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

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

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

#rockbox log for 2011-03-01

00:00:42 Join T44 [0] (~Topy44@
00:01:03 Join pamaury_ [0] (
00:01:23 Join Unhelpful_ [0] (~quassel@rockbox/developer/Unhelpful)
00:01:41 Join Robdgreat_ [0] (
00:02:27kugelthomasjfox: I'm assuming the shutdown code works for you?
00:02:46 Join linuxstb_ [0] (
00:02:46 Quit linuxstb_ (Changing host)
00:02:46 Join linuxstb_ [0] (~linuxstb@rockbox/developer/linuxstb)
00:02:56 Quit sideral (Remote host closed the connection)
00:03:08thomasjfoxkugel: Oh yes, I really tortured it. 25 shutdowns with sigaltstack and 25 with SDL threads.
00:03:29thomasjfoxkugel: Thanks for fixing it :)
00:03:30 Join sideral [0] (~sideral@
00:03:30 Quit sideral (Changing host)
00:03:30 Join sideral [0] (~sideral@rockbox/developer/sideral)
00:04:08 Join viv [0] (
00:04:28 Join n17ikh| [0] (
00:04:46 Part viv
00:05:06CIA-70New commit by thomasjfox (r29467): Move sleep timer code outside of PLATFORM_NATIVE ifdef so RaaA can access it ...
00:05:10pamaury_sideral: hello, I tried to see which files were left open on usb mount and could only get the currently playing file...
00:05:52sideralpamaury: I always get at least one more file: a font file
00:06:32 Join stripwax [0] (
00:06:41pamaury_actually, I wrote some code to print the file names and except if I did a mistake, the font file is never left open
00:06:51sideralAnd sometimes a third, which I haven't identified yet −− I guess it could be a DB-related file during DB update
00:07:21sideralmaybe the font file is specific to my theme?
00:07:21 Join crwll [0] (
00:07:21pamaury_I can provide you with the code do log the file names, if it helps identifying them
00:07:33pamaury_perhaps, I know nothing about fonts
00:07:33sideralpamaury: Yes, that would be nice
00:07:56 Quit froggyman_ (*.net *.split)
00:07:56 Quit Keripo1 (*.net *.split)
00:07:57 Quit pamaury (*.net *.split)
00:07:57 Quit tchan (*.net *.split)
00:07:57 Quit domonoky (*.net *.split)
00:07:57 Quit Topy44 (*.net *.split)
00:07:57 Quit robin0800 (*.net *.split)
00:07:58 Quit Unhelpful (*.net *.split)
00:07:59 Quit MagusG (*.net *.split)
00:07:59 Quit ehntoo (*.net *.split)
00:07:59 Quit YPSY (*.net *.split)
00:08:00 Quit linuxstb (*.net *.split)
00:08:00 Quit ender| (*.net *.split)
00:08:01 Quit crwl (*.net *.split)
00:08:02 Quit n17ikh (*.net *.split)
00:08:02 Quit Robdgreat (*.net *.split)
00:08:02 Nick n17ikh| is now known as n17ikh (
00:08:33 Join MagusG [0] (
00:08:35kugelsideral: that worked, cool
00:09:09pamaury_sideral:, you need to disable dircache and enable logf
00:09:26sideralkugel: Great to hear! Now the two of us can commit to the 3.8.1 branch like crazy :)
00:09:37CIA-70r29467 build result: 468 errors, 0 warnings (thomasjfox committed)
00:09:44kugelpamaury_, sideral: I know buffering doesn't close the last file if it's buffered partially. I tried to close it but it introduced noticeable audio glitches
00:09:49sideralThanks pamaury, I'll apply it to my build and report back
00:09:55mshathlonxp486 errors :D
00:09:58kugelsideral: AlexP can too :P
00:10:06thomasjfoxstill no high score :(
00:10:25***Saving seen data "./dancer.seen"
00:10:32AlexPThere is no 3.8.1 branch :)
00:10:46kugelyou need to seriously break some often included header to get a high score :)
00:11:13 Join saratoga [0] (9803c6dd@gateway/web/freenode/ip.
00:11:36 Quit linuxguy3 (Ping timeout: 250 seconds)
00:12:54 Join ehntoo [0] (
00:12:58 Join shai [0] (
00:13:01 Join robin0800 [0] (
00:13:10 Quit sideral (Remote host closed the connection)
00:13:20 Join froggyman_ [0] (~seth@
00:13:27 Join sideral [0] (~sideral@
00:13:27 Quit sideral (Changing host)
00:13:27 Join sideral [0] (~sideral@rockbox/developer/sideral)
00:13:31sideralAlexP: Is there not?
00:13:32kugelTheSeven: your fat commits created some warnings in the android build
00:13:37 Join linuxguy3 [0] (
00:13:47TheSevenkugel: hm? what kind of warnings?
00:13:51TheSevencan you paste them?
00:13:55AlexPsideral: No, just 3.8
00:14:02kugelswap16/32 redefined
00:14:07 Join Ypsy [0] (
00:14:15 Nick Ypsy is now known as YPSY (
00:14:15AlexPsideral: Any backports will normally go in 3.8, then a 3.8.1 will be tagged
00:14:22*TheSeven hasn't touched that
00:14:23CIA-70New commit by thomasjfox (r29468): Fix red
00:14:28sideralAlexP: I see
00:14:53thomasjfoxha, commit 29-468 fixes 468 build errors
00:15:21 Join Keripo [0] (
00:15:33sideralBTW, I cannot play back m4a files in my trunk-based build any more −− hope that's cause by post-branch changes
00:15:42sideralperhaps jhMikeS' buffering changes
00:15:58stripwaxI'm sortof following FS #11863 ... is that a bug in 3.8 or just svn?
00:16:57kugelthomasjfox: heh, nice one
00:17:16[Saint]kugel: The touch area "menu" goes to the WPS if there's music playing now?
00:17:23sideralstripwax: Both, AFAIK. Some fixes went into 3.8 already, and some were made after the branch
00:17:41kugel[Saint]: does it?
00:17:43[Saint]Or, did it always used to do that....?
00:17:54[Saint]it does, yes.
00:18:02[Saint]I swear this is a new thing ;)
00:18:10kugelif it does what the menu button does, it goes to the main menu and back to where you come from (if you press it in the main menu)
00:18:26kugelI suspect it always did
00:18:39CIA-70r29468 build result: 1 errors, 0 warnings (thomasjfox committed)
00:19:06[Saint]Hmmm, I have only just noticed's quite convenient.
00:19:06 Nick Robdgreat_ is now known as Robdgreat (
00:19:12 Quit Robdgreat (Changing host)
00:19:12 Join Robdgreat [0] (~rob@unaffiliated/robdgreat)
00:19:43thomasjfoxkugel: Your build host is fscked up :o)
00:19:57stripwaxsideral - (or maybe MikeS or Buschel) - and the bit that's unfixed is vorbis-specific only?
00:19:57thomasjfoxhome/kugel/compilers/bin//m68k-elf-ar: /home/kugel/rockbox-client/builds/build-mpiohd200/apps/codecs/lib/ffmpeg_bitstream.o: No such file or directory
00:20:45sideralstripwax: I haven't been following these changes that closely, sorry
00:20:46*[Saint] wonders how hard it would be to add kinetic scrolling to pictureflow.
00:20:59[Saint]It looks *really* nice on RaaA btw
00:21:02kugelthomasjfox: nah, that's just a superflous / in $PATH (I hope so anyway)
00:21:25*stripwax should try 3.8 and see if it's more stable than svn
00:21:28thomasjfoxkugel: does the build run a parallel build?
00:21:53thomasjfoxkugel: see;type=mpiohd200
00:21:56kugelwith -j9
00:21:56TheSevenkugel: sounds like the one in /home/kugel/rbdev/rockbox-git/firmware/export/system.h should ne #ifndef'ed
00:22:03thomasjfoxlooks like a typical parallel make problem
00:23:16[Saint]kugel: Is there a reason that Hotkey isn't defined for any application builds?
00:25:07kugelI don't think so
00:25:37JdGordonhotkey should be automatically enabled for touchscreen in config.h
00:25:38kugelI quite possibly copy&pasted <target>.h from the wrong one :p
00:26:01[Saint]right...well, it'd be cool if it were enabled.
00:26:17[Saint](I don't mean that to sound demanding)
00:28:56thomasjfoxkugel: Is there a way to retrigger a build? I'll take it that I can ignore the failure on the build host
00:29:34 Join jhMikeS [0] (
00:29:34 Quit jhMikeS (Changing host)
00:29:34 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
00:30:09kugelthomasjfox: just ignore it
00:30:38thomasjfoxkugel: Ok. I'm just surprised it didn't see something like this pop up before
00:31:42 Quit mshathlonxp (Quit: fall into sleep)
00:32:42*thomasjfox definately needs more sleep and the sleep timer isn't working for RaaA yet
00:34:27stripwaxdata abort at 0004e860 with 3.8 . that's after database autoupdate has finally completed to 100% (although playback paused). is that the same symptom as fs#11863? (i find it hard to tell)
00:37:46sideralstripwax: Don't think so
00:37:49stripwaxmaybe it's unrelated.
00:39:05sideralstripwax: If it's reproducible with a release build, please report it to the bug tracker
00:39:26sideral(with the steps / data that reproduce it)
00:39:43 Quit ender` (Quit: Consultation, n. Medical term meaning "to share the wealth.")
00:40:01stripwaxsideral - yep, will see if it's reproducible. data aborts due to database update/initialize seem hard to get good reusable test cases for
00:40:34stripwaxi'll try deleting + reinitalising database first
00:41:21sideralWhat version generated the old database?
00:41:54stripwaxsideral - some post 3.7 svn build, but i wouldn't know the exact build unfortunately
00:42:01 Quit linuxguy3 (Read error: Operation timed out)
00:42:15sideralThere has been a DB schema change since 3.7. Possibly the logic that checks for database incompatibility is broken.
