (Ping timeout: 240 seconds)
15:23:04 Join bieber [0] (
15:24:03*TheSeven doesn't understand that usb codfe
15:24:17*kugel read coffee
15:24:19pamauryThe usb code is not easy to follow
15:24:46TheSevensomeone's calling sectorcache_invalidate, but I can't seem to figure out who
15:25:19TheSevenit's probably coming through a disk_mount_all() call, but why on earth would anybody call that on usb *insertion*?
15:25:44pamaurybecause on ums, noone can access the disk so the usb code unmount everything
15:25:48pamaurysee usb_slave function I think
15:26:15pamaury(in usb.c)
15:26:16TheSevenyeah, but it does actually *mount*, not unmount it on insertion
15:27:19 Join evilnick_B [0] (0c140464@rockbox/staff/evilnick)
15:27:24pamauryno, that doesn't make sense
15:27:48JdGordonkugel: I was wondering if you'd done anything about /.rockbox or ROCKBOX_DIR for RaaA yet?
15:28:31 Join jgarvey [0] (
15:29:51kugelnot yet
15:30:13pamauryTheSeven: if I look at the code, it does mount when usb is unplugged but does not umount when usb is plugged
15:30:18pamaurygevaerts: can you confirm that ^
15:30:25JdGordonkugel: do you know how you're going to do it yet?
15:30:30 Quit jgarvey (Client Quit)
15:30:56 Join jgarvey [0] (
15:30:58kugelI hope I can replace ROCKBOX_DIR with something that's not a compile time constant for RaaA
15:31:01TheSevenpamaury: i've placed a panicf in front of all calls to disk_mount_all in usb.c, and none of those gets triggered
15:31:13TheSevenso it must be somebody else that's calling it, but who?
15:31:23pamauryno idea
15:31:24TheSeventhe only other place i can find where it is called is on boot
15:31:43kugelthen change PLUGIN_DIR etc to /usr/local/lib, CONFIG_DIR to $HOME, and so on
15:31:54pamaurydid you grep the entire tree ? Even in app/ for the default handler ?
15:32:28kugelpamaury: your sim usb driver would be helpful now, no? :)
15:32:54 Join anewuser [0] (anewuser@unaffiliated/anewuser)
15:33:20pamauryMy usb sim driver behaves differently from a real target, and with all the modification you did to the tree, it would not compile :(
15:33:29JdGordonkugel: ok, damn... probably wont be helpful for the d2 then
15:34:05kugelJdGordon: I think it would be inconvinient for the d2 anyway, because you'd need the OF to update rockbox
15:34:14kugel(assuming rockbox usb works on that thing)
15:35:02pamauryTheSeven: and you said to put a panic on every call and no one gets triggered ?
15:35:12TheSevenboth of the usb.c ones have one
15:35:18TheSeventhe only other relevant one is
15:35:18TheSeven../../apps/main.c:570: rc = disk_mount_all();
15:35:44pixelmakugel: also me... you said somethng about the FM quickscreen - do you have plans to revive it? Your changes to the c200 keymap back then cause a slight bug (you also decrease volume now at the same time you would call the quickscreen from the radio), this bug stayed even after removing the FM quickscreen action and the fix is probably different depending on whether you won't need the quickscreen or you do
15:35:49TheSevenah, indeed, that's usb-related
15:35:53TheSevenwhere is the boot one then?
15:36:31TheSevenhmm, nope, that's early usb
15:36:37 Join DerPapst [0] (
15:36:59kugelpixelma: didn't I use the same button as the wps quickscreen on?
15:37:35 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
15:37:52pixelmakugel: I think you did but that doesn't matter for the problem
15:38:36 Join hebz0rl [0] (
15:39:02pixelmait's an action (or prevent another action being hit at the same time problem) and the fm has a context on its own, also some special actions on its own
15:39:52pixelmaI didn't notice until yesterday as I prefer my other targets for radio
15:41:59 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
15:43:29 Join chaos [0] (~chaos@gentoo/user/ch4os)
15:43:30 Join panni_ [0] (
15:43:52TheSevenJdGordon: where is FONT_UI declared?
15:44:10JdGordonfont.h probably
15:44:14 Quit pjm0616 (Ping timeout: 272 seconds)
15:44:38 Join pjm0616 [0] (~user@
15:44:42TheSevenoh, right, an enum
15:44:49TheSevendidn't grep properly for that
15:45:57TheSevenjudging from that, FONT_UI can't be unloaded at all
15:47:04TheSeven if (font_id >= SYSTEMFONTCOUNT && pf)
15:47:18TheSevenSYSTEMFONTCOUNT is below FONT_UI in the enum
15:47:30JdGordonoh right
15:48:06JdGordonyou could change that :)
15:48:39JdGordonI think you could also load a dummy font file
15:49:31JdGordonthat will cause the fd to close at least
15:49:35JdGordonnot too pretty though
15:50:10 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
15:51:09JdGordondoes anyone have any suggestions how to go about replacing the skin engine? right now I'm commenting out pretty much anything which stops it compiling when I remove the old code, but that is huge and messy
15:52:01 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
15:54:00TheSeventhe next one is a directory handle on /.rockbox/wps/UniCatcher
15:54:43TheSevenwell, that's the alphabetically very last folder on it - dircache?
15:57:48TheSevenindeed - that one seems to be coming from dircache
15:58:01JdGordonmore likely last dir opened?
15:58:11TheSevenif i disable dircache it's gone
15:58:38TheSevenand i'm testing with usb plugged in during boot, so that it'll enter usb mode as early as possible
16:01:56TheSevennext one: /.rockbox/viewers.config
16:02:28TheSevenwho could be holding a handle to that file?
16:07:25***Saving seen data "./dancer.seen"
16:09:19pamauryhehe, file handle mess
16:09:37pamaurythat was a trap
16:09:41 Join MethoS- [0] (~clemens@
16:14:11 Join halmi [0] (
16:19:56 Quit bmbl (Quit: Bye!)
16:20:16S_a_i_n_tTheSeven: Can you confirm keyclick is broken on Nano2g?
16:20:40S_a_i_n_tIt only "clicks" if playback is happening, silent otherwise.
16:20:54 Join retroj [0] (~retroj@unaffiliated/retroj)
16:20:58TheSevenhm, actually i haven't heard a single click recently
16:21:01S_a_i_n_t(for me anyway...just noticed it now)
16:21:03TheSevenmight be disabled, let me check
16:21:29JdGordonTheSeven: is that with dircache on?
16:21:41TheSevenJdGordon: what?
16:22:03S_a_i_n_tI also notice this stops the "hiss" during startup, which confirms my theory that something in keyclick makes that awful hiss before an audio event
16:22:19TheSevenJdGordon: no, with dircache disabled
16:22:48*TheSeven didn't notice a hiss
16:22:56S_a_i_n_t(explanation: when I boot my Nanos, there is an awful hissing noise until the first keyclick, or until playback starts)
16:22:56JdGordonwoops... read_config() in filetypes.c is leaking a fd
16:23:25S_a_i_n_tquite noticable on my targets, better IEMS perhaps.
16:23:45ucchansorry, I quit soon. I must go to bed. funman, if you read this comment, please read FS #11399.
16:23:47TheSeveni just get a loud click when it initializes the codec
16:24:14TheSevenJdGordon, pamaury: apparently that cache is a bug attractor :-)
16:24:28TheSevenit already caught four unrelated bugs
16:24:40CIA-8New commit by 03jdgordon (r27014): close the viewers.config file fd
16:24:44pamaurythat's because they are lots of evil things related to open handles that no one want to touch
16:24:52S_a_i_n_tTheSeven: You gots clicks? I checked myself, and keyclick is definitely enabled, yet I have not keyclick unless playback is happening :/
16:25:04 Quit ucchan (Quit: Leaving...)
16:25:13JdGordonwe should be able to detect this in the sim no?
16:26:43CIA-8r27014 build result: All green
16:27:27TheSevenS_a_i_n_t: worksforme, r26788:26996M-100621
16:28:21*S_a_i_n_t scratches his head.
16:28:57S_a_i_n_tI shall try building that revision.
16:28:58*TheSeven is updating to r27014M
16:29:12 Quit halmi (Quit: halmi)
16:29:33 Join halmi [0] (
16:30:04S_a_i_n_tI'm quite annoyed you can't hear the hiss to be honest though :/
16:30:25S_a_i_n_tits quite annoying, for me.
16:30:41JdGordonTheSeven: lsof shows the sim only leaving open the font files... so you might be almost done :)
16:31:12S_a_i_n_tAfter boot there's a static hiss that goes away as soon as there is an audio even like voice, keyclick, or playback.
16:31:33S_a_i_n_tbut is constant until then.
16:31:39TheSevenJdGordon: I already extended that font thing: for (i = FONT_UI; i < MAXFONTS; i++) font_unload(i);
16:31:55TheSevenS_a_i_n_t: never had that
16:32:25TheSevenmight have been caused by one of those pcm changes
16:32:43JdGordon.playlist_control stays open after stopping playback
16:32:50S_a_i_n_tI've had it through the history of the port.
16:33:18S_a_i_n_tAt first I thought it was crappy IEMs, so I got better ones, and it got worse :/
16:33:30TheSevenhow loud is that?
16:33:31S_a_i_n_ts/worse/more noticable/
16:34:07S_a_i_n_tquite noticable at -25dB, earsplitting at +6
16:34:22S_a_i_n_tlike a white noise, static.
16:34:49 Join dfkt [0] (dfkt@unaffiliated/dfkt)
16:35:42S_a_i_n_tI tested it by cranking the volume to =^, turning the player off, booting and not touching anything.
16:35:53S_a_i_n_targh... "+6"
16:36:36S_a_i_n_tit dissappears as soon as there's a keyclick, voice event, playback starts etc then doesn't come back.
16:36:54JdGordonbuilding the DB doesnt leak any fd's :)
16:36:57S_a_i_n_tas if something isn't init'ed properly at boot.
16:37:17TheSevenhm, my codec doesn't seem to produce this
16:37:34TheSevenhowever, i have noticed increased power consumption until playback is started for the first time
16:37:39S_a_i_n_tweirs, all my Nanos do, regardless of the build.
16:37:39TheSevencould have the same root cause
16:38:42TheSevenyours are both rev. 7 devices?
16:38:43S_a_i_n_tIt happens in my patched builds, in clean SVN, on both Nano 1st and 2nd Gen.
16:38:47TheSevenmine is a rev. 9
16:38:58S_a_i_n_thow do I check this?
16:38:59TheSevenah, even first gen? that's funny
16:39:11S_a_i_n_tyeah, even first gen.
16:39:43S_a_i_n_t(sorry to distract you from you cache stuff ;))
16:41:49S_a_i_n_tHowever, this "hiss" isn't happening on the Nano2g with broken keyclick, which makes me suspicious of keyclick as being the cause.
16:42:00JdGordonTheSeven: calling playlist_shutdown() before starting usb is probably a good idea
16:42:06JdGordonthat will close the control file
16:42:14 Join toffe82 [0] (~chatzilla@
16:42:23TheSevenfrom which thread?
16:43:13TheSevenyay! your viewers.config fix killed the last one of them
16:43:51JdGordonmain thread
16:45:08JdGordonthat should maybe be added to clean_shutdown() also
16:45:43 Join halmi_ [0] (
16:46:08JdGordonah, it is there already, in sustem_flush()
16:47:23 Join mikroflops [0] (
16:47:45 Quit halmi (Ping timeout: 245 seconds)
16:51:57evildorkIf I use Rockbox, will I be able to organize my music by custom tags such as Composer>Mood>Opus, etc.?
16:52:00 Join Ed___ [0] (~chatzilla@
16:53:26*TheSeven glares at *very, very* outdated comments in apps/main.c
16:54:24 Join n1s [0] (~n1s@rockbox/developer/n1s)
16:54:30JdGordonsee! comments are a waste of time!
16:55:05 Quit Ed__ (Ping timeout: 260 seconds)
16:55:56pamauryevildork: what is mood and opus ? And what do you mean by organize ? With the database, using a custom tagnavi file, you can navigate your music files as you want with the proper tags
16:56:01TheSevendoes the main thread have an event queue? if yes, where are the handlers?
16:56:37pamauryYou should have a look at the default handlers in misc.c
16:57:19TheSevenaha, misc.c. very descriptive file name.
16:57:32pixelmaevildork: do you want to organise purely on a per tag basis (in a database) or could you imagine to just stuff your music into different subfolders?
16:57:47CIA-8New commit by 03jdgordon (r27015): Make the title area go up one level in the lists again. This Doesnt work if you have the title in a seperate viewport in the sbs, but that isnt really ...
16:58:29pamauryTheSeven: that's the treasure file. If you don't know where to look, look in misc.c ^^
16:59:28CIA-8r27015 build result: All green
17:00:03 Quit bieber (Ping timeout: 264 seconds)
17:00:16 Join bieber [0] (
17:02:22TheSevenS_a_i_n_t: btw, keyclick is fine on r27014M
17:02:50 Quit Ed___ (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
17:03:11TheSevensector cache exclusive access collision while trying to create a folder in the root directory
17:04:19 Join lestatar [0] (
17:06:00 Part Zagor
17:06:52pamaurycan you identify who tries to open it ?
17:07:36TheSevenit'll probably be more interesting to know who has already locked it
17:10:18 Join DerPapst1 [0] (
17:11:01TheSevenhmm, the root directory is already opened non-exclusively when exclusive access is requested to add an entry
17:11:11TheSeventhat *might* be the very same directory handle
17:12:09 Quit DerPapst (Ping timeout: 240 seconds)
17:12:56lestatarhello all...don't wanna crowd the channel - i have a few rb v2 fuze questions please... :-)
17:13:25n1slestatar: just ask, if anyone can answer, they probably will
17:13:35lestatargreat thanks...
17:14:07lestatarfirst time here, so let me get this outta the way...all you rb project folks deserve medals...awesome awesome work :-)
17:14:35lestatarok, so v2 fuze works great for me so far, but at, online help is down
17:14:49kugeljust ask...
17:15:06lestatari loaded a few albums onto 16gb microsdhc card...not all songs showing up while browsing card
17:15:11 Join liar [0] (
17:15:27 Part retroj ("Channel buffer killed")
17:15:30lestatarhowever, all appear and play properly viewing artist file browsing for instance
17:16:18n1sso they don't show up in the dayabase?
17:16:19lestatarupdate of db seems to initiate [refreshing in background message], but after several hours the added songs still do not list under card browsing
17:16:35lestatarthey show up fine under artist
17:16:49lestatarnot while directly viewing card contents
17:17:07lestatarnot an issue with the card or rb necessarily as i already have 1500 files on the card
17:17:15lestatarjust these last few albums i added
17:17:16n1sso they show up in the database but *not* in the file browser?
17:17:20S_a_i_n_tare they "hidden"?
17:17:35S_a_i_n_tI mean the file properties
17:17:40lestatarnot hidden file attribs
17:17:53lestatarthough by practice, all my mp3 files are set for readonly
17:18:16n1slestatar: just to be on the safe side, set your file view setting to "all" and check again
17:18:41lestatarok, thanks...will give it a shot...
17:18:56lestatarstupid sandisk of is refreshing media again... :-(
17:19:15lestataranother question re: playlists
17:19:34n1sif that doesn't show them, they either are in another dir or something is very wrong
17:20:00lestatarusing of and a couple 3rd party apps, i was able to create mixed content playlists successfully on fuze, using songs from both main mem and card...
17:20:07lestataris this possible in rb?
17:20:33S_a_i_n_tIIRC, yes.
17:20:35lestatarand is it possible for rb to read/import .pla playlists i have previosuly created with above mixed contents?
17:20:46n1sno, m2u only
17:21:17lestatarah, ok...thought i skimmed somewhere in manual that import could be done...guess only for m3u
17:21:46lestatarno worries, i should create my mixed content playlists directly on the fuze/rb?
17:22:09S_a_i_n_tprobably easiest.
17:22:42*TheSeven is running into a problem
17:22:55S_a_i_n_tTheSeven: Try walking ;)
17:22:55lestatarok, thanks again...will def review manual and try all suggestions. you folks are great :-D
17:23:07lestatarcheers all :-)
17:23:51TheSevento fix that properly, we'll probably need a mutex protecting directory access
17:25:51TheSevenor just write with non-exclusive access
17:30:54TheSevensomehow I'm losing the settings during shutdown
17:30:59gevaertspamaury: yes, I've seen that too
17:31:07 Quit evildork (Quit: CGI:IRC)
17:31:39pamaurygevaerts: seen what ?
17:31:57gevaertswhat you asked me two hours ago :)
17:32:31JdGordonTheSeven: probably means the disk idle callback thingamy isnt being called
17:33:19pamaurygevaerts: is this normal ? Shouldn't the code unmount everything on usb plug ?
17:33:58 Join Jaigoda [0] (
17:34:02gevaertspamaury: in my not very informed opinion, yes
17:34:19gevaertsNot unmounting seems like asking for trouble sooner or later
17:34:20pamaurythe same for me :)
17:38:48 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
17:39:43*TheSeven wonders how the cache can cause settings not to be saved
17:39:43 Quit Jaigoda (Quit: CGI:IRC (EOF))
17:39:47 Join webguest44 [0] (
17:40:14 Quit webguest44 (Client Quit)
17:40:25*kugel grumbles
17:40:43JdGordonTheSeven: if the cache stops the disk spinning up the callback wont be called
17:40:45kugelI would like to commit the CONFIG_PLATFORM patch but I'm running into trouble with dependency generation
17:40:58gevaertskugel: anything specific?
17:41:13saratogafunman: I've seen a few people report that their fuze hard locked when plugged into USB, and could only be recovered by letting the battery drain
17:41:39saratogadoes the Fuzev2 enter some particular mode when trying to reboot into the OF ?
17:41:56 Join petur [0] (~petur@rockbox/developer/petur)
17:42:22TheSevenJdGordon: shouldn't the settings be saved during shutdown anyway?
17:42:30JdGordonwell yes
17:43:02 Join stoffel [0] (
17:45:06kugelgevaerts: yea, font.h includes sysfont.h, but dep gen doesn't see that with my patch (so sysfont.h is created too late)
17:45:16gevaertsah, fun
17:45:31kugelit looks to me as if dep gen would take config.h into account. I assume it works in svn because -DSIMULATOR comes from the command line
17:45:41 Join bertrik [0] (
17:46:00kugelwould *not*
17:46:13gevaertsIs this in gsoc-app-target?
17:46:21kugelno, gsoc-ifdef-cleanup
17:46:32 Quit bertrik (Client Quit)
17:47:45 Join bertrik [0] (
17:48:03kugelnot quite up-to-date, I can push the sync'd one
17:49:03gevaertsthat might help
17:49:28 Join FlynDice [0] (
17:49:51 Join funman [0] (~fun@rockbox/developer/funman)
17:50:01kugelhm no, even moving the include sysfont.h out of the ifdef doesn't add it to the dep list
17:50:36bertrikFlynDice, could you test FS #11423 for me on your ams targets?
17:51:18kugelhm, push doesn't work
17:51:45FlynDicebertrik: Not immediately but some time today should work.... what am I testing for?
17:52:02funmanbattery life has increased a bit on clipv1 : 20 minutes or so
17:52:54kugelgevaerts: I needed −−force for some reason, so pulling won't work for you I think (just re-checkout the whole branch)
17:53:23funmanperhaps because of r26953/r26954; but then I hope we see that it's not causing all those skips/glitches in audio playback
17:53:23bertrikFlynDice just see if the radio still works (tunes successfully)
17:53:51FlynDicebertrik: sure
17:54:33lestatarhi back...the one with the missing files under microsd file view
17:54:34kugelhm, the bug is somewhere else
17:54:42lestatarchanged settings to file, show all...
17:54:47kugelfont.h doesn't add sysfont.h dependency in svn too
17:55:06 Quit GeekShadow (Quit: The cake is a lie !)
17:55:12lestatarviewd the card...a few songs from addition last night still do not appear
17:55:20gevaertskugel: I'm looking at any sim build I guess?
17:55:39lestatarbut again show fine under artist browing... any more ideas? :-/
17:56:11S_a_i_n_tlestatar: How is the dir structure organised?
17:56:19S_a_i_n_ttoo many files in one dir perhaps?
17:56:25lestataron card, music/tunes
17:56:27lestatarthat is it
17:56:43S_a_i_n_ttry sorting into subdirs.