00:45:23sideralthomasjfox: Is the last commit considered OK despite the build error?
00:45:24 Join linuxguy3 [0] (~timj@
00:45:31thomasjfoxsideral: yes
00:45:45sideralOK, thanks
00:47:16CIA-70New commit by sideral (r29469): DB import: Correctly import previously exported resume offsets
00:48:40thomasjfoxkugel: After testing on my host, I think the SDL audio backends needs some work. It also clicks on every seek / touchscreen activity
00:49:00kugelon a pc?
00:49:16stripwaxsideral - hrm, database abort at 0004ee74, on a Initialise Now with 3.8.
00:49:22thomasjfoxkugel: yes
00:49:33thomasjfoxkugel: x86_64 Fedora 14 box
00:49:51stripwaxalthough i just saw your above commit. reckon i should try svn latest?
00:50:26sideralstripwax: Could also be a metadata parsing problem. Can you enable metadata logging in the debug menu prior to initializing the database?
00:50:32stripwaxi do have an exported database, which presumably gets imported as the final step of database initialise? (it appears to dara abort only right at the very end)
00:50:45stripwaxwill do. with 3.8; or with svn latest?
00:51:06CIA-70r29469 build result: All green
00:51:14sideralstripwax: I believe the import of the exported data is done only when requested explicitly / manually
00:51:42sideralMy commit is not related to DB init, only to DB import
00:51:58kugelTorne: do you know what's the deal with the ipod video 64mb ata hack?
00:51:58sideralstripwax: with 3.8, please
00:52:22TheSevenkugel: which one of the hacks?
00:52:29Tornenot really. it's, er, to make the physical sector thing work.
00:52:43kugelespecially why it can't use a standard mutex
00:52:44Tornei'm not sure what the problem is with using a regular lock though
00:52:48sideralstripwax: Also / alternatively, you could try DB import of your old export instead of DB init
00:52:51Torneobviously it needs a mutex
00:52:58Tornebut i'm not sure why it uses that weird hacky implementation
00:53:04Torneor how that one even works
00:53:39kugelTheSeven: standard mutex also has a corelock
00:53:58stripwaxsideral - i thought the DB import would just merge in the modifications (play counts, ratings, etc) into the database?
00:54:13stripwaxanyway, reiniting db now with metadata cache enabled.
00:54:44stripwaxfyi dircache is on, load to ram is true, and there is an existing database (presumably loaded to ram at this point)
00:55:19kugelI can't seem to spot why it needs a separate mutex implementation
00:55:19sideralstripwax: I'm not sure. I've only ever used DB export/import as a backup mechanism, not for merging DB data
00:55:46Tornekugel: well someone tried taking it out and it doesn't work :)
00:56:02kugelit seems to work really the same
00:56:15TheSevenkugel: at the first glance it looks to me as if it is actually some kind of more lightweight mutex that isn't IRQ-safe
00:56:21Torneoh? maybe i'm not remembering right
00:56:31Tornei thought someone found it broke stuff to not use it
00:56:41kugelif the thread has the lock, then count up and return, otherwise get the corelock and lock the current thread until the mutex is released by the owner
00:56:53kugelbetween switch_thread()s the corelock is released
00:57:18kugelyes, it doesn't disable irqs, is that the reason?
00:57:37TheSeventhis might have something to do with thread blocking behavior as well
00:57:43CIA-70New commit by sideral (r29470): autoresume: Optimize playlist resume (manual, bookmark, or after power-on) ...
00:57:56TheSeventhis one more or less polls for the lock, while mutexes probably do some more sophisticated blocking
00:58:28kugelit doesn't block (as in ask the scheduler to block it) the thread, but it busy loops for the lock to get free
00:58:36kugelso it's much more like the spin lock, isn't it?
00:58:38 Quit thomasjfox (Remote host closed the connection)
00:59:13TheSevenyeah, it looks like a latency-optimized lightweight mutex to me
00:59:13kugelif so we should instead call it spin lock and move it to the other kernel primitives
00:59:49kugelbut why does the bigger ipod video have the lightweight one, while the others use normal mutexes?
01:01:39CIA-70r29470 build result: All green
01:01:42 Quit Jerom (Quit: Leaving.)
01:02:56sideralstripwax: Please be sure to post your results. You'll find your metadata-access log in /metadata.log. I need to go offline now.
01:03:10stripwaxsideral - will do. thanks
01:03:36stripwaxi'm guessing metadata.log is sync'd while building db -
01:03:48sideralYes it is. Is it slow? :)
01:03:58stripwaxso it contains the last metadata at the point of the data abort (when it gets there)?
01:04:20stripwaxshould i post just the last part of the metadata.log then, or all of it?
01:05:13sideralJust the last file name, along with the file if possible
01:05:16TheSevenhm, looking at the places where it is used, I really don't understand what the matter is with a regular one
01:05:17kugelmutexes aren't interrupt safe anyway, I wonder why they deal with them
01:05:40TheSevenkugel: not interrupt safe in terms of you may not use them in IRQ handlers
01:05:40sideralstripwax: Thanks for helping debug this!
01:05:54 Join DJeXeCute [0] (
01:05:56stripwaxthank you!
01:06:07DJeXeCuteI have a question
01:06:35TheSevenand as we have a cooperative multitasking kernel, I don't see why they would need to care about IRQs on singlecore systems
01:07:08 Quit sideral (Quit: Leaving.)
01:07:13TheSevenDJeXeCute: well, seems like my crystal ball is broken, it failed to guess what your question is :)
01:07:17DJeXeCuteBolth my Onboard and my USB soundcard are fried. I was woundering if I could somehow use my Ipod Mini 1st gen as a USB soundcard?
01:07:22kugelTheSeven: exactly
01:07:28kugeldealing with corelocks should be sufficient
01:07:31 Quit pamaury_ (Remote host closed the connection)
01:07:54kugelthe normal mutex disables IRQ before switch_thread(), but I can't see how
01:07:58TheSevenDJeXeCute: in theory that might be possible. someone worked on that some time ago, but i don't think it ever got far enough to be really usable
01:08:02kugelI mean why it's needed
01:08:12kugeljhMikeS: ?
01:08:25TheSevenkugel: maybe it's acutally switch_thread that needs this?
01:08:55kugelenabling IRQ is the first thing switch_thread() does
01:09:06TheSeventhat seems to be pretty much pointless then :)
01:09:27kugeloops no
01:09:31 Quit liar (Ping timeout: 240 seconds)
01:10:55kugelit disables irq itself rather early, and not all callers of switch_thread() disable irq so I still don't see the point
01:11:35kugelthe mutex has the advantage that threads blocked on the mutex are unblocked in order, so they should be better latency wise
01:12:22kugelwith that spin lock it's random which thread is unlocked first, and high proprity threads might get penalties because of over yielding
01:12:53jhMikeSkugel: ??
01:13:57kugeljhMikeS: do you know why ata.c has this custom mutex implementation? and why mutex_lock disables irqs?
01:14:35DJeXeCute[0]=&sev[0]=&pri[0]=&due[0]=&reported[0]=&cat[0]=&status[0]=open&percent[0]=&opened=&dev=&closed=&duedatefrom=&duedateto=&changedfrom=&changedto=&openedfrom=&openedto=&closedfrom=&closedto= Is that the project Id be looking for to use my ipod as a usb soundcard?
01:15:04jhMikeS1) yes, because there is an odd problem with iPod video that's yet to be tracked down 2) all scheduling disables IRQs because of interaction with interrupt wakeups
01:15:32stripwaxsideral - for the logs - how annoying, it now seems to be aborting while committing database
01:15:44 Quit mudd1 (Read error: Operation timed out)
01:16:10kugeljhMikeS: I can understand that for event_queues but not for a mutex
01:16:34stripwaxbut it sortof seemed ok right up until the moment I enabled Auto Update = Yes and rebooted. I got a data abort right after the reboot, then on a subsequent reboot I got a data abort at committing database step 4/9 (reproducibly).
01:17:02jhMikeSkugel: yes, for a mutex when sleeping because of interaction with the run lists and other parts
01:17:49jhMikeSalso, intaction of corelocks and interrupts on > 1 core
01:18:35jhMikeSone can't look at one function and infer because it's not called in ISR context or use by them that the disable is unimportant since things connect more broadly
01:18:41kugelcorelock should take care of multi issues, no?
01:19:17jhMikeSno, you don't want a corelock that is locked by one core to deadlock on itself if an ISR on that same core happens to try to claim it
01:19:40kugelIMO block_thread() and friends should disable irqs, not their callers in kernel.c
01:20:30 Quit DerPapst (Quit: Leaving.)
01:20:38jhMikeSthen we'd be doing alot more disables/enables
01:20:41kugelif a mutex is locked from an isr things go wild anyway
01:21:48jhMikeStrue, but things are woken from an ISR, things that locking a mutex also touches. the connections are farther down
01:23:33kugeljhMikeS: do you happen to know why this custom mutex is used on a 64mb ipod? I'd expect it's used on all, or a standard mutex is used on all
01:24:00kugelI can't see a difference from normal mutexes except that it doesn't disable irq and busy waits for the lock to get free
01:26:01 Quit robin0800 (Quit: Leaving)
01:26:02kugelyour commit message is a bit unclear to me
01:27:55DJeXeCuteCan someone please help me I found the project for using the ipod as a USB Soundcard but need assistance in getting it set up?
01:28:07*TheSeven wonders if mpegplayer can take advantage of LCD DMA
01:28:16jhMikeSkugel: because 80MB ipod video had a problem with the UI becoming unresponsive that the object hack seemed to alleviate. I suspect its to do with sleeping the core at a particular time but I'm not sure _why_ it's effective.