17:56:49gevaertsor increase the limit
17:57:14S_a_i_n_taha! bingo!
17:57:28lestatarmy max entry was already 3000
17:57:36lestatari am nowhere near that
17:57:41 Quit pamaury (Remote host closed the connection)
17:57:42S_a_i_n_tand, does it exceed 300 entries?
17:57:46lestatarprolly like 1500-1700 files on card
17:57:48S_a_i_n_toh, ok.
17:57:56pixelmathey should still show up, just not appear until you scroll down enough
17:58:19lestatarbeen looking pretty carefully since i noticed this...
17:58:30lestatar1 album has like 12 songs
17:58:38lestatarin artist view they all appear...
17:58:41lestatarall play properly
17:58:56lestatarjust odd in microsd view only like 4 songs appear in list
17:59:05pixelmathat sounds like you are browsing the database?
17:59:08lestatarand i would hate to do subdirectories...
17:59:12S_a_i_n_tI find it very odd that they show in the database, but not in the filebrowser :/
17:59:36lestataryes, am new, but this seems odd to me as well
17:59:48lestatarjust increased limit to 5000
17:59:57lestatarshould i do update of rb db anyway?
18:00:27lestatarwhat should directory cache setting be?
18:00:29gevaertsif you have less files than the limit, increasing the limit isn't going to help
18:00:45gevaertskugel: "font.h doesn't add sysfont.h dependency". Where?
18:01:13lestatari notice when i browse the card, a message says dir buffer that a clue?
18:01:16kugelI'm getting compile errors in pdabox (which includes plugin.h, and font.h from there)
18:01:17gevaertsI mean, as a dependency of what file?
18:01:32JdGordonlestatar: are the filenames pretty long?
18:01:32gevaertsah, ok. I'll look at that one
18:01:47JdGordonthat message does suggest uppding the max files setting
18:01:48kugelfrom /home/kugel/rbdev/rockbox-git/apps/plugins/pdbox/PDa/src/d_arithmetic.c:11:
18:02:08lestatarnot crazy long, but some yes, perhaps
18:02:29lestatarthough again, i already had plenty of files viewable under card browsing with even longer filenames
18:03:03lestatarthis is a minor issue...i just find it odd is all... :-)
18:04:23 Quit bertrik (Quit: >home)
18:04:27kugelgevaerts: FIRMLIB depends on sysfont.[oh] as per firmware.make, maybe there's the problem
18:05:18kugelsysfont.h gets created with make bin
18:05:30TheSevenhow does that storage idle callback work?
18:07:27***Saving seen data "./dancer.seen"
18:08:01 Quit Rob2222 (Ping timeout: 272 seconds)
18:08:19 Nick fxb is now known as fxb__ (
18:08:33 Join Rob2222 [0] (
18:08:53gevaertskugel: what's interesting is that othe files in pdbox/PDa/src do have sysfont.h in their dependencies
18:09:13kugelhm, I haven't looked after these files
18:11:20 Quit bieber (Ping timeout: 260 seconds)
18:11:32 Join bieber [0] (
18:13:04kugelgevaerts: d_arithmetic.c should depend on m_pd.h (the only file it includes) but it doesn't
18:13:20kugelthe files included by m_pd.h get lost
18:14:53gevaertskugel: I think I see what's going on
18:14:54kugelgevaerts: ha! FIXEDPOINT is not defined in dep gen, but during compilation
18:15:20gevaertsNow how to fix that...
18:15:24TheSeveni'm explicitly calling call_storage_idle_notifys(true); before flushing the cache, and the settings are still not saved!
18:16:03kugelpdbox.make adds it to the command line
18:17:32kugelgevaerts: I have no idea how dep gen works :(
18:17:43 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
18:17:45kugelI assume it takes config.h and runs gcc -E through all .c files?
18:18:15kugelno, not gcc -E, but it collects their #includes
18:18:54 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
18:19:06 Join webguest51 [0] (
18:19:20funmansee mkdepfile in tools/functions.make
18:20:36 Join DerpDerp [0] (48cc921f@gateway/web/freenode/ip.
18:20:49DerpDerpI did this right.
18:20:56DerpDerpThis is the IRC for rockbox, right?
18:21:20 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
18:21:42 Quit webguest51 (Client Quit)
18:21:45 Join G4Oblivion [0] (
18:21:46DerpDerpWell, if anyone would be willing to help me out, that'd be great.
18:22:01funmannot if you don't tell the problem
18:22:10TorneAsk your question, don't speculate on whether people will help you :)
18:22:54DerpDerpFair point. So, I unstalled rockbox on my Sansa Fuze v2 today, and it was working fine right up until I plugged it into the USB port, at which point, the screen went black.
18:23:06DerpDerpIt hasn't turned back on since, so I'm wondering if there's anything I can do.
18:23:26Tornedo you mean "installed"? :)
18:23:36kugelgevaerts: that bug is in svn too, but I wonder why my patch exposes it
18:24:09gevaertskugel: bad luck I'd say
18:24:40kugelI guess I better wait for Zagor to look at it?
18:24:41DerpDerpWell, used the manual install instructions I found using mkamsboot.
18:24:49TorneDerpDerp: it's supposed to reboot to the original firmware when you insert the USB cable.
18:24:56gevaertsI hope to have a fix in a few minutes
18:25:02kugeloh great
18:25:06Tornewhere did you get your versoin of mkamsboot?
18:25:27DerpDerpFrom this link:
18:25:31 Join pamaury [0] (
18:25:31 Quit pamaury (Changing host)
18:25:31 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
18:25:31Tornealso, have you tried holding the power switch for a really long time
18:25:43Torneif it's crashed you may need to hard reset it that way, i believe it can take 30-40 seconds
18:25:48DerpDerpYeah, I held it down for upward of a minute.
18:26:41Tornei assume plugging/unplugging the cable further doesn't do anything either?
18:26:45 Quit G4Oblivion (Quit: CGI:IRC (Ping timeout))
18:26:49DerpDerpSadly, no.
18:27:15Tornedo you know what version of Rockbox you installed?
18:27:42funmanDerpDerp: see the last line
18:27:43bertrikPossibly there's still something running on the fuze and you need to wait until the battery discharges completely
18:28:11TheSeveni just can't manage to make that stupid ipod save it's setting with the cache enabled
18:28:17DerpDerpI installed the most recent build, I think.
18:28:32TorneThe most recent build changes many times per day, though..
18:29:17DerpDerpAh, I see. Well, I just did this a couple hours ago using the build I posted up there.
18:29:19Torneanyway, yes, you might have to wait for it to discharge before you can make it work again, unfortunately
18:29:59DerpDerpGotcha. Thanks for the help! What usually causes something like this to happen?
18:30:19funmanIf you find a probable cause please tell us because we have no idea
18:30:26DerpDerpWill do!
18:30:31funmanalthough plugging USB cable seems to be a somewhat common factor
18:31:10DerpDerpIt was turned on when I plugged it in, so next time (if there is a next time for this particular player), I might power it down first.
18:31:40funmanwhat does 'stmfd sp!, {r0-r2}' do exactly: sp -= 3; sp[0] = r2; sp[1] = r1; sp[2] = r0; ?
18:32:28funmanand sp will point to the last item pushed on the stack : r2 ?
18:32:49TheSeventhe other way round
18:32:58TheSevenr0 will be at the lowest address IIRC
18:33:23TheSevenbut it's been some months since i last touched that kind of thing ;-)
18:33:43funmani have a memory dump of c200v2, i try to unwind the stack myselfd
18:33:51Tornelowest register goes to lowest address
18:34:14CIA-8New commit by 03gevaerts (r27016): Remove plugin-specific defines from the pdbox makefile, and add them to m_pd.h, which is the global pdbox header. This makes sure that the defines are ...
18:34:54funmanso sp always points to lowest register pushed?
18:35:04Torneif it's not been moved to make room for locals, yes
18:35:19kugelgevaerts: I had that idea as well, I hoped their would be a way to not ignore *.make
18:35:32gevaertsno easy way as far as I can see
18:35:38funmanok thanks!
18:36:07CIA-8r27016 build result: 24 errors, 0 warnings (gevaerts committed)
18:37:58 Quit DerpDerp (Quit: Page closed)
18:39:42 Join pamaury_ [0] (
18:39:58CIA-8New commit by 03gevaerts (r27017): Fix include file
18:40:01TheSevenpamaury: uh oh. file system corruption.
18:40:44pamaury_ah, bad news
18:41:50CIA-8r27017 build result: All green
18:42:03pamaury_TheSeven: is it due to the new cache system ?
18:42:26TheSeventhe last sector of every file written contains garbage
18:42:39TheSevenand it looks like some fat changes didn't get committed
18:42:41*pamaury_ thinks that have the simulator read/write into a file would be of great help here
18:43:07 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
18:43:36 Quit pamaury (Ping timeout: 264 seconds)
18:43:40kugelgevaerts: sim compiles now
18:43:41pamaury_Are you sure there are no hanging buffers, that everything is committed to the cache ta leats ?
18:43:50TheSeveni doubt this
18:43:59TheSeventhe problem is most probably somewhere in file.c
18:44:02 Quit Strife89 (Quit: Reboot to Windows.)
18:44:51pamaury_Did you check that rockbox flushes everything at poweroff ?
18:45:02TheSeveni enforced that
18:45:17CIA-8New commit by 03bertrik (r27018): Fix FS #10474 - FM tuner on Gigabeat S fails to tune last frequency on start
18:45:21TheSeventhat stuff should actually have been committed when the file was closed
18:46:04pamaury_Are you sure all the files are closed ?
18:46:15TheSeveni hope so
18:46:32 Join Strife89 [0] (
18:46:53CIA-8r27018 build result: All green
18:46:55TheSeveni just copied 43 music files using the file browser, if those wouldn't have been closed, it would have ran out of handles for sure
18:46:59pamaury_And the garbage at the end happens for every file written ?