01:29:22stripwax80GB rather than 80MB? (i.e. large physical sector size)
01:29:24jhMikeSkugel: yeah, it's not different from normal other than it doesn't do true blocking and doesn't invoke prio inheritance
01:29:48kugelit also doesn't disable irqs
01:30:38jhMikeSthe yield will of course but it doesn't alter the contents of the run list so it doesn't do it there
01:30:38TheSevenso it basically prevents the core it's running on from sleeping
01:30:42kugeldoes the ipod video use dma or so for lcd updates?
01:30:53kugelperhaps the standard mutex makes irqs disabled for too
01:31:05TheSevenmaybe this is to alleviate a similar problem like the one I just had with ce-ata where latencies while waiting for something were way too long
01:31:30jhMikeSthey shouldn't be disabled for more than a microsecond, no
01:31:42TheSevenwhat about the following
01:31:48TheSeven-storage can be accessed from both cores
01:32:11DJeXeCuteThanks for the help everyone bye:P
01:32:13 Quit DJeXeCute ()
01:32:34TheSeven-if both cores access it, it's likely that control of the lock alternates between them very often
01:32:44jhMikeSTheSeven: on PP I wouldn't do that, but that because of lack of cache coherency
01:33:00TheSeven-if the waiting core is sleeping, it might be possible that it will only wake up on the next tick
01:33:26jhMikeSit won't sleep if waiting for a spinlock
01:33:50jhMikeShow could it? there's a thread running
01:33:51TheSevenit won't sleep if it's waiting for a corelock or that ata lock, but it will sleep if it's waiting for a mutex, right?
01:34:10TheSeven(if the mutex is locked on the other core)
01:34:10jhMikeSinterrupt will be enabled immediately upon waking it though
01:34:26TheSevenhow are cores woken up on PP?
01:34:30jhMikeSthe other core explicitly wakes the other when it unblocks one of its threads
01:35:06kugeljhMikeS: if this one doesn't deal with sleeping and priority inheritance I can imagine it works better in some situations. shouldn't we move it into kernel.c as spin lock?
01:35:10jhMikeSthread-pp.c, by setting the processor control registers to wake up the other core
01:35:10TheSevenso it can actively wake it, and doesn't need to wait for some tick or something similar?
01:36:01jhMikeSno, nothing waits for ticks. if a thread is made runnable, the core is always immediatly woken and any interrupt races between ISR serving and going back to sleep are also avoided
01:36:28TheSevenheck, how can mpegplayer get more fps than test_fps? :)
01:37:17jhMikeStwo cores sending and acking messages to each other through a queue on PP takes ~12 microseconds
01:37:20 Join krazykit [0] (
01:37:53TheSevenah, boosting :/
01:38:03TheSevenlinuxstb_: still around?
01:38:19TheSevenjhMikeS: 01:36] <TheSeven> heck, how can mpegplayer get more fps than test_fps? :)
01:38:30jhMikeSdoes it? :)
01:38:39TheSevenyes, if mpegplayer boosts and test_fps doesn't
01:38:51jhMikeSah, ok :)
01:38:56kugelI also find it suspicious that it works better for the 64mb ipod only. I strongly suspect it works better on all (assuming it works indeed better)
01:39:06*TheSeven wonders if LCD DMA pays off if the frame rate in test_fps drops from ~42 to ~27fps when activating it
01:39:27kugelanyway, IIRC all ipod video builds are 64MB with MAX_PHYS_SECTOR_SIZE enabled, IIUC
01:39:41jhMikeSkugel: I think only the 64MB also had the 80MB drive. I know the 30MB is ok with the regular mutex
01:39:58kugelthe builds are unified now
01:40:00jhMikeSiirc it wasn't always so
01:40:26gevaertsjhMikeS: except IIRC some refurbished players have random boards basically
01:40:36jhMikeSgigabeat S uses ATA DMA and not the hacky object
01:41:28kugelwell, it's possibly not hacky. spinlocks are a generally accepted barrier
01:41:50jhMikeSgevaerts: interesting. this all started years ago and the only thing showing symptoms needing the hack was 80MB video. it's got nothing to do with large sector support either. one can enable that on anything and not have a problem with the mutex.
01:42:32gevaertsjhMikeS: I mean, if it's related to the disk, I'm pretty sure people were having problems before the build unification
01:42:55jhMikeSthey were indeed
01:45:01jhMikeSkugel: If the 80MB issue gets fixed, I'll delete that stuff from ata.c in a second and clean things back up
01:45:25TheSeventest_fps is going nuts!
01:45:26*kugel swaps jhMikeS G and M key ;)
01:45:36TheSeven121864.122888fps RGB full screen
01:45:54TheSevenCPU frequency: -443338684 MHz
01:46:09jhMikeSlol...what's that about?
01:46:22kugeljhMikeS: do you think it's fixed now? effecively that "hack" is active on all ipod video builds now isn't it?
01:47:36jhMikeSkugel: Noone's tested without it to my knowledge lately. I've tried several times and the fs tasks quickly follow for the 80GB users
01:47:42*TheSeven suggests just trying it
01:48:11*kugel has little hope
01:48:33TheSevendrop a few testing builds somewhere and have some volunteers check if something starts acting up
01:48:59*TheSeven wonders who originally wrote that hack, and why
01:49:09TheSevenor rather how it was determined that that thing was needed originally
01:49:12jhMikeSkugel: Jiggabytes it is :)
01:49:42iamben80MB ipod... you got me excited that it might work on my rio pmp300 (64MB) player
01:49:43*jhMikeS wrote the object hack in ata.c (for the reasons we've been talking about)
01:50:24TheSevenyeah, but how did you determine that this is what was needed to fix the problems?
01:50:49 Quit [Saint] (Disconnected by services)
01:50:51 Join S_a_i_n_t [0] (S_a_i_n_t@
01:50:58TheSeveni mean, if I was hunting for a bug, I wouldn't just try and slap a bunch of mutex re-implementations at it
01:51:24kugelI'd call you crazy if you did :p
01:51:45TheSevenyeah, but what would trigger someone to even suspect that the mutexes might be involved?
01:51:56jhMikeSbecause of an earlier implementation of something that didn't sleep (before prio inherit) that was used in ata.c. when going back to pure mutex, the problem showed up. I then wrote the hack to immitate what was going on before and the problem mysteriously disappeared again.
01:52:22TheSevenah, so you basically locally reverted a kernel change for the ata driver...
01:52:59kugelso it's the priority inheritance that exposes this behavior?
01:53:09jhMikeSI had a hunch, spit that out, and said, "please try this".
01:53:54 Quit stripwax (Quit:
01:54:06kugeljhMikeS: what do you say to my earlier question?
01:54:20jhMikeSI think it's the possibility of the core sleeping while running ata code on that particular device. I believe I even tried priority disabled entirely without effect.
01:54:51jhMikeSwas that one one?
01:55:32 Quit S_a_i_n_t (Ping timeout: 264 seconds)
01:55:53kugel"if this one doesn't deal with sleeping and priority inheritance I can imagine it works better in some situations. shouldn't we move it into kernel.c as spin lock
01:56:00kugel"if this one doesn't deal with sleeping and priority inheritance I can imagine it works better in some situations. shouldn't we move it into kernel.c as spin lock
01:56:03jhMikeSI tried alot of crap, threw it at the owners and gave up until somehow I got an 80GB video in my possession to track it down.
01:56:06DBUGEnqueued KICK kugel
01:56:06kugel"if this one doesn't deal with sleeping and priority inheritance I can imagine it works better in some situations. shouldn't we move it into kernel.c as spin lock
01:56:20kugel"if this one doesn't deal with sleeping and priority inheritance I can imagine it works better in some situations. shouldn't we move it into kernel.c as spin lock?"
01:56:39jhMikeSoh, we did have that in the first place
01:57:25jhMikeSAlot of fast CPU waste cycles if they just spin, esp. with a slow thing like a disk. They could save power instead.
01:57:55kugelit shouldn't replace mutexes, but in certain cases spin locks are the better choice
01:58:01TheSeven...which brings up last week's argument again...
01:59:26jhMikeSI'm not sure they'll save anything since the scheduler isn't really much overhead and we'll have just another damned object in the kernel (the fewer the better imo)
01:59:56 Join [Saint] [0] (S_a_i_n_t@
02:00:13CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
02:00:13*TheSeven asks for a tick-less high-resolution scheduler once again :)
02:00:31TheSeventhis could save us quite a bunch of yielding loops
02:02:00jhMikeSwhy are the an issue? most of the time threads are asleep anyway unless the codec thread is busy
02:02:09jhMikeS*is there an
02:02:13kugeljhMikeS: obviously it works better in at least one case, and there are quite possibly more
02:02:55kugeland this is in fact a kernel object, whether or not it's in kernel.c. so unless it's removed entirely it should be in kernel.c (perhaps with a HAVE_SPINLOCK_OBJECT)
02:03:38jhMikeSoy, we're going all the way back to 2006/7 ?? :)
02:04:09kugelthe cases where you'd consider a spin lock are the cases where it's not likely a second thread enters the critical section anyway (because the critical section is supposed to be short)
02:05:15kugelno we aren't
02:05:30jhMikeStrue, but then a mutex doesn't do much either if not contended (quite on purpose)
02:06:36 Nick Unhelpful_ is now known as Unhelpful (~quassel@rockbox/developer/Unhelpful)
02:06:55kugelthe current mutex is still the better choice in most cases
02:08:38 Join DJeXeCute [0] (
02:08:51kugeljhMikeS: but if it's contended it does a lot more
02:08:52jhMikeSkugel: a disk access sure isn't a short cs though :)
02:09:25kugelright. that confuses me also. why is that spinlock working better in this situation?
02:10:03jhMikeSkugel: obscure hardware issue I think since it's unique to one setup
02:10:15DJeXeCuteCan someone point me in the right direction for making changes like adding the Ipod USB Soundcard project to my ipod mini 1st gen?