18:47:39funmanTheSeven: plugins check if any file is still opened when they exit (but file browser isn't a plugin)
18:48:18pamaury_funman: TheSeven spent hours fixed leaking file handles in the code :)
18:48:42 Part Viaken ("WeeChat 0.2.6")
18:48:56pamaury_funman: I think uchida disapprove your last commit regarding the text viewer
18:49:12pamaury_He went on channel but you weren't online
18:49:14funmanpamaury_: thanks, i am subscribed to the flyspray task so i saw it
18:49:30funmani missed him by a few
18:49:37TheSevenAnd the garbage at the end happens for every file written ? < yes
18:50:14pamaury_TheSeven: can you put your code on FS or on pastebin. I can have a look at it
18:50:14funmantext_viewer crashes in gui_statusbar_draw()
18:50:52 Nick pamaury_ is now known as pamaury (
18:51:01 Quit pamaury (Changing host)
18:51:01 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
18:51:35 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
18:52:00TheSevenI'll commit some of it (fixes for bugs that surfaced while developing it) first
18:52:25 Join domonoky1 [0] (
18:52:25pamauryyes you are right to do so, you did a great job at tracking this issue it seems
18:52:45funmanJdGordon: gui_statusbar_draw is an awfully large function :/
18:52:53TheSeveni'm not sure if i should commit that font one though
18:53:04CIA-8New commit by 03kugel (r27019): Rockbox as an application: Replace many occurences of #ifdef SIMULATOR with #if (CONFIG_PLATFORM & PLATFORM_HOSTED) (or equivalently). ...
18:53:12pamauryavoid it for now
18:53:16 Quit domonoky (Ping timeout: 265 seconds)
18:53:35kugelnext commit will add the application target
18:53:38pamauryThat would probably cause a major confusion + bug reports + hangry people
18:53:40 Quit swilde (Remote host closed the connection)
18:54:28CIA-8New commit by 03theseven (r27020): Fix ata_idle_notify on iPod Nano 2G
18:55:13CIA-8r27019 build result: 72 errors, 88 warnings (kugel committed)
18:55:41 Join Jerom [0] (~Jerom@
18:57:00CIA-8r27020 build result: 72 errors, 88 warnings (theseven committed)
19:00:57 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
19:08:15JeromI like the new feature of this bot
19:08:27Bagder808 points!
19:09:01funmaninfo gcc says : `-O' also turns on `-fomit-frame-pointer' on machines where doing
19:09:04funman so does not interfere with debugging.
19:09:35 Join webguest43 [0] (
19:10:24CIA-8New commit by 03kugel (r27021): Fix yellows and reds (one was a type, the other was rather incorrect before already)
19:10:35 Quit webguest43 (Client Quit)
19:10:38 Join webguest57 [0] (
19:10:46funmanon ARM i don't see the frame pointer being saved with -O, I don't understand how doing so doesn't interfere with debugging
19:10:51 Quit webguest57 (Client Quit)
19:11:16 Join G4Oblivion [0] (
19:12:12CIA-8r27021 build result: 123 errors, 6 warnings (kugel committed)
19:12:29G4OblivionHello, I need help getting rockbox source. I'm trying to download with cygwin. I keep getting the .sig error.
19:12:37JeromNice :D
19:12:45G4OblivionI tried cmd with -x and nothing happened
19:13:09S_a_i_n_tG4Oblivion: That is explained in the wiki IRC
19:13:18S_a_i_n_tadd the -x to the .exe
19:14:37S_a_i_n_tsorry, to the shortcut to the .bat file
19:15:05funmanno dev have been using Sansa AMSv1/AMSv2 recently?
19:15:17*kugel uses it all day
19:15:52G4OblivionLike this "setup.exe -x"? I'm using RB wiki.
19:16:22funmankugel: with a build >= r26954 ? no songs skipping or weird audio glitches? which model is it?
19:16:30*kugel wonders why he even changed backlight.h
19:16:44kugelfunman: my build on it is older I think
19:16:45 Join M3DLG [0] (
19:17:08funmansaratoga and a lot of people report problems, i think r26953/r26954 cause them
19:17:13 Quit DerPapst1 (Quit: Leaving.)
19:17:25funmanbut i don't want to believe it because test_disk was fine :'(
19:17:35Bagder1236 points and we have a HIGHSCORE!
19:17:45Jeromfuzev2, r26981 I had no problem this afternoon
19:18:00funmanTheSeven: can cached sectors be read/written before storage_(read/write)_sectors() return ?
19:18:00S_a_i_n_tG4Oblivion: Easiest way I found to doit was make a shortcut, right click it, then change the "target to include " -x"
19:18:41ranmafunman: I haven't been using it to listen to audio lately ^^;
19:19:01G4OblivionOk. Thanks, I will try it.
19:19:04bertrikfunman, I haven't seen a problem on my clipv1 and clip+
19:19:06funmani dump/clean the dcache for the provided buffer before doing the transfer so it should be safe unless some code loads/store to/from this buffer while the transfer is in progress
19:19:13kugelfunman: I used invalidate_dcache_range (not clean_) in FS #10805, I can't remember audio glitches
19:19:25funmanwith mp3 i can hear one glitch per album perhaps
19:19:52funmanbut given the weird problem with text_viewer and some ipods, eabi transition could have broken something else
19:20:45funmanranma: btw do you have an idea on how gdb is unwinding the stack without the frame pointer?
19:21:03funmanperhaps it uses some hints from the .elf file?
19:21:40funmansee FS #11399, the backtrace I found is way off from what gdb tells
19:22:25 Quit amiconn (Disconnected by services)
19:22:27 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
19:22:30G4Oblivionugh, its doing the same thing cmd does. I start it but nothing pops up. I don't see it under processes either. without -x its fine.
19:22:42G4OblivionI will try again tomorrow when I have more time. Thanks for the help.
19:22:45 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
19:22:49funmanr11 is #0 so i wonder how it gives a backtrace at all
19:23:16funmansaratoga: have you run test_codec again since the PLL@248MHz commit was reverted?
19:23:46saratogano, should i?
19:24:12S_a_i_n_targh, G4Oblivion: Are you using " -x" or "-X"
19:24:22funmanS_a_i_n_t: i can test, reading ABI / rb forums I'm not sure if the problems are fixed; and also someone reported glitches with Fuzev1
19:24:28funmansaratoga: ^
19:24:49S_a_i_n_ta: it needs a space, and b: it needs to be a capital X
19:24:59CtcpPing from gevaerts!~fg@rockbox/developer/gevaerts
19:25:19S_a_i_n_t"C:\Cygwin\Local_Package_Directory\Setup\setup.exe -X" is my example, woks fine.
19:25:27funmansaratoga: did test_codec crash or did you just play the .ape file normally?
19:25:36saratogai played them normally
19:25:44saratogaape didn't crash, it jus tplayed static after a while
19:25:51funmanthey won't fit in audiobuffer anyway for test_codec
19:26:01ranmafunman: I'd assume it's using debug information from the .elf
19:26:27funmanperhaps we can use -O with -fno-omit-frame-pointer
19:26:38kugelfunman: very possible, you can run gdb without debug info if you give it the .elf file
19:26:52CIA-8New commit by 03kugel (r27022): I shouldn't have touched backlight.c at all :\ Also fix the yellow in peakmeter.c
19:28:02 Quit G4Oblivion (Quit: CGI:IRC (Ping timeout))
19:28:33 Join G4Oblivion [0] (
19:28:37CIA-8r27022 build result: All green
19:29:24G4Oblivionoh, wow. My bad, I forgot to use a capitial X. Thank you very much.
19:29:33S_a_i_n_tNo problem.
19:30:02G4OblivionI tried that with CMD and it didn't work so I thought it was supposed to be a lower case.
19:30:30S_a_i_n_tI actually don;t need to add the -X at all, I have no idea why some people do.
19:30:49ranmafunman: BTW;revision=26946 was causing me problems, I haven't checked yet if it's the buggy mps change or the switch to the _range() variants
19:30:54S_a_i_n_tI just checked for your benifit that it indeed works fo rme.
19:30:59funmansaratoga: i'm testing ape (starting from c1000)
19:31:02funmanon fuzev2*
19:31:30funman-> noise on c1000 after 50 seconds
19:31:56 Join dfkt_ [0] (~dfkt@unaffiliated/dfkt)
19:32:26funmanranma: why is mps buggy?
19:32:35 Quit G4Oblivion (Client Quit)
19:34:59ranmaIt depends on the endpoint, you were using the uninitilized counter value
19:35:16ranmai.e. it should be in the for loop
19:35:21funmanoh right i see it ..
19:35:47 Quit dfkt (Ping timeout: 265 seconds)
19:36:25 Join playing-the-fool [0] (
19:36:54ranmaAnd my "I wonder why the compiler didn't warn" comment is stupid, of course the build system doesn't see it since it's not enabled by default yet :)
19:38:59funmani should've paid attention to it anyway :/
19:39:37funmanif i undefined STORAGE_WANTS_ALIGN i don't get white noise (so far), but it can also be due to the aligned path being taken less often
19:40:15saratogafunman: did you think the current AMSv2 volume patch was ok?
19:40:57funmani should ask you, not the reverse :p
19:41:10funmani see nothing wrong with it
19:41:11saratogawell i think its ok, but what do i know
19:41:28saratogaok i'll commit it later and see if anyone complains
19:41:43 Quit playing-the-fool (Quit: Ухожу я от вас)
19:41:48saratogaFWIW i think volume could probably be increased on AMSv1 using the same method, but i'm not sure and people aren't complaining
19:42:05saratogai remember dfkt saying volume on the clipv1 was lower then the OF when he tested
19:42:34Jeromsaratoga: Are you increasing the current volume on AMSv2 ?
19:42:53JeromMay I ask why ?
19:43:04saratogawhy not?
19:43:13JeromNot deafen people
19:43:18funmanit'll only increase maximal volume, you're free to make it lower
19:43:41JeromI'm currently at -20 for normal usage...
19:43:42saratogaif you don't want loud volume don't turn it up
19:43:59JeromBut well not my problem ^^
19:44:34funmani think i'll leave this probable SD bug in trunk while TheSeven is messing with storage
19:44:46saratogai don't get the SD bug
19:44:49saratogatest disk runs forever
19:44:56saratogabut playback screws up in seconds
19:44:56TheSevenfunman: I'm not about to commit anything soon
19:45:10funmanTheSeven: well you might spot some bugs in the process
19:45:11TheSevenif whatever you fixed is in the target tree, it won't collide
19:45:29funmansaratoga: i commented out STORAGE_WANTS_ALIGN and it's now playing c4000 ok
19:45:48funmanTheSeven: no no, i added bugs, not fixed them :P
19:45:54saratogaahhh so its a playback/buffering.c bug?