02:10:26***Saving seen data "./dancer.seen"
02:10:37DJeXeCuteI already found the project and gitpulled the branch
02:11:08jhMikeSkugel: of course storage priority might have changed the game a bit
02:12:22kugelah right, so disabling could be worth a try
02:12:45*jhMikeS is up for it
02:14:37kugelbut I maintain that if it's not removed entirely it should be in kernel.c
02:15:00DJeXeCute:/ i give up
02:15:02 Quit DJeXeCute ()
02:19:14TheSevenkugel: if you do that, you'll need to make it singlecore aware
02:19:23TheSeven=> more ifdef hell
02:19:42jhMikeSkugel: I wanted to discourage its growing tendrils into everything, making dumping it harder if the reason for its existence goes away.
02:20:12kugelTheSeven: core_lock/unlock is stubbed for single core
02:21:16jhMikeSthat wakeup object is now everywhere when I could have just extended semaphores properly and made them ISR compatible
02:21:50*jhMikeS does have a patch for that too
02:23:41kugeljhMikeS: it's not everywhere. only few targets use it
02:24:32 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
02:25:12kugelisn't it just the potential thread_switch() that makes semaphore_release isr incompatible?
02:25:14 Quit evilnick_B (Quit: Page closed)
02:27:04jhMikeSkugel: that and it doesn't disable them when claiming the object. for some situations semaphore_wait_w_tmo is also needed (already did it though). they are a bit bigger per object and it's a little bit more code, but not much.
02:28:30kugelwakeup_wait has more code than semaphore_wait
02:29:02jhMikeSbut it does both timeout and infinite blocks
02:30:17kugelanyway, go for isr safe semaphores! :)
02:30:29kugelI already wondered why they can't be used
02:30:33jhMikeSthe experimental patch: (not sure it would apply now)
02:31:05CIA-70New commit by gevaerts (r29471): Recalibrate build costs
02:31:18jhMikeSkugel: because things took a strange turn :)
02:31:21*TheSeven wwonders how irq-safe mutexes or semaphores are supposed to work
02:31:50JdGordondont they disable the irq when you get a lock that needs to worry about that?
02:31:59jhMikeSthe IRQ can up the sem, not down it
02:32:32jhMikeSIt can also poll it with a timeout of zero
02:32:44CIA-70New commit by gevaerts (r29472): Add sdl application to the build system
02:33:27 Quit dfkt (Read error: Connection reset by peer)
02:36:58jhMikeSor semaphore_wait_w_timeout(&sem, 0) (same thing)
02:37:24kugelbut the semaphores lose the ability to switch the thread if a higher prio one was waiting on it, that's a bit sad isn't it?
02:37:35jhMikeSso yeah, ISR can poll it down but won't block
02:37:58jhMikeSThe patch should have a solution by not doing it if IRQ is disabled already
02:39:04kugelah nice
02:40:35kugelis that safe?
02:41:00*kugel doesn't know
02:41:03jhMikeSfor ARM it is, sine IRQ is always disabled by the core
02:41:27jhMikeSolder arm anyway, not the newer stacked IRQ stuff
02:41:40kugelthat's not easy on RaaA I'm afraid
02:42:53kugelbut on RaaA you can check whether you're in an "irq" or not (pthread_self() == main_thread)
02:43:54jhMikeSthe detect should be a target implementation I expect
02:45:19kugelor perhaps just an semaphore_release_from_irq()?
02:45:30jhMikeSon coldfire the sr should have the current ISR priority, which doesn't match the fully-enabled value
02:45:56 Quit linuxguy3 (Ping timeout: 264 seconds)
02:45:56 Quit T44 (Ping timeout: 264 seconds)
02:46:13jhMikeSyes, I had considered that as well
02:47:35 Join linuxguy3 [0] (
02:47:51kugelthe (pthread_self() == main_thread) doesnt work with sdl threads
02:48:30jhMikeSisn't there a current thread id function on SDL?
02:48:57jhMikeSor just the current thread pointer?
02:50:00kugelthe rockbox threads just sdl threads as well as the isrs
02:50:03jhMikeSiirc, one problem I had was having a pointer for the main thread since it was never created through SDL in the first place.
02:50:21kugelyou can get it with pthread_self() in any event
02:51:11kugelin fact we do it already (thread-unix.c)
02:51:32CIA-70New commit by theseven (r29473): iPod Classic: Disable boosting, it seems to cause random lockups
02:51:37kugelthat only works on unix of course
02:51:53jhMikeSand fibers on Windows?
02:52:19CIA-70New commit by theseven (r29474): iPod Classic: Use DMA (and double buffering) for LCD updates
02:52:45 Quit GeekShadow (Quit: The cake is a lie !)
02:52:56kugeljhMikeS: yes, fibers are used on windows. but I meant you can call pthread_self() even if you're only dealing with SDL threads
02:52:57 Quit linuxguy3 (Ping timeout: 252 seconds)
02:52:59jhMikeSof course, sim_enter_irq_handler can set a flag (I think I removed it doing that)
02:53:20kugelnot implemented on android ;)
02:54:03kugelthe android port doesn't lock out interrupts but that hasn't been a problem so far
02:54:21jhMikeSis it preemptive?
02:54:56 Join linuxguy3 [0] (
02:54:57kugelwell, partly, the rockbox threads of course aren't, but the tick handler and inpute events are
02:55:07kugelaudio callback also
02:56:04jhMikeSthat could corrupt the linked lists and other things
02:56:37CIA-70r29473 build result: All green
02:59:38kugelonly the tick handler can possibly deal with kernel objects, the others can't
02:59:56TheSevenkugel: sure?
03:00:01TheSevenwhat about wakeups?
03:00:03kugelon android, yes
03:00:37kugelthere's no wakeups involved in the other handlers
03:00:47jhMikeSthe scheduler has mysterious internal dependencies (but that's not unusual for the sort of thing)
03:01:20 Quit MethoS- (Remote host closed the connection)
03:01:29TheSevenkugel: how does the audio callback acquire more data then?
03:01:59kugelget_more_callback? is there a wakeup involved?
03:02:25jhMikeSaudio callback priority is not handled by the scheduler because it's not usually disabled on target except through pcm_play_lock/unlock
03:03:09TheSevenhm, that's a unidirectional ring buffer, which can in theory be done without kernel primitives
03:03:13jhMikeSon arm, it's often FIQ, and on coldfire it's priority 6
03:03:42jhMikeSit's an overcomplicated one :\
03:04:48CIA-70r29474 build result: All green
03:05:11TheSevennevertheless it might involve a wakeup to signal the codec thread that it may continue working
03:06:55TheSevenso if the codec thread starts blocking for that wakeup while it's being signalled, it might possibly get lost and block infinitely long, depending on whether there's a timeout
03:07:21kugelcodecs don't use wakeups
03:07:41kugelnothing in the core does, they're only used in a target specific way
03:08:05TheSevenhow are the buffering and codec threads and the pcm callback communicating then?
03:09:09kugeldon't ask me :)
03:09:21 Part iamben
03:09:27TheSevenspinlock-like yielding loops!?
03:09:59jhMikeScodec thread just polls the buffer and calls sleep()
03:11:15jhMikeSmpegplayer, same thing but much simpler
03:13:53 Quit mystica555 (Ping timeout: 240 seconds)
03:14:44 Quit mystica555_ (Ping timeout: 264 seconds)
03:16:32 Quit linuxguy3 (Read error: Operation timed out)
03:18:25 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
03:19:02 Join mystica555_ [0] (
03:19:34 Join linuxguy3 [0] (
03:26:57 Join mystica555 [0] (
03:27:21*jhMikeS finds 26 files with wakeup_init
03:27:43kugelso much?
03:28:04jhMikeSmips and arm files
03:29:36*jhMikeS wonders if that was meant as irony relative to his "everywhere" comment :)
04:04:05 Quit Keripo (Quit: Leaving.)
04:05:16 Join Keripo [0] (
04:06:14 Nick froggyman_ is now known as froggyman (~seth@
04:06:24 Quit froggyman (Changing host)
04:06:24 Join froggyman [0] (~seth@unaffiliated/froggyman)
04:08:49 Join kugel_ [0] (~kugel@rockbox/developer/kugel)
04:10:30***Saving seen data "./dancer.seen"
04:10:44 Join Topy44 [0] (~Topy44@
04:11:58 Quit kugel (Ping timeout: 240 seconds)
04:18:43JdGordondamn I hate being at work when i have an epiphany :p
04:19:01JdGordonOK, 2 more tiny changes to the touchscreen tags:
04:19:27JdGordon1) Add an "action" which does nothing so we can know if a region is pressed but that region is used to control the wps and not really do anything else
04:20:16JdGordon2) Add a "setting set" action which will let a touch region set a specific setting to a given value instead of just the next/prev value which we have now
04:20:43JdGordonthis would allow awesome stuff like popping up a submenu with the 4 replay modes and being able to press the one you want instead of having to cycle them
04:22:44JdGordon1) is simple to add, 2) is going to be interesting...
04:23:26JdGordonit might even make sense to make a new tag completly for that one so %T doesnt get too confusing (although i don't tihnk it is)
04:24:32 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:24:32 Quit amiconn (Disconnected by services)
04:24:32 Quit pixelma (Disconnected by services)
04:24:34 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:24:36 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:24:52 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:25:05JdGordon%T(label, x, y, width, height, *setting_set*, *repeat_mode*, *off*)
04:25:43JdGordon[Saint]: ^
04:25:47jhMikeSouch in usb-drv-as3525.c !!
04:27:26jhMikeSone should not signal then immediately re-init the object :\
04:42:07 Join The_Pwny [0] (
04:47:07 Quit Topy44 (Read error: Connection reset by peer)
04:47:08 Quit Barahir (Ping timeout: 250 seconds)
04:48:47 Join Barahir [0] (
04:51:52 Quit TheSeven (Ping timeout: 260 seconds)
04:52:42 Join Topy44 [0] (~Topy44@
04:52:43 Quit Keripo (Quit: Leaving.)
04:57:01 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
05:01:06CIA-70New commit by jdgordon (r29475): Add a touchaction "none" which does nothing when pressed. This will allow some fancy controls to skins in combination with recent %T/%Tl changes
05:05:06JdGordon8min builds :(
05:09:29CIA-70r29475 build result: All green
05:15:55jhMikeSwhatever happened to the < 2 min builds?
05:16:20JdGordonFor these 181 builds, the following 31 build clients participated:
05:16:40JdGordonEffective round speed was 4225 points/second, making us 70% efficient.
05:16:44JdGordonwe need more servers
05:16:51jhMikeSmoar servers!
05:17:56 Join Rob2223 [0] (
05:21:13 Quit Rob2222 (Ping timeout: 240 seconds)
05:37:22 Quit GodEater (Ping timeout: 260 seconds)
05:38:06 Join GodEater [0] (
05:38:06 Quit GodEater (Changing host)
05:38:06 Join GodEater [0] (~bibble@rockbox/staff/GodEater)
05:43:08 Join Keripo [0] (
05:59:04 Join L-Strife89 [0] (~Strife89@
06:10:34***Saving seen data "./dancer.seen"
06:10:48 Quit timccc (Quit: Leaving.)
06:11:08 Join timccc [0] (~timccc@
06:24:31 Join Keripo1 [0] (
06:27:14 Quit Keripo (Ping timeout: 241 seconds)
06:29:43 Join Horschti [0] (~Horscht@xbmc/user/horscht)
06:32:50 Quit Horscht (Ping timeout: 240 seconds)
06:47:40 Quit bzed (Ping timeout: 272 seconds)
06:48:35 Join bzed [0] (
06:56:43 Join stoffel [0] (
07:01:28 Join Xerion_ [0] (
07:02:50 Quit Xerion (Ping timeout: 240 seconds)
07:02:51 Nick Xerion_ is now known as Xerion (
07:11:34 Quit L-Strife89 (Ping timeout: 240 seconds)
07:28:55 Quit BHSPitMonkey (Remote host closed the connection)
07:33:41 Quit [Saint] (Ping timeout: 246 seconds)
07:34:31 Quit stoffel (Ping timeout: 250 seconds)
07:41:15 Join Buschel [0] (
07:59:35 Join leavittx [0] (~lev@
08:04:34 Join [Saint] [0] (S_a_i_n_t@
08:05:22CIA-70New commit by Buschel (r29476): Switch off ATA DMA for all PP targets. There is sufficient evidence that ATA DMA causes sporadic lockups and static noise on several PP based players. ...
08:05:32Buschel[Saint]: hi, could you test the nano1g audio stuff?
08:09:15JdGordonBuschel: going to backport that for 3.8.1?
08:09:53BuschelJdGordon: yes. but not before Thursday evening. I am leaving for a short business trip today
08:10:38***Saving seen data "./dancer.seen"
08:10:47 Join mudd1 [0] (
08:11:54 Quit [Saint] (Ping timeout: 272 seconds)
08:12:05Buschelsame for FS #11973 as soon as it is verified
08:13:58 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
08:15:01 Join [Saint] [0] (S_a_i_n_t@
08:15:45 Quit bluebroth3r (Read error: Operation timed out)
08:18:09 Join bluebrother [0] (
08:18:09 Quit bluebrother (Changing host)
08:18:09 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
08:20:04 Quit Buschel (Quit: ChatZilla 0.9.86 [Firefox 3.6.13/20101203075014])
08:21:39 Join fyrestorm [0] (
08:24:02 Join LinusN [0] (~linus@rockbox/developer/LinusN)
08:26:16CIA-70New commit by jdgordon (r29477): Remove code duplication in some generic skin touch action handling. ...
08:28:21 Join ender` [0] (
08:34:19amiconnHmm, seems like the big build boxes have fallen off the buildserver network
08:34:30amiconnMy two clients are working like mad.
08:34:49JdGordonI'm reinstalling ubunutu on my desktop and will reconnect it up
08:35:00JdGordondunno how much it wil help, but hopefully a bit
08:35:47amiconn"6 builds remaining" That's still Buschel's commit
08:39:56CIA-70r29476 build result: All green
08:41:55Zagoryeah we lost all roolku's machines
08:42:27amiconnI hope the hard timeout isn't set too low then, at least
08:43:13amiconnLast round took almost 35 minutes
08:48:36 Quit amee2k (Ping timeout: 250 seconds)
08:51:43 Quit t0rc (Quit: Give someone code, help them with one project. Teach someone to code, help them rule the world.)
08:52:24JdGordontime to drop some builds also?
08:52:35JdGordon35min is the bad old days :)
08:53:08Zagorif roolku is permanently gone, we need to do a recruiting drive and perhaps prune the build list a bit
08:53:19Zagorbut we should first see if that is the case
08:53:51CIA-70New commit by wodz (r29478): Move ata_mmc.c into target tree as it is SH (ondio) specific. Associated header file is left intact as it seems to be used in many places for ...
08:53:58 Join wodz [0] (
08:55:13 Join amee2k [0] (
08:57:31amiconnBuild speed should be a bit better for subsequent builds due to ccache
09:19:22 Quit amee2k (Ping timeout: 246 seconds)
09:20:49 Nick crwll is now known as crwl (
09:26:08wodzdamn this build rund takes ages
09:26:35Zagor1 build remaining...
09:27:07CIA-70r29477 build result: 0 errors, 4 warnings (jdgordon committed)
09:33:41GodEaterJust wanted to pop in to say I thought kugel's suggestions for GSOC were all great ideas. They all look to be around the right amount of work - he's clearly inspired!
09:38:45wodzZagor: any estimated time for the current rund - I have to leave soon unfortunately
09:39:27 Quit mystica555_ (Ping timeout: 260 seconds)
09:39:44wodzI think TheSeven's idea about test layer is also very interesting and can be turned into GSOC proposal as well
09:39:52ZagorI'm afraid not. I added an additional client but I wouldn't bet below 30 mintues
09:40:51wodzMaybe we should prepare vm machine to serve as build client. This will attract windows users maybe?
09:41:21 Quit Keripo1 (Quit: Leaving.)
09:42:36Zagorfirst we need to know if roolku's machines are gone permantently or not
09:43:36Zagorof course, more clients is *always* good. but I don't know if we really have a lot of windows users wishing to run build client vms?
09:44:39Zagorit certainly can't hurt though, if someone fancies preparing it
09:46:06wodzZagor: The question is more if our build system is prepared for this. This should be as much automated as possible meaning that server side should assign unique name for example
09:46:25Zagorbut there's a whole list of other clients that have dissappeared as well. not just roolku.
09:46:59Zagorwodz: ah, no we're not prepared for "cloned clients".
09:47:24wodzZagor: maybe worth considering
09:47:30 Quit factor (Read error: Connection reset by peer)
09:47:51 Join amee2k [0] (
09:50:48 Quit Topy44 (Read error: Connection reset by peer)
09:50:58ZagorI wonder if a simple start/install script isn't a better idea. I think people are likely to want to be able to follow their contribution in the lists.
09:52:01wodzthat's an option yes
09:52:47JdGordonwould vm's even give us a worthwhile speed bump?
09:52:59*JdGordon will fix his warnings shortly
09:53:04 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
09:53:14wodzJdGordon: that depends how many people will run it
09:54:00ZagorJdGordon: right now, every sheeva box or atom netbook would help
09:54:46wodzJdGordon: but currently running build client need some afford which may discourage people from contributing to our distributed build system
09:55:11 Quit amee2k (Ping timeout: 260 seconds)
09:55:16wodzrunning vm is cheap solution this days
09:55:40Zagorat midnight we had 15 non-roolku clients, now we have 7 (and that is only because I just added a new!)
09:57:42pixelmaa pity I don't have time to update or prepare my VMs before today in the evening or so
09:57:49CIA-70r29478 build result: 0 errors, 4 warnings (wodz committed)
09:57:49 Join roolku [0] (
09:58:05wodzZagor: About clonned clients - build server may mark machine with cleverly build hash which will be written on first run in some config file. People will be able to track down their contribution by this
09:58:22 Join n1s [0] (
09:58:22 Quit n1s (Changing host)
09:58:22 Join n1s [0] (~n1s@rockbox/developer/n1s)
09:58:26wodzNot my warnings :-)
09:58:30JdGordondoes anyone know why the sim builds arent warning like they used to?
09:58:38pixelmaah roolku, welcome :)
09:58:46roolkuhi :)
09:59:17roolkuit looks to me that some change on the build server may have killed all my clients?
09:59:45roolkuthey hang and the last message is: 2011-03-01 04:09:21 Killed build sdlapplication
09:59:47Zagorroolku: I don't know what has happened. the server is unchanged.
09:59:58Zagorbut it appears to have affected a lot of clients.
10:00:00pixelmaor connection problems?
10:01:04JdGordonthe required toolchain config stuff got messed up?
10:01:04roolkuwhat is sdlapplication? it is not a simulator build?
10:01:06 Part LinusN
10:01:22Zagorpixelma: no, I can connect to for example bagder's build client but it still absent
10:01:41pixelmait's to do with the Rockbox as an app port - one for SDL
10:01:49CIA-70New commit by jdgordon (r29479): fix yellow
10:01:56 Join Topy44 [0] (~Topy44@
10:02:23roolkuthey didn't drop all at once, but starting from 29473 - as if they got something they couldn't chew
10:02:36Zagoroh, gevaerts added that in 29472. that looks like it could the problem.
10:03:38roolkuI started b38 again and it seems to work - I'll wait a bit before I go through the trouble of restarting them all
10:04:01Zagorroolku: yes, we'll need to verify the sdlapp addition
10:04:43 Join factor [0] (~factor@
10:05:04pixelmaah, the SDL app needs to set aprefix during configure, maybe that causes the trouble
10:05:11Zagoryes, "../tools/configure −−target=sdlapp −−type=n" blocks waiting for LCD width ...
10:05:22*Zagor whips gevaerts
10:05:23pixelmaand that
10:05:44 Join einhirn [0] (
10:06:06JdGordonhe did say something about blaming other people if anything went pear shaped last night :)
10:07:53JdGordonwhats the package to get makeinfo?