19:46:10funmansaratoga: no i think it's a race condition
19:46:11dfkt_saratoga, that was a very old test with RB on the clipv1, and yes, RB was quieter than the OF... should i repeat it with a new build?
19:46:20 Nick dfkt_ is now known as dfkt (~dfkt@unaffiliated/dfkt)
19:46:34saratogadfkt: if you want, I'm not sure if its even possible to increase :)
19:46:41funmani can play back a full album (in mp3) and hear only 1 small glitch (repeated pattern or skipping some seconds) on the full time of the album
19:47:08saratogathe clipv1 uses a different DAC, but it looks to me like its not set to max volume, though maybe it doesn't go any higher even if set higher
19:47:47saratogaFWIW it seems to me like the clipv1 is quite loud already
19:47:49funmanc5000 isn't realtime :o)
19:50:28n1sit is... on the beast
19:50:53 Join Buschel [0] (
19:51:42funmann1s: this is on fuzev2 (armv5/240MHz)
19:52:21Buschelcan an as3525 guru take a look at FS #11419 ?
19:52:25n1sthen your observation is probably correct :)
19:52:56funmanBuschel: i think it's the problem we were just talking about
19:53:38Buschelah, ok :o)
19:54:21 Join DerPapst [0] (
19:56:03funmannot sure what I should do then, perhaps revert and put these in a patch on flyspray?
19:56:39 Quit CGL (Remote host closed the connection)
19:57:15 Quit DerPapst (Client Quit)
20:01:12dfktsaratoga, actually, OF was as loud on max volume as RB was on +6 (and distorted accordingly) -
20:03:39 Join Buschel_ [0] (
20:04:49 Join itcheg [0] (
20:06:46 Quit Buschel (Ping timeout: 248 seconds)
20:07:29***Saving seen data "./dancer.seen"
20:09:47 Quit itcheg (Quit: CGI:IRC (Ping timeout))
20:11:55Strife89What happened to the links to voice files on the Current Build page?
20:12:29kugelfunman: that would probably be better
20:12:38Strife89Or rather, the Daily build page.
20:12:47kugelfunman: add the patches to FS #11419
20:13:02kugelfunman: err, FS #10805
20:13:34funmani should check these patches and find what was wrong then
20:14:24kugelfunman: I wonder if the cache needs to be invalidated after the transfer on reads?
20:14:44funmanafter or before shouldn't matter afaiu
20:14:52funmanas long as we don't access the range
20:15:07kugelalso, why isn't the cache handled by the dma code? any dma access needs cache coherency
20:16:08funmanhm yes it makes sense
20:19:16 Quit mc2739 (Ping timeout: 260 seconds)
20:19:53funmani think we should manage to make Torne use his clipv2 more so he gets to fix all the problems
20:20:49pixelmaStrife89: follow the link to the users' ml ;) Manuals were affected too hence the topic... both were fixed today but will only be back to normal tomorrow
20:21:09TheSevenhm, splashf()ing a lot during file i/o doesn't seem to be a good idea
20:21:13 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
20:21:28funmanTheSeven: draw something in a corner of the screen and lcd_update_rect() should be faster
20:21:30 Quit kugel (Remote host closed the connection)
20:21:43TheSeveni doubt that this would have helped much
20:21:47 Join stripwax [0] (
20:21:51TheSevenit somehow messed up the font loading
20:21:57Strife89pixelma: Ah, didn't see that ... For some reason XChat is showing the tail end of the topic. :/
20:22:04Strife89pixelma: Thanks, though. :)
20:22:10TheSeveni ended up with a main menu where only about half of the letters were displayed
20:22:15TheSevenlooked funny :-)
20:24:32CIA-8New commit by 03bieber (r27023): Theme Editor: Fixed some compiler warnings
20:26:28CIA-8r27023 build result: All green
20:27:26*ranma just successfully wrote a file via usb using rockbox for the first time
20:28:27 Join Luca_S [0] (
20:28:35gevaertsranma: congratulations!
20:28:36Luca_Sranma: woot! congrats :D
20:28:39funmanranma: how large?
20:28:45*gevaerts remembers that feeling :)
20:29:07*gevaerts also remembers finding the corrupted bytes afterwards :)
20:29:32ranmafunman: 6084736 Bytes
20:29:42funmanwow :) congrats :)
20:30:37Jeromranma: please tell me it's for AMSv2
20:31:46 Quit funman (Quit: free(random());)
20:32:05ranmaJerom: AMSv1 ;)
20:32:40JeromI wanna cry ^^
20:32:43 Quit stripwax (Quit:
20:32:50ranmamd5sum is fine too
20:33:04simonrvnoh damn. well, it's a start. :)
20:33:23Luca_SJerom: feel free to work on it :-)
20:33:32JeromI'm trying
20:33:59JeromBut well I'm a "normal" developer I don't work on embedded devices
20:34:05*TheSeven thinks he spotted the bug
20:34:17gevaertsJerom: you have to start with something :)
20:34:36JeromWell I'm reading pamaury's code and I can also make sens out of it
20:34:49 Join stripwax_ [0] (
20:34:55 Quit t0rc (Quit: Leaving)
20:35:55JeromBut It seems that the setup packet is never comming
20:36:03Luca_Shuh.. the patch for PLL@384MHz on AMSv2 was reverted?
20:36:13JeromYes Luca_S
20:36:28 Quit MethoS- (Remote host closed the connection)
20:37:06ranmaJerom: I'm wondering if the clocking is set up correctly in the v2 usb driver
20:37:25ranmaIf the 48MHz usb clock is wrong that would explain why it won't receive the setup packet
20:37:52JeromThe only interrupts that we are receiving is bus reset
20:37:55JeromAnd enum done
20:38:18ranmaWell, those don't depend on the usb clock
20:38:31ranmabus reset is just when both lines are pulled to 0
20:38:41Luca_Sthe revert message says it caused problems with uSD... I didn't experience them :/
20:38:51ranmaenum done is based on timeouts in the controller I think
20:38:53 Quit wombat23 (Remote host closed the connection)
20:39:06JeromThese people experienced problems
20:40:02 Quit halmi_ (Quit: halmi_)
20:40:18TheSevensweet :)
20:40:24Strife89I suspect I'm having issues with disk corruption, but might anyone care to decipher this error? :
20:40:24TheSevenpamaury: it's working
20:40:42Strife89"Prefetch abort at FE21A1DA (0)"
20:40:57JeromWell I cant try to copy some of the nano2g init code over to amsv2
20:40:58ranmaJerom: I was wonering that because I do "CGU_USB = 1<<5 /* enable */ | (CLK_DIV(AS3525_PLLB_FREQ, 48000000) / 2) << 2 | 2; /* source = PLLB */"
20:41:00Strife89This occured while booting Rockbox.
20:41:04pixelmaStrife89: is this an Ipod?
20:41:13Strife89pixelma: An iPod Color.
20:41:15ranmaAnd the amsv2 driver just sets the enable bit without setting up the divider settings
20:41:40Jeromranma you think I should try copying over your init line ?
20:41:46bertrikPLLB is not enabled by default as far as I know
20:41:47pixelmaamiconn reported unstability issues with recent builds on a Color too yesterday (IIRC)
20:41:54Luca_SJerom: I just read that thread. Rockboxed FuzeV2 have always been a bit picky regarding to uSD, I had some troubles myself at some point
20:41:55pixelmaor insability
20:42:12pixelmaStrife89: you know what I mean ;\
20:42:13ranmaJerom: I think it can't hurt to try, but of course the bit meanings might be different...
20:42:17bertrikWe do know now how to configure and enable it though
20:42:19Strife89pixelma: :)
20:42:24JeromOk will try it
20:42:27Luca_Smight just be that some wait need better tuning
20:42:35TheSevenhm, should i commit the "unload fonts during USB" fix?
20:42:50Strife89pixelma: I still suspect disk corruption. Earlier the bootloader couldn't even FIND rockbox.ipod.
20:43:16Strife89pixelma: I rebooted to Disk Mode and recopied the file.
20:43:19JeromTime to drain the battery again :p
20:43:21pamauryTheSeven: what was the problem ?
20:43:39Strife89And this time, I got the crash.
20:43:40pixelmasame thing reported by amiconn yesterday or the day before it
20:43:51TheSevenpamaury: me being stupid :)
20:44:17pamaurynothing unusual then :P
20:44:39S_a_i_n_tamiconn reported issues with a 5G Video IIRC
20:44:59 Join wombat23 [0] (
20:45:04Strife89pixelma: If it means anything, I'm still using the testing version of that bootloader from a couple of months back.
20:45:21n1sStrife89: a disk scan never hutrs of course
20:45:36CIA-8New commit by 03Buschel (r27024): Disable unneeded parts of mpc's file-I/O interface.
20:45:59S_a_i_n_tStrife89: It shouldn't make a difference, that bootloader made it into SVN
20:46:10Strife89n1s: Yeah, that was my next move.
20:46:19*TheSeven wonders why usb on nano2g is so unstable lately
20:46:38S_a_i_n_tTheSeven: I find it stable, but dog slow.
20:46:43Strife89Care to refresh my memory? Isn't it "chkdsk /f X:" ?
20:46:49Strife89Where X is the drive letter?
20:47:01S_a_i_n_tI mean, it doesn't drop out or anything (for me) but transfer is rather painfull
20:47:01TheSevenS_a_i_n_t: The drive keeps disappearing in the middle of a transfer
20:47:28CIA-8r27024 build result: All green
20:47:29n1sStrife89: i think most windows command line tools have this /? thing that tells you how to use them
20:47:41TheSevenS_a_i_n_t: how slow is it? which storage size? which flash variant?
20:47:44CIA-8New commit by 03jethead71 (r27025): iPod 3G: Ensure wheel repeats are not generated unless the same keycode is posted more than once in a row, regardless of velocity. Remove 'was_hold' ...
20:47:51TheSevenI'm getting 4-5MB/s on my 8GB nano2g
20:47:56Jeromranma: Where do you pull this 48Mhz ?
20:49:02S_a_i_n_tTheSeven: I'm getting about 3~4MBps on 4GB nano2g
20:49:14TheSeventhat seems to be OK
20:49:16S_a_i_n_tsame on the 2 and 8 GB ones
20:49:17TheSeventhe OF can't be much faster
20:49:25n1sStrife89: anyway, as pixelma said this sounds more like some fun bug, prefetch abort is usually some weird bug
20:49:26CIA-8r27025 build result: 19 errors, 79 warnings (jethead71 committed)
20:49:30ranmaJerom: 384MHz PLLB divided by 8
20:49:52ranmaYou may have to enable the pllb first (see usb-drv-as3525.c)
20:50:21S_a_i_n_tthe OF seems to easily do 4~5 to 5MBps flat, but, I'm really lucky to get 3~4MBps with RB
20:50:39S_a_i_n_tthat is the peak speed, it can drop to less than 1MBps
20:52:03Strife89n1s: So it would seem. chkdsk turned up some fragment in .rockbox, but that was it. I copied rockbox.ipod again and got the same result.
20:52:25Strife89I hope this bug doesn't affect the iPod Videos.
20:52:27 Join Xerion [0] (
20:52:37JeromYou got to love embedded devices, registers, registers some more registers ^^
20:53:28bertrikJerom, pamaury need help setting up the PLL?
20:53:57Strife89Might the prefetch problem have anything to do with the Database or dircache?
20:54:02JeromWell it seems that USB needs PLLB for it's clock
20:54:31JeromAnd it disabled by default so I copied some of ranma code over to see if I can enable it and receive something from the usb bus
20:55:17bertrik240 MHz from PLL A / 5 = 48 MHz too, or do you really need PLL B?
20:55:22pamauryI don't know. On my clip+ it is correctly enumerated and responds to bus reset. If the clock was badly setup, it would not work at all
20:55:49ranmapamaury: No, the bus reset and enumeration don't depend on getting the clock right I think
20:56:04ranmaBus reset is just pulling both data lines low, clock is not relevant
20:56:05Jerombertrik: Honestly I have no idea :p
20:56:28pamauryBut enum speed is likely to require the clock
20:56:29ranmaFor enumeration it'll probably just fall back to FS or LS if the clock is wrong
20:56:36bertrikbut enumeration (assuming you mean setting the USB device address) does send a packet, right?
20:56:37pamauryIt reports HS
20:57:10Jeromenumeration happens before setup
20:57:14pamauryno, the "enumeration" is the speed detection
20:57:16ranmaHmm, if it reports HS that might be an indication the clock is alright
20:57:47 Quit Buschel_ (Ping timeout: 276 seconds)
20:57:49pamauryIt reports HS something like 10 times and after linux fallbacks to FS but still I can't make it work
20:57:56pamaurytoday I will try to work without DMA, doing PIO
20:58:00 Join stripwax [0] (
20:58:29bertriksometimes it can also help to run without interrupts, just polling the interrupt routine
20:59:08n1sStrife89: as amiconn said it feels like something fundamental, and is probably either a bug in the new toolchain or a previously latent bug in rockbox, uncovered by the new toolchain
20:59:28 Join Buschel [0] (
21:00:00ranma just says it does a 'hs chirp' for enumeration, maybe it's too tolerant on that part?
21:00:07 Quit stripwax_ (Ping timeout: 252 seconds)
21:00:23pamaurybertrik: but polling what ? I would poll the status register, that is the one which triggers interrupt :)
21:00:34pamauryOk, let's try with the clock then
21:00:58 Quit Luca_S (Quit: CGI:IRC (EOF))
21:02:10 Join Luca_S [0] (
21:02:29pamauryThe OF, it if does stick with the init sequence, sets CGU_USB to 0x20 which means usb phy clock enable, div=1,clk=clk_main
21:02:40pamauryIF the as3525 datasheet applies
21:03:36pamauryHow can I change that ?
21:03:55TheSevenshould I commit that?
21:04:24Jerompamaury: I'm currently trying ranma's clocking
21:04:54S_a_i_n_tTheSeven: S_a_i_n_t doesn't see why not.
21:05:18ranmapamaury: Do we use the same PLLA clock as OF? Maybe the controller has an internal register for clock division?
21:05:34pamauryI don't know
21:05:35 Quit rasher (Read error: Connection reset by peer)
21:06:01 Quit Luca_S (Client Quit)
21:06:34pamauryThe PHY part of the controller is basically a mystery. No two controllers in the nature do the same thing :( The only thing I know is what I reversed engineered
21:07:24JeromI forgot to remove uSB card... OF taking ages to scan the sd
21:07:30Strife89Let's see what happens if I delete .glyphcache, .playlist_control, config.cfg, DATABA~8.TCD (looks like it got corrupted :/ ), and nvram.bin; then try to boot.
21:08:19pamauryIt does not seem to change something
21:08:28pamaury(I tried ranman clock setup)
21:08:46Strife89What do you know? It boots. :D
21:08:52bertrikpamaury, what clock frequency does the USB peripheral require?
21:09:05pamauryI don't know
21:09:35 Part watto
21:09:47bertrikI thought it usually is 48 MHz for USB2.0 devices
21:09:52 Join rasher [0] (
21:09:52 Quit rasher (Changing host)
21:09:52 Join rasher [0] (~rasher@rockbox/developer/rasher)
21:09:59bertrik*hi-speed devices I mean
21:10:10Strife89n1s: I think I just discovered something.
21:10:36gevaertsIsn't it also often 60MHz?
21:10:36 Join Luca_S [0] (
21:10:49pamauryYes but I can't be sure, I don't know what the usb controller *needs*
21:10:53Strife89I deleted some of the cache/temporary files from.rockbox, as listed three minutes ago.
21:11:05Strife89This time, Rockbox booted.
21:11:12Strife89(This is on the iPod Color.)
21:11:26pamaury48MHz seems to be common and sensical so let's stick with it
21:11:29Strife89As a test, I started a song and then changed themes.
21:11:38Strife89Then I shut down Rockbox and rebooted.
21:11:52pamauryI'll check with the usb analyzer if something changed
21:12:00Strife89It was successful.
21:12:08pamauryTheSeven: thanks, I'll have a look at later
21:12:13S_a_i_n_tStrife89: None of those files are present in the .zip install thogh correct?
21:12:21Strife89S_a_i_n_t: Correct.
21:12:23S_a_i_n_tsomething changed in their format?
21:12:39Strife89Or in the way Rockbox handles them.
21:12:56S_a_i_n_tNice catch
21:13:06pamauryJerom: any success ?
21:13:34pamaurybertrik: when did the main clock changed recently ?
21:13:45pamaury(on as3525v2)
21:13:50JeromIt changed forth than back again
21:14:12Strife89S_a_i_n_t: Spread the word, then? I think this warrants plenty of investigation.
21:14:15bertrikindeed, now it's back at 240 MHz again
21:14:36S_a_i_n_tThat's the awesome thing about IRC logs ;)
21:15:01n1sStrife89: ah, that could make sense, the new abi probably changes rules for struct layouts in memory so some of those binary files could get loaded wrongly
21:15:31 Quit stoffel (Remote host closed the connection)
21:15:51pamauryAnyway, I will try with PIO, see if I have more luck.
21:15:55n1sif amiconn reads the logs it would be interesting if he still has said files and could investigate
21:16:43*S_a_i_n_t wonders if Strife89 can pull said files out of the recycler
21:17:03n1sTheSeven: it (FS #11428) sounds like The Right Thing (TM)
21:17:14Strife89S_a_i_n_t: You're in luck, I backed up the entire .rockbox folder before updating it on the iPod.
21:17:39n1sStrife89: would be interestin gif you could work out which file(s) cause it
21:17:56TheSevenn1s: it will make the HID mode text look a bit ugly
21:18:21n1sworking and ugly > pretty and broken :)
21:18:39Strife89n1s, S_a_i_n_t: I dunno how well I could do that. I'll give it a try, but what say I package those files and send 'em to you?
21:19:06n1salthough i suppose we could preload needed glyphs for the usb screen...
21:19:12S_a_i_n_teasy to tell, try installing the build you backed up, then delete the files one by one.
21:19:21S_a_i_n_tor, one at a time, then try booting.
21:19:25n1sStrife89: i have no ipod 4g
21:19:44 Join vmwareuser [0] (
21:20:04Strife89n1s: Ah.
21:20:15Strife89S_a_i_n_t: Works for me.
21:20:16 Quit stripwax (Ping timeout: 245 seconds)
21:20:33Strife89I'll package a .zip with the files anyway, just in case.
21:20:58S_a_i_n_tyeah, be good to keep it aside to see exactly what broke, then why.
21:21:10 Quit BHSPitMonkey (Ping timeout: 265 seconds)
21:21:44 Join stripwax [0] (
21:21:54 Quit stripwax (Client Quit)
21:22:06Strife89Here's the .zip.
21:22:11Strife89On to testing.
21:22:29vmwareuserCan someone help me to get Rockbox to compile for a 5.5G iPod on the Vmware platform? It worked for me when I last tried a few months ago, but with the current svn revision the make script just throws heaps of "arm-elf-eabi-gcc: command not found" errors ...?
21:22:51TheSevenyou need to install the new (eabi) toolchain
21:23:04TheSevenwe've switched compilers some days ago
21:23:04Luca_Svmwareuser: I had the same problems, and after a few hours of trying, I simply downloaded an ubuntu vm off the net fails there
21:23:30 Join katg [0] (
21:24:15n1svmwareuser: run the script in the tools dir of the checkout (as root) and select the arm-eabi toolchain
21:24:53TheSevenany pp ipod owners around?
21:25:00 Quit Topy44 (Ping timeout: 260 seconds)
21:25:05Luca_Sand if it works, please report how you did it :P
21:25:06TheSevencould anybody have a look why the hardware keyclick patch is failing on pp?
21:25:07vmwareusernls: thanks a lot, I'll try that now.
21:25:27saratogayou'll probbaly have to install some other packages before it'll build
21:25:39TheSevenvmwareuser: also make sure to first remove the old toolchains from the PATH before running
21:25:51Jerominstall texinfo
21:25:54Luca_Ssaratoga: it failed saying 'please update config.guess'
21:25:55JeromYou will probably need it
21:26:06 Join kaylinsigswort [0] (
21:26:12TheSevenif you don't do this, the toolchains will seem to build properly, but will fail in odd ways while compiling rockbox
21:27:13saratogai didn't have to remove anything from the path when building under ubuntu FWIW
21:27:36TheSevensaratoga: but you probably didn't have the old arm-elf toolchain installed
21:27:55saratogayes I do
21:28:00TheSevenit failed for me (and several others) if it finds those while building the new ones
21:28:06saratogamy machine has been a build server for a number of years
21:28:07TheSevenubuntu karmic over here
21:28:22 Join Topy44 [0] (
21:28:31TheSevenhm, interesting
21:28:33vmwareuserTheSeven: removing the old toolchain on the vmware machine means removing /usr/local/arm-elf/bin from PATH?