10:08:12ZagorJdGordon: texinfo
10:08:55JdGordonand do any targets still use the old arm toolchain?
10:09:31Zagorwe should take that out of the "all" option in rockboxdev
10:10:42***Saving seen data "./dancer.seen"
10:13:51CIA-70New commit by wodz (r29480): Move drivers/i2c.c into target tree as it contains SH specific bits. Leave associated header file intact as it is used in many places for historical ...
10:16:16wodzOk, that's my 0.02$ to make codebase a bit cleaner
10:16:22 Quit mudd1 (Ping timeout: 264 seconds)
10:20:41CIA-70New commit by zagor (r29481): Add LCD width and height to sdlapplication.
10:23:05Zagorroolku: now you can restart your clients
10:23:14ZagorI'll send a mail to the rbclient list
10:25:32CIA-70r29479 build result: All green
10:26:20 Join The_Pwny_ [0] (
10:27:34 Quit The_Pwny (Ping timeout: 276 seconds)
10:29:42 Join amee2k [0] (
10:29:49roolkuZagor: I have restarted the n?? clients ... will do the rest later today
10:31:28 Quit The_Pwny_ (Ping timeout: 276 seconds)
10:33:30wodzdamn this sd/mmc things are scattered across target tree files, firmware/export/sd.h, firmware/export/sdmmc.h, firmware/export/ata_mmc.h
10:33:37JdGordonoh, the install paths got changed?
10:33:43CIA-70r29480 build result: All green
10:34:09wodzwhat a mess
10:35:40 Join DerPapst [0] (
10:36:00 Quit wodz (Quit: Leaving)
10:37:49 Join bicyclerepairman [0] (~Sam@
10:39:29bicyclerepairmanhow do I go about donating some hardware?
10:39:43bicyclerepairmanI have an old creative Zen that's not useful to me,
10:39:52bicyclerepairmanand I know you chaps are working on a port to it, so...
10:39:59bicyclerepairmanit's yours if you want it.
10:42:17Zagorthat's nice. are you in the uk?
10:43:24Zagorwhich model is it?
10:47:48CIA-70New commit by gevaerts (r29482): Fix the builds file properly now. I'll never do this stuff again after my regular bed time!
10:49:06JdGordonis the wiki outdated? is arm in the arch list eabi?
10:49:07Zagoroops :)
10:49:15ZagorJdGordon: eabi
10:51:03JdGordonand m68k? saying i've got the wrong version (which i just installed with
10:53:40*gevaerts doesn't understand why his calibration script didn't choke on that bad line
10:53:56JdGordonm68k-gcc452 apparently shold prob be updated :)
10:54:51gevaertsJdGordon: the instructions say that you should tweak according to your local setup IIRC :)
10:55:07JdGordonyes, but it doesnt say what the excpeted arch's are
10:55:17JdGordonor legal ones
10:55:19*JdGordon is in
10:56:40bicyclerepairmanYeah I'm in the UK.
10:57:32bicyclerepairmanZagor, it's a Creative Zen Micro
10:57:58bicyclerepairmanfirmware version is 2.21.02,
10:58:11 Join LinusN [0] (~linus@rockbox/developer/LinusN)
10:58:12bicyclerepairmanstorage capacity is 2 gig incase you care :-)
10:58:41bicyclerepairmanand it works OK but the headphone jack is a bit squoogley
10:58:54bicyclerepairmanso that you have to hold the wire to get any sound.
10:59:11bicyclerepairmanBut I tihnk this doesn't matter from a developer's point of view
11:03:34 Join GodEater_ [0] (
11:03:34 Quit GodEater_ (Changing host)
11:03:34 Join GodEater_ [0] (~bibble@rockbox/staff/GodEater)
11:06:24 Quit GodEater (Ping timeout: 250 seconds)
11:17:04Zagorany brit volunteering to be mail drop?
11:18:42bicyclerepairmanbrb, got an appointment at the bank.
11:20:08 Join mudd1 [0] (~cmertes@2001:638:504:20e0:221:70ff:fe83:655e)
11:28:02 Quit panni_ (Read error: Connection reset by peer)
11:28:13 Join panni_ [0] (
11:30:36 Join MethoS- [0] (~clemens@
11:36:07 Join francesco_ [0] (
11:36:40francesco_can i set crossfading playback and relative settings from config file or similar?
11:37:04n1sfrancesco_: yes
11:37:51francesco_n1s: can u point me to the relative docs/how to?
11:39:00n1sfrancesco_: see the "Config file options" section in the manual's appendix
11:43:17francesco_n1s: thx
11:47:50 Quit roolku ()
11:49:11 Quit simonrvn (Read error: Connection reset by peer)
11:49:50 Quit [Saint] (Quit: I'm only going to Heaven if it feels like Hell, I'm only going to Heaven if it tastes like caramel...)
11:53:06bicyclerepairmanI'm back
11:53:13bicyclerepairmanZagor, did you find anyone?
11:54:24 Join dfkt [0] (dfkt@unaffiliated/dfkt)
11:55:49pamauryZagor: what do you mean by "mail drop" ?
11:57:24Zagorpamaury: to recieve the device. and forward it to any developer who expresses interest.
11:58:07pamauryok, I guess it's better if it's someone in UK
11:59:48ZagorI just thought it would be easiest. but if nobody volunteers, and if bicyclerepairman doesn't mind sending it internationally, I can take it
12:00:12bicyclerepairmanwell where do you live, Zagor?
12:00:29bicyclerepairmanoh that's simply beautifu :-)
12:00:31pamauryI don't min too, France is not too far also :)
12:00:32bicyclerepairmanI love Sweden
12:01:37bicyclerepairmanwell I don't really mind sending it internationally, but I wonder if you could help me out with the postage?
12:01:43bicyclerepairmanwhoever's interested
12:04:04Zagorpossibly. but let's wait a bit to see if someone local raises a hand.
12:10:45***Saving seen data "./dancer.seen"
12:13:08francesco_i have a feat request (or better a bug fix) for which i'll be ready to pay a (small) amount of money for. A free player would be also provided to the interested developer. Anyone?
12:13:53JdGordonwhy dont you say what it is?
12:14:13francesco_jdGordon: it's task #11664
12:14:16JdGordonand bare in mind we dont code for money, we do it to have fun
12:14:32francesco_jdGordon: sure.
12:17:31Tornefrancesco_: That's a bug that a lot of people already care about and are working on
12:17:36TorneIf we knew how to fix it it would be fixed already
12:17:50TorneOffering money/players is not likely to make any difference
12:17:56Torneit'll get fixed when we work out what the problem is
12:22:34 Join swilde [0] (
12:27:00francesco_Torne: ok, i was just trying to offer some help to speed things up. Maybe i've choosen the wrong mean. Anything else i can do to help? (apart from coding...)
12:27:35Tornethere is almost certainly no way to help without making a technical contribution
12:27:42Tornewhen a new patch is proposed, test it
12:27:58Tornebut other than that.. probably not
12:29:57francesco_Torne: ok thanks
12:30:50n1sverifying bugs can also be helpful if they are only affecting certain hardware for example
12:31:06Torneadd the bug to your watchlist and it'll email you whenever soemone adds a comment
12:31:16Tornen1s: this is "usb doesn't work on amsv2"
12:31:28Tornen1s: so there's not a lot of verification to do :)
12:32:10n1si understood him as asking how to help in a wider sense, not just for that particluar bug
12:34:20TheSevenTorne: do you know what exactly the matter with amsv2 usb is?
12:34:58n1swhy is logoswapper in our wiki?
12:34:59TheSevenIIRC this is the same OTG that I am working with in several ipods, and I've had some funky behavior with these recently as well
12:35:00Tornesometimes the device hard locks and can't be reset
12:35:12Tornesince there's no hardware reset on those devices
12:35:20Torneyou have to wait for the battery to run all the way down
12:35:31TorneThis is why USB is still disabled on them :)
12:35:37TheSevenwhich luckily happens a lot faster than on an ipod :)
12:35:58Tornethat's all i know about it anyway
12:36:15TheSevenok, so this is actually a pure software bug in the driver?
12:36:20Torneno idea.
12:36:46TheSeventhe behavior i recently experienced was linux having trouble to communicate with that OTG, while windows was working fine
12:37:06TheSevenand sometimes some lockups as well, but these apparently went away after raising Vcore a bit
12:39:25 Join [Saint] [0] (S_a_i_n_t@
12:44:23JdGordonhey [Saint]
12:47:46 Join wodz|work [0] (
12:48:01 Quit rasher (Ping timeout: 272 seconds)
12:48:07 Quit [Saint] (Ping timeout: 250 seconds)
12:49:08wodz|workAbout amsv2 lockups - why don't set watchdog? This at least may make the problem less harmfull and make testing safer
12:50:13wodz|workwe will know at least if tick tasks are still running in locked up state
12:50:21 Quit panni_ (Quit: ( :: NoNameScript 3.81 :: ))
12:50:33 Join Jerom [0] (~jerome@
12:51:49 Join [Saint] [0] (S_a_i_n_t@
12:52:16wodz|workTheSeven: I experience various funny effects with USB on linux with receintish kernel. This is surly problem on linux side.
12:52:39TheSevenyeah, but it doesn't happen with other hardware, so we must be triggering it somehow
12:52:45TheSevenand thus we must be able to work around it
12:53:38wodz|workTheSeven: not true - I have problem with PP5020E for example and some cardreaders\
12:54:12francesco_ok, i throw my 2 cents, based on my experiences: OF boots up even if player powered after it gets connected to power trough the usb cable+power adapter
12:54:14TheSevenyeah, but it doesn't happen to every single pendrive out there
12:54:41TheSevenso there must be something we can do about it
12:54:47 Join rasher [0] (
12:54:47 Quit rasher (Changing host)
12:54:47 Join rasher [0] (~rasher@rockbox/developer/rasher)
12:55:00pamauryI think jhMikeS suggested that #11664 was a SD bug and not a USB one, after all the usb fixes committed to trunk
12:55:50wodz|workTheSeven: Maybe we trigger it somehow but the bug *IS* in linux ehci driver
12:56:08TheSevenpamaury: that would be easy to check
12:56:12TheSevenjust remove SD?