21:28:54TheSevenyes, and deleting that directory is also a good idea
21:29:23TheSeventhe new toolchains will be installed to /usr/local/bin anyway
21:29:42S_a_i_n_tI was going to say that, not only did I need to remove the path, I needed to remove the old toolchain dir also
21:30:03n1sit failed for me too but i had a slightly older arm-elf-eabi in my PATH
21:30:04*S_a_i_n_t is glad TheSeven has this problem also.
21:30:18S_a_i_n_tthat'd be the third person I've seen have that trouble
21:30:33 Quit Luca_S (Quit: CGI:IRC)
21:30:39S_a_i_n_twoo! 4 then counting n1s
21:30:46S_a_i_n_tI'm not alone ;)
21:31:37n1sS_a_i_n_t: as long as the old toolchains aren't in your PATH you should be fine (i have tried that=
21:32:21S_a_i_n_tI needed to remove the toolchain dir as well.
21:32:26TheSevenn1s: I'm not completely sure, but it's very unlikely that they were in the path when I build the new ones, yet it failed
21:32:46TheSeveni ended up removing all *arm-elf*, *sh-elf*, ... stuff from /usr/local
21:33:22TheSevenactually the old ones were never in the path for the root account
21:33:24n1sbtw jhMikeS posted a thread in the forums with an updated vmware 'm not completely sure, but it's very unlikely that they were in the path when I build the new ones, yet it failed
21:33:24n1s<TheSeven> i ended up removing all *arm-elf*, *sh-elf*, ... stuff from /usr/local
21:34:32n1seh, sorry, not intended paste
21:38:33Jeromis the PLLA static or can it change ?
21:38:56saratogait can be adjusted but it doesn't reclock during normal use
21:39:39 Quit katg (Read error: Connection reset by peer)
21:43:16bieberJdGordon: If an identifier is specified in a Vi tag, does that make it hidden to begin with?
21:45:26TheSevenliar: What's the status of FS #10894?
21:46:53*FlynDice reads about uSD problems after adjusting PLLA freq and wonders if saratoga might have a AMSv2 with uSD to check frequencies on like he did for v1's
21:48:33amiconnS_a_i_n_t: I have no G5. I noticed the problem on my iPod Photo
21:48:52amiconnIt *might* actually be the database code
21:49:22Strife89The database code .....
21:49:29amiconnWhile a complete rebuild is meant to be that, there have been reports that it might not under certain conditions
21:50:06amiconnThe usual suggestion in such a case was to delete all .tcd files, then reboot and rebuild
21:50:10saratogaFlynDice: with the spec analyzer? would that tell us anything new? i think we have a good idea what the frequencies are
21:51:13amiconnPerhaps a somewhat broken .tcd file (in whatever way) makes the database code enter a weird state it won't recover from
21:52:29amiconnIt would explain the various weird effects I've seen. First a data abort at boot. Then, after rebooting and starting said rebuild (with the old .tcd files in place) neverending database scan with the statistics in the debug menu acting up
21:52:30FlynDicesaratoga: I don't think there's documentation for the divider we use to set the card clock rate, I just made an educated guess and it worked.
21:52:40n1sif that can be confirmed, isn't there some kind of version for these files that can be bumped?
21:53:10amiconnAfter reset, another try to rebuild the db finally worked
21:54:02amiconnPerhaps the db code needs some sanity checks when loading the .tcd files?
21:54:23S_a_i_n_twouldn't the .tcd files mean (in some cases, most often in fact, unnecessarily) that the db would be rebuilt on every update?
21:54:24amiconnIf there are pointers, a broken .tcd may potentially cause overwrites of arbitrary memory
21:54:35saratogacould use one of these to do it
21:54:43S_a_i_n_ts/wouldn't the/wouldn't versioning the/
21:55:19saratogathough is it that hard to pull apart a clip+
21:56:04*bertrik was about to mention that sniffer
21:56:09 Quit kaylinsigswort (Remote host closed the connection)
21:56:11amiconnStrife89: Eh, you also had a corrupted .tcd file (shortname only)?
21:56:25Strife89amiconn: Yep.
21:56:33*amiconn had that as well
21:56:55amiconnImo that's another hint towards the db code
21:57:09*S_a_i_n_t has seen his database go haywire when a .tcd was currupted.
21:57:10*TheSeven wonders why screendump could stkov
21:57:15bertriksaratoga, it mentions it sniffs SPI traffic, but I guess we can measure at least the clock rate with it
21:57:30S_a_i_n_tthe file included a ~ after it got chewed by whatever chewed it.
21:57:38saratogawell it gives you all the pins, so i could clip a probe to it and measure the clock rate
21:57:56amiconnThe ~ is part of the standard scheme for shortnames
21:57:58n1sS_a_i_n_t: it would trigger a rebuild only when the version was bumped
21:58:02saratogaall my equipment is analog anyway
21:58:23amiconnn1s: Versioning wouldn't protect from broken files
21:58:46Strife89S_a_i_n_t, n1s: I can't seem to reproduce the error. :/
21:58:49amiconnIt might still be a good idea though
21:58:55n1samiconn: true, but in this case it could have been used (if anyone would have anticipated the breakage)
21:59:22S_a_i_n_tStrife89: Even if you write the old backed up build back to the iPod?
21:59:33Strife89S_a_i_n_t: Yup. :/
21:59:40bertrikI guess the rockbox fund could pay for it (and pay for a full-size SD card sniffer too)
21:59:50n1sbut then i don't know how we would deal with runtime data
22:00:46Strife89S_a_i_n_t: I have a hunch that it had something to do with a theme failing to load, but that's all it is - a hunch.
22:01:13saratogaits 15 dollars i don't really mind, though it might be quicker to just pop open the player
22:02:11Jerompamaury: I got an out ep int !
22:02:12saratogaon the e200v2 you could get at the sd contacts just by opening the device
22:02:33pamauryJerom: \o/ how ?
22:02:44Jerom /* PHY clock */
22:02:46Jerom CGU_USB = 1<<5 /* enable */
22:02:48Jerom | (CLK_DIV(AS3525_PLLA_FREQ, 60000000)) << 2
22:02:50Jerom | AS3525_CLK_PLLA; /* source = PLLA */
22:02:55Jeromsry for copy pasting
22:03:46pamaurywhere does this 60Mhz comes from ?
22:04:01JeromSomeone said it sooner ^^
22:04:26gevaerts60MHz makes lots of sense as a byte clock on usb
22:07:06pamauryhum, doesn't work on mine, perhaps I have bad modification on the usb code. You are using my svn code ?
22:07:32***Saving seen data "./dancer.seen"
22:07:42 Join blu3t0oth [0] (
22:09:15 Join toffe82_ [0] (~chatzilla@
22:11:02*TheSeven is wondering if he should commit FS #11428 right away or wait for further input
22:11:06 Quit toffe82 (Ping timeout: 245 seconds)
22:11:39Strife89Depends on what it is. :)
22:12:04CIA-8New commit by 03bieber (r27026): Theme Editor: Reworked information passing among render functions, now loads all viewports and shows Custom UI viewport in blue
22:13:17TheSevenhm, nobody complains
22:13:33pamauryJerom: it works ! That's awesome
22:13:41pamauryI will finally be able to write this driver
22:13:46Jerom:) I'm happy to
22:13:54blu3t0othRockbox is awesome!
22:13:59CIA-8r27026 build result: All green
22:14:12JeromIt will be a good day when we can finally drop the OF
22:14:15pamauryThat's incredibly lucky that the svn code is ok, it would not work on my local version
22:14:24pamauryI need to figure this out
22:15:51pamauryhum, there was a as3525 clock change, I wasn't at head; perhaps it explain things
22:17:27vmwareuserTheSeven: I installed the new toolchain using tools/ (after I removed the arm-elf directory from /usr/local). I reran tools/configure, and now the make command fails after only two lines of output: "CC apps/action.c", followed by "/home/user/rockbox/apps/action.c:26:18: error: lang.h: no such file or directory". Any idea why this happens? I'm using a fresh SVN checkout of Rockbox.
22:17:32saratogahead is back to the way it was before I think
22:17:41pamauryOkay, I will commit my changes
22:17:45saratogavmwareuser: make clean
22:18:45vmwareusersaratoga: Oops. Thanks : )
22:18:53CIA-8New commit by 03theseven (r27027): Allow UI fonts to be unloaded (needed for FS #11428)
22:19:34CIA-8New commit by 03theseven (r27028): Unload fonts when entering UMS mode to close their file handles. Closes FS #11428
22:20:19pamauryJerom: you definitely need to be in the credits for that :)
22:20:28pamauryTheSeven: you finally committed it :)
22:20:36CIA-8r27027 build result: All green
22:20:47TheSevenalso a nice revision number :-)
22:22:17CIA-8r27028 build result: 10 errors, 1 warnings (theseven committed)
22:22:42 Quit n1s (Quit: Lmnar)
22:22:51TheSevenurgh. charcell.
22:24:59 Join binaryhermit [0] (
22:24:59CIA-8New commit by 03theseven (r27029): Fix charcell red.
22:25:33CIA-8New commit by 03pamaury (r27030): as3525v2-usb: add a few missing define for completeness, finish reorganization of the header
22:25:40CIA-8New commit by 03pamaury (r27031): as3525v2-usb: add support to derive usb clock from pllb, correct endpoint listing, simplify a few things.
22:25:48CIA-8New commit by 03pamaury (r27032): as3525-usb: change clock handling
22:25:55CIA-8New commit by 03pamaury (r27033): as3525v2-usb: definitely switch to this clock handling, it is the good one
22:26:45CIA-8r27029 build result: All green
22:27:13pamauryJerom: what is your real name ? So I can add you to the credits
22:27:45JeromJérôme Heil
22:28:14JeromDon't look up my name on google, it's not me :p
22:28:29CIA-8r27033 build result: All green
22:28:57pamaurydon't worry, it's just that we only put real name in credits
22:34:42 Part blu3t0oth ("Verlassend")
22:34:43CIA-8New commit by 03pamaury (r27034): Add Jérôme Heil to CREDITS for figuring out the usb clock problem on as3525v2.