12:56:34TheSevenwodz|work: nevertheless we need to get rid of it somehow
12:57:37wodz|workTheSeven: I'd love to see this fixed unloading (or unbinding) ehci when I want to access my ipod is a bit silly
12:58:00TheSevenif possible without a reboot at all
12:58:06pamauryI haven't tested my usb patch for quite some time, perhaps I should see what is the current state
12:58:32wodz|workTheSeven: But on my mini1g even disk mode exhibits the same problem
12:59:09TheSeventhat problem being that it works, but is terribly slow because of bus resets after every couple of bulk packets?
12:59:38TheSeven(often observed as mount locking up because scanning the FAT may take like an hour)
12:59:53 Join simonrvn [0] (simon@2001:470:8c85:11fe::c0a8:195)
13:00:26wodz|workTheSeven: yes but I think it sometimes goes as bad as hard usb disconnect
13:02:54wodz|workTheSeven: Do you know something about the root of this behaviour? Google isn't very verbose about it
13:03:30TheSevenIIRC it has been bisected, so we might want to have a look at the diff
13:04:16wodz|workTheSeven: could you provide the source of this info?
13:05:23TheSevennot exactly our behavior, but similar enough to be caused by the same issue
13:06:22 Part LinusN
13:08:28TheSevenespecially sounds interesting
13:10:04pamaurythat looks really similar to our problem
13:13:54 Quit wodz|work (Ping timeout: 240 seconds)
13:13:54 Part Zagor
13:14:37 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
13:16:07TheSevenyeah, but sadly seems to be off-topic in that launchpad-thread
13:16:21*TheSeven just spotted a possible build system flaw!
13:16:27 Join wodz [0] (
13:16:35TheSevenit doesn't seem to notice if svn up fails because of e.g. insufficient permissions
13:16:48TheSevendoes that mean that it will produce old builds without anyone noticing?
13:17:05wodzstill the only solution provided is to unload ehci
13:18:17n1sunloading ehci is kind of annoying with usb connected mouse and keyboard
13:19:06pamauryyou need to get the command line right if you don't want to reboot :)
13:19:50TheSevenin theory you can just unbind that one device and pass it over to uhci
13:19:51Zagordoesn't svn return non-zero on bad permissions?
13:21:02wodzeven worse on ubuntu where they decided to compile this into kernel
13:22:11 Quit sasquatch (Quit: WeeChat 0.3.2)
13:23:09 Join sasquatch [0] (
13:23:09 Quit scorche (Disconnected by services)
13:23:09 Join scorche` [0] (~scorche@rockbox/administrator/scorche)
13:26:19 Quit Zagor (Read error: Connection reset by peer)
13:26:26 Quit wodz (Quit: CGI:IRC (Ping timeout))
13:28:38 Join Guest85077 [0] (
13:28:45 Quit Guest85077 (Changing host)
13:28:45 Join Guest85077 [0] (~bjst@rockbox/developer/Zagor)
13:31:10CIA-70New commit by jdgordon (r29483): Add an ability to set a setting to a specific value with a touchscreen action. ...
13:33:17 Join TheLemonMan [0] (
13:34:35JdGordoncan we convert genlang to c? it really slows down the builds :/
13:34:35*JdGordon is an impatient bugger
13:35:37 Part Guest85077
13:36:59 Quit Jerom (Quit: Leaving.)
13:36:59n1syeah genlang is slow and i have been thinking about doing that but i don't want to :)
13:36:59 Quit timccc (Ping timeout: 240 seconds)
13:38:01n1siirc someone said it could be made faster while still remaining perl but i don't know how
13:38:01JdGordonZagor: gevaerts: is sane?
13:38:06JdGordonI keep getting these server connection stalled errors :(
13:39:08 Join timccc [0] (~timccc@
13:40:58DEBUGReceived signal 15 (SIGTERM), terminating (snapshot: fplrun.c line 385)
13:40:58***Saving seen data "./dancer.seen"
13:45:46***Started Dancer V4.16
13:45:46***Connected to on port 6667
13:45:46***Logfile for #rockbox started
13:45:47Mode"logbot :+i" by logbot
13:45:51***Server message 501: 'logbot :Unknown MODE flag'
13:45:51 Join logbot [0] (
13:45:51 Join timccc [0] (~timccc@
13:45:51 Join TheLemonMan [0] (
13:45:51 Join scorche` [0] (~scorche@rockbox/administrator/scorche)
13:45:51 Join sasquatch [0] (
13:45:51 Join simonrvn [0] (simon@2001:470:8c85:11fe::c0a8:195)
13:45:51 Join rasher [0] (~rasher@rockbox/developer/rasher)
13:45:51 Join [Saint] [0] (S_a_i_n_t@
13:45:51 Join swilde [0] (
13:45:51 Join dfkt [0] (dfkt@unaffiliated/dfkt)
13:45:51 Join francesco_ [0] (
13:45:51 Join MethoS- [0] (~clemens@
13:45:51 Join mudd1 [0] (~cmertes@2001:638:504:20e0:221:70ff:fe83:655e)
13:45:51 Join GodEater_ [0] (~bibble@rockbox/staff/GodEater)
13:45:51 Join bicyclerepairman [0] (~Sam@
13:45:51 Join DerPapst [0] (
13:45:51 Join amee2k [0] (
13:45:51 Join einhirn [0] (
13:45:51 Join factor [0] (~factor@
13:45:51 Join Topy44 [0] (~Topy44@
13:45:51 Join n1s [0] (~n1s@rockbox/developer/n1s)
13:45:51 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
13:45:51 Join ender` [0] (
13:45:51 Join fyrestorm [0] (
13:45:51 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
13:45:51 Join leavittx [0] (~lev@
13:45:51 Join Xerion [0] (
13:45:51 Join bzed [0] (
13:45:51 Join Horschti [0] (~Horscht@xbmc/user/horscht)
13:45:51 Join Rob2223 [0] (
13:45:51 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
13:45:51 Join Barahir [0] (
13:45:51 Join pixelma [0] (quassel@rockbox/staff/pixelma)
13:45:51 Join amiconn [0] (quassel@rockbox/developer/amiconn)
13:45:51 Join kugel_ [0] (~kugel@rockbox/developer/kugel)
13:45:51 Join mystica555 [0] (
13:45:51 Join linuxguy3 [0] (
13:45:51 Join krazykit [0] (
13:45:51 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
13:45:51 Join YPSY [0] (
13:45:51 Join froggyman [0] (~seth@unaffiliated/froggyman)
13:45:51 Join shai [0] (
13:45:51 Join ehntoo [0] (
13:45:51 Join saratoga [0] (9803c6dd@gateway/web/freenode/ip.
13:45:51 Join MagusG [0] (
13:45:51 Join crwl [0] (
13:45:51 Join n17ikh [0] (
13:45:51 Join linuxstb_ [0] (~linuxstb@rockbox/developer/linuxstb)
13:45:51 Join Robdgreat [0] (~rob@unaffiliated/robdgreat)
13:45:51 Join Unhelpful [0] (~quassel@rockbox/developer/Unhelpful)
13:45:51 Join tchan1 [0] (
13:45:51 Join balintx [0] (
13:45:51 Join Dr_Agasa [0] (
13:45:51 Join antil33t [0] (
13:45:51 Join mc2739 [0] (~mc2739@rockbox/developer/mc2739)
13:45:51 Join parafin [0] (
13:45:51 Join eGen_ [0] (
13:45:51 Join cjcopi [0] (
13:45:51 Join jepler [0] (~jepler@emc/developer/pdpc.professional.jepler)
13:45:51 Join Loto [0] (~nfs@xbmc/user/Loto)
13:45:51 Join CapsAdmin [0] (
13:45:51 Join knittl [0] (~knittl@unaffiliated/knittl)
13:45:51 Join Mephistopheles [0] (
13:45:51 Join AlexP [0] (~alex@rockbox/staff/AlexP)
13:45:51 Join jordan` [0] (
13:45:51 Join FOAD [0] (~dok@
13:45:51 Join logiclost [0] (
13:45:51 Join Llorean [0] (~DarkkOne@rockbox/user/Llorean)
13:45:51 Join olejorgenb [0] (
13:45:51 Join avacore [0] (
13:45:51 Join alexbobp [0] (
13:45:51 Join guymann [0] (
13:45:51 Join FoolOnHill [0] (
13:45:51 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
13:45:51 Join niekie [0] (~niek@CAcert/Assurer/niekie)
13:45:51 Join literal [0] (
13:45:51 Join kkit|sh [0] (
13:45:51 Join Rondom [0] (
13:45:51 Join Zarggg [0] (
13:45:51 Join maraz [0] (
13:45:51 Join plux [0] (
13:45:51 Join Dreamxtreme [0] (~Dre@
13:45:51 Join markun [0] (~markun@rockbox/developer/markun)
13:45:51 Join Zambezi [0] (Zulu@
13:45:51 Join Galois [0] (
13:45:51 Join Torne [0] (~torne@rockbox/developer/Torne)
13:45:51 Join soap_ [0] (~soap@rockbox/staff/soap)
13:45:51 Join z35 [0] (
13:45:51 Join tmzt [0] (~tmzt@
13:45:51 Join aevin [0] (eivindsy@unaffiliated/aevin)
13:45:51 Join Utchy [0] (
13:45:51 Join ps-auxw [0] (~arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
13:45:51 Join Elfish [0] (amba@2a01:4f8:100:90a1:abc:abc:abc:abc)
13:45:51 Join ranmachan [0] (
13:45:51 Join gevaerts [0] (~fg@rockbox/developer/gevaerts)
13:45:51 Join ack` [0] (
13:45:51 Join yosafbridge [0] (
13:45:51 Join Kohlrabi [0] (
13:45:51 Join ved [0] (
13:45:51 Join Battousai [0] (~bryan@gentoo/developer/battousai)
13:45:51 Join simabeis [0] (
13:45:51 Join Hadaka [0] (
13:45:51 Join Guinness [0] (
13:45:51 Join Farthen [0] (
13:45:51 Join zu_ [0] (
13:45:51 Join dionoea [0] (~dionoea@videolan/developer/dionoea)
13:45:51 Join feisar- [0] (
13:45:51 Join Strife89 [0] (~Strife89@
13:45:51 Join merbanan [0] (
13:45:51 Join ThomasAH [0] (
13:45:51 Join CIA-70 [0] (~CIA@
13:45:51 Join preglow [0] (thomj@rockbox/developer/preglow)
13:45:51 Join TBCOOL [0] (
13:45:51 Join Slasheri_ [0] (
13:45:51 Join advcomp2019_ [0] (~advcomp20@unaffiliated/advcomp2019)
13:45:51 Join @ChanServ [0] (ChanServ@services.)