22:34:50pamauryNow the hard work begins
22:36:51CIA-8r27034 build result: All green
22:37:45JeromI copied most of the code from usb-s3c6400.c over but now that you changed the defines ^^
22:38:28pamauryAt the end, we might merge the two drivers
22:38:31simonrvnwoohoo Jerom
22:38:58JeromThx \o/
22:39:29pamauryHowever, the two usb controllers have different versions. As we don't have the datasheet for the 2.2 nor the 2.6, we difference between them are unclear. The linux patch I have suggest they are really small but...
22:41:41 Join anewuser [0] (anewuser@unaffiliated/anewuser)
22:42:47Jerombtw I noticed a strange behavior from the controller, now that we ACK the interrupt it wait 5s before it sends the default reply
22:43:19 Join piggz [0] (~piggz@
22:43:47piggzhi..when the database is updating in the background, is there a way to tell its status, like, if its finished?
22:44:00pamauryWe need to write a proper handling for this
22:44:51Jerompiggz: When the small hd or sdcard icon disappear
22:45:30piggzJerom: beside the clock?
22:45:55JeromYes it show "disk" activity
22:46:12JeromWhen there is no more activity the database as finish its work
22:46:40 Quit Buschel (Ping timeout: 260 seconds)
22:46:47Jerompamaury: You don't wanna use a struct wakeup ?
22:48:13vmwareuserFor the record, the current SVN revision compiled just fine on the vmware image that is available for download on After deleting the old toolchain, running to install the new one and running a 'make clean' in the build directory, compilation worked without errors. Now I've got a current version of Rockbox with my beloved wikiviewer plugin installed. Thanks!
22:48:39pamauryI don't know what you call a struct wakeup but yes I will use a wakeup object. But for now, I need to write in/out handlers, see if it correctly handles controls transfers and also check both HS and FS
22:49:25piggzJerom: cool, cheers, its gone off now
22:49:35 Quit hebz0rl (Ping timeout: 276 seconds)
22:49:49 Quit vmwareuser (Quit: CGI:IRC)
22:55:41*TheSeven will go to sleep
22:55:52kisakhave a nice sleep
22:55:55TheSeveni hope there will be some comments in fs#11429 tomorrow :-)
22:56:30Jerompamaury: When the usb core request a endpoint, the direction is from the host point of view ?
22:58:16 Quit evilnick_B (Quit: Page closed)
22:58:22 Join Meekrab [0] (
23:02:09 Quit bmbl (Quit: Bye!)
23:05:12 Join detaos [0] (
23:05:23MeekrabWho wants to pick up the zune project again?
23:07:19kisakisn't the easiest way to use Hide-a-pod to port to the zune?
23:08:32kisakI mean ... noone appears to be active at this time, feel free to find where someone else left off and dig in
23:09:03BagderMeekrab: go ahead, lead the way
23:09:21Bagderafaik the zune was never even sold outside of north america
23:09:31MeekrabI was dicking with the ability to inject code into a running machine, but sofar i need a 4096bit RSA encrypted hash
23:10:36 Quit Topy44 (Ping timeout: 260 seconds)
23:12:43 Join Topy44 [0] (
23:13:17 Quit Jerom (Quit: Jerom)
23:14:09MeekrabIf I designed a simple yet powerful OPEN SOURCE DMP, would I be allowed to sell the hardware with rockbox?
23:14:31Bagderas long as you also offer the source code to it
23:14:39MeekrabWell yeah, hence open source
23:19:30 Join MethoS- [0] (~clemens@
23:20:03 Join Jerom [0] (~jerome@
23:20:40MeekrabBasically I have the idea for an ARM based unit with a 320x480 screen, 30gb HDD, SD expansion, AV out, and a common USB B cable output and custom dock port (dubbed the geekport)
23:21:21piggzthat would be an ipod video then? ;)
23:21:25MeekrabI'd call it the BeatBox
23:21:30piggz- the sd
23:22:20domonoky1Meekrab: maybe you want to look into what the lyre project is trying todo..
23:22:23MeekrabBasically it'll look like a zune with a bigass screen and buttons on the side of the unit
23:23:08 Join AzureWurk [0] (
23:23:15JeromMeekrab: You should look at this thread for a start:
23:23:18pamauryTheSeven: shouldn't your usb driver call usb_core_transfer_complete even on transfer timeout ?
23:24:38AzureWurkquestion: is it normal for the I2SO clock on the fuze v2 to drp from 240mhz to 5mhz when playing ogg files?
23:24:45MeekrabI project that it'll cost about $200, with your basic stuf
23:24:52Meekrablike usb, headphones, etc
23:25:05 Join fdinel [0] (
23:25:51MeekrabI'm thinking I'm going to base it around the zune, and use the toshiba HDD and familiar batteries except without the vendor lock in on firmware/Os
23:27:08Meekrabwould 64mb for buffering ram be overkill? after all the hdd does suck alot out of the battery on spinup
23:27:49JeromWell if you wanna use a hdd
23:27:58JeromThen I don't think it is
23:28:01dfkt<Meekrab> Basically it'll look like a zune with a bigass screen <−− the current zune hd has the same screen resolution /// <Meekrab> Basically I have the idea for an ARM based unit with a 320x480 screen, 30gb HDD <−− nowadays i guess 32gb of flash memory would be more useful
23:28:11BagderMeekrab: i'd suggest you join the people
23:28:50MeekrabHard drives have a longer running life
23:29:01Bagderwhat do you base that on?
23:29:17MeekrabFlash memory has limited writing cycles and is then pooched
23:29:27Meekrabhard drives can be popped out and replaced
23:29:35Bagderdo you know how many cycles?
23:29:37Meekrabplus their cheap :)
23:29:43Bagderdo you know about wear leveling?
23:29:51MeekrabA few billion, with wear leveling
23:30:00Jeroman 1.5" hdd isn't cheap imho
23:30:02Meekrab10 years of constant access on a regular basis
23:30:14MeekrabWho said anything about the little HDDs?
23:30:23MeekrabI was thinking SATA 2.5 HDDs :P
23:30:38Bagderhdds are in fact much more likely to fail
23:30:51simonrvnmore mechanical things to break
23:31:05JeromI broke my ipod 5.5G that way
23:31:08AzureWurkbut unlike hdd when flash cells fail they just cant erase data most tims, hdds die they are pooched and so is everything on them :P
23:31:21JeromSlipped off my hand...
23:31:23BagderMeekrab: but yes, lots of people consider building their own. for a while. then they realize the problems and stop... :-)
23:31:40MeekrabI guess Compactflash would be a viable replacement
23:31:45Meekrabit still uses the ATA spec
23:32:05AzureWurki would use a par o swapable micro ssd's
23:32:06Meekrab(sorry to be biased but I really love programming for ATA spec storage like SD and CF and regular HDD0
23:32:08ender`it's quite easy to find CF->IDE adapters of all sizes
23:32:20ender`but there's also SD->IDE adapters
23:32:32Meekrabbut why not build my machine around CF for main storage?
23:32:47ender`CF cards are huge
23:32:51Bagderbecause CF is old and buried?
23:33:02AzureWurkcf is a bit large for capacity
23:33:33Meekrabbut it has better hardware control in my opinion
23:33:48ender`CF spec is still being extended (IIRC, the last addition added SATA support)
23:34:17MeekrabIf I place a CF card on a zune, it takes up about 1/3 of its shapespace
23:34:24 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
23:34:42Meekraba 1.8HDD, it takes up 2/3 its shapespace
23:35:15MeekrabI'm basing my idea on two good facts
23:35:23AzureWurka drive like asus eeepc's use is alot smaller
23:35:35Bagderwell, pick whatever you want. Rockbox should be able to use it. good luck!
23:35:55MeekrabAverage End users want to have a customised player that looks cool and does exactly what they want, like play games, doom etc
23:36:56Meekrabthe other guy, who knows about the hardware will know what to do and probably took it apart lots of times, and he knows when and where to get replacment parts, for the storage it could be as easy as taking a trip to wally world and getting an even bigger CF
23:38:06Meekraband even though it is a Warranty nightmare, it'll be completely modular like a standard desktop PC
23:38:15Meekrabhence why I'd sell it as DIY
23:39:01AzureWurkimho after having replaced 4-8gb of the above for ppl they are what i would use for a pmp
23:39:29Meekraband I absolutely refuse to encorporate ZIF or ribbon cables of any sort in my machine
23:39:56AzureWurknaa those just snap in like ram
23:40:10Bagdercan we get back to Rockbox talk now?
23:40:14 Quit komputes (Remote host closed the connection)
23:40:24AzureWurkquestion: is it normal for the I2SO clock on the fuze v2 to drp from 240mhz to 5mhz when playing ogg files?
23:41:10CIA-8New commit by 03bertrik (r27035): Apply FS #11423 - Use udelay in AMS driver for FM radio I2C
23:41:38 Join Milardo2 [0] (
23:42:57CIA-8r27035 build result: All green
23:43:14 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
23:44:21Milardo2hey does anybody know what information if any is needed for the meizu port?
23:45:42AzureWurkwhats the addy to the revision page:
23:48:06 Quit M3DLG (Ping timeout: 248 seconds)
23:50:16S_a_i_n_trevision page?
23:50:52simonrvnrecent changes or whatever?
23:52:28S_a_i_n_tAzureWurk: I trust you recieved the custom Fuze V2 build I compiled for you? You may (or may not) notice it includes a few other patches, but there is nothing in there that is too drastic.
23:52:43 Quit detaos (Remote host closed the connection)
23:56:04 Join detaos [0] (
23:57:51S_a_i_n_tIt uses saratoga's volume patch (revision #4), shutdown in menu, a prettier graphic for the battery discharge curve in Debug, ...and a few more I can't remember. Most of the patches in my tree are iPod specific so won't show up in the build I gave you, but a few will so I thought I'd mention it in case you think "Hey! Where did 'X' go?!?" the next time you update your build.
23:58:09S_a_i_n_tAzureSky: AzureWurk: ^^

Previous day | Next day