13:45:51 Join jae_ [0] (
13:45:51 Join yawny [0] (
13:45:51 Join 45PABT3DY [0] (~iq@unaffiliated/iq)
13:45:51 Join Kuitsi [0] (
13:45:51 Join pikytcus1 [0] (
13:45:51 Join pjm0616 [0] (
13:45:51 Join scorche|sh [0] (
13:45:51 Join B4gder [0] (~daniel@rockbox/developer/bagder)
13:45:51 Join jfc [0] (
13:45:51 Join [fred] [0] (fred@
13:46:47 Join _Zagor [0] (
13:46:47 Quit _Zagor (Changing host)
13:46:47 Join _Zagor [0] (~bjst@rockbox/developer/Zagor)
13:47:28 Nick _Zagor is now known as Zagor (~bjst@rockbox/developer/Zagor)
13:48:37JdGordonalthough the ordering is what might be interesting... it drops the server and continues to build, then the server comes back, tries to delete the build dir then asks for that build
13:50:38 Join hebz0rl [0] (
13:53:26 Join kronflux [0] (~kronflux@
13:55:19 Join wodz [0] (
13:55:49wodzn1s: About genlang - profiling will be the very first step:
13:58:42CIA-70r29483 build result: 15 errors, 0 warnings (jdgordon committed)
14:02:03CIA-70New commit by jdgordon (r29484): fix red
14:05:14 Quit wodz (Quit: CGI:IRC)
14:06:37 Join Sochiro [0] (~Sochiro@
14:08:37CIA-70r29484 build result: All green
14:12:54 Quit antil33t (Read error: Connection reset by peer)
14:13:04 Join antil33t [0] (
14:13:30 Quit shai (Ping timeout: 250 seconds)
14:23:10 Quit francesco_ (Remote host closed the connection)
14:25:47 Quit linuxguy3 (Read error: Operation timed out)
14:29:50 Join linuxguy3 [0] (
14:32:21 Join ender| [0] (
14:34:18 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
14:36:13 Quit krazykit (Quit: awe yeeeeeee)
14:55:12 Quit Sochiro (Read error: Connection reset by peer)
14:55:40 Join Sochiro [0] (~Sochiro@
15:02:58 Join komputes [0] (~komputes@ubuntu/member/komputes)
15:05:34 Join evilnick_B [0] (0c140464@rockbox/staff/evilnick)
15:08:35 Quit Sochiro (Ping timeout: 240 seconds)
15:15:06 Join LinusN [0] (
15:15:06 Quit LinusN (Changing host)
15:15:06 Join LinusN [0] (~linus@rockbox/developer/LinusN)
15:25:45 Join _jhMikeS_ [0] (
15:25:45 Quit _jhMikeS_ (Changing host)
15:25:45 Join _jhMikeS_ [0] (~jethead71@rockbox/developer/jhMikeS)
15:25:45 Quit jhMikeS (Disconnected by services)
15:25:46 Nick _jhMikeS_ is now known as jhMikeS (~jethead71@rockbox/developer/jhMikeS)
15:35:04 Quit linuxguy3 (Read error: Operation timed out)
15:36:16 Join linuxguy3 [0] (
15:45:48***Saving seen data "./dancer.seen"
15:47:02 Quit mudd1 (Quit: Ex-Chat)
15:48:08 Quit linuxguy3 (Read error: Operation timed out)
15:50:41 Join linuxguy3 [0] (
15:51:28 Join Sochiro [0] (~Sochiro@
16:01:17 Quit n1s (Quit: Lämnar)
16:02:04 Join panni_ [0] (
16:06:39 Quit FoolOnHill (Remote host closed the connection)
16:06:59 Quit leavittx (Read error: Operation timed out)
16:18:01 Quit sasquatch (Quit: WeeChat 0.3.2)
16:18:26 Join sasquatch [0] (
16:36:52 Part LinusN
16:40:18 Join FoH [0] (
16:41:09 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
16:51:36 Join Sochiro_ [0] (~Sochiro@
16:51:45 Quit Xerion (Read error: Connection reset by peer)
16:52:18 Quit pamaury (Remote host closed the connection)
16:52:34 Join Xerion [0] (
16:55:26 Quit Sochiro (Ping timeout: 240 seconds)
16:59:15 Part swilde ("ERC Version 5.3 (IRC client for Emacs)")
17:00:57 Quit t0rc (Quit: Give someone code, help them with one project. Teach someone to code, help them rule the world.)
17:02:56 Part Zagor
17:03:21 Join user890104 [0] (
17:04:00 Quit Sochiro_ (Read error: Connection reset by peer)
17:04:25 Join Sochiro_ [0] (~Sochiro@
17:07:09 Quit TheLemonMan (Quit: free(me))
17:11:10 Join toffe82 [0] (
17:13:37 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
17:15:07 Join pyro_maniac_ [0] (
17:16:45 Quit Sochiro_ (Ping timeout: 240 seconds)
17:19:06saratogaTheSeven: the problem with AMSv2 USB seems to be that the controller isn't initialized or clocked right, and so occasionally drops the USB connection (IIRC)
17:19:54 Quit antil33t (Read error: Connection reset by peer)
17:20:06 Join antil33t [0] (
17:20:46TheSevensaratoga: but why would that cause a software lockup?
17:21:19saratogai don't think it locks up
17:21:30saratogai think the transfer just times out and the host resets the connection
17:23:11 Nick tchan1 is now known as tchan (
17:23:22 Quit tchan (Changing host)
17:23:22 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
17:26:53TheSevensaratoga: some hours earler someone claimed that they hard lock in a way where they can't even be rebooted, unless you let the battery run down
17:27:21saratogathat happened when we tried to reboot into the OF from rockbox to use its USB mode
17:29:39TheSevenhuh? that chat sounded like we are rebooting into the OF to prevent exactly that behavior when trying to use RB usb
17:29:45 Join sideral [0] (~sideral@rockbox/developer/sideral)
17:30:33TorneOh, er, yes. i am mistaken :)
17:30:45Tornewe are currently requiring that people manually boot the OF to get USB
17:31:05Tornebecause rebooting locks it up sometimes and using the rockbox usb stack is unreliable
17:31:10saratogaautomatically rebooting tending to hard lock players
17:31:21sideralAs a data point, I've been running my ClipV2 with enabled USB for more than a month now, and it's perfectly stable. Caveat: I needed to revert r29169
17:31:33saratogaAFAIK USB transfers don't freeze the player, but they don't work all that well on a lot of systems, particularly windows IIRC
17:32:24TheSevendo we have more precise information about how those transfers fail?
17:32:51TheSevenis it CRC errors on the connection or similar things? or some higher-level problem?
17:33:19TheSevenalso, which transfer directions are affected?
17:33:35TheSevenany difference between the two available drivers?
17:33:36sideralTheSeven: Don't think so. I remember people reporting the occasional "usual nonresponsive Windows system", and retrying the connection fixed it
17:35:59saratogayou should ping pamaury about it
17:41:44sideralsaratoga: r29169 is jhMikeS' commit, and pamaury doesn't want to have anything to do with the SD driver :)
17:45:30 Join Xerion_ [0] (
17:45:51***Saving seen data "./dancer.seen"
17:47:37 Quit Xerion (Ping timeout: 276 seconds)
17:47:39 Nick Xerion_ is now known as Xerion (
17:49:59 Quit timccc (Ping timeout: 252 seconds)
17:51:29 Join u42p [0] (
17:54:10 Join n1s [0] (
17:54:10 Quit n1s (Changing host)
17:54:10 Join n1s [0] (~n1s@rockbox/developer/n1s)
17:54:15u42paloha. i half bricked my sansa clip+ or something. it was acting goofy the past days, not starting or hanging/crashing. this morning i had the great idea of using RockboxUtility (linux) to install the latest rockbox on it. the window kinda froze, the sansa said writing every now and then but it seemed to be very slow. well, i had to go so i killed it... i could boot the sansa firmware afterwards, gui froze every now and then but i could listen well
17:54:16u42pto music.
17:54:47u42pnow the player was off over the day and wont display anything anymore. i held power or power+home for ages with no luck. if i plug it in i get "Disk /dev/sdf: 4 MB, 4231680 bytes"
17:54:53u42p"Disk /dev/sdf doesn't contain a valid partition table"
17:55:08u42pany idea how to recover from this?
17:57:18gevaertsu42p: may help
17:57:29gevaertsBe careful with those things though
17:57:48u42pi really really really do not want to open it up :\
17:59:43TorneYou don't have a choice.
18:00:08TorneInterrupting the flashing process is pretty much guaranteed to be fatal
18:00:19Torneto unbrick it you need to short two terminals inside.
18:01:08 Quit Xerion (Quit: )
18:01:11u42pi was not flashing the bootloader, so that might not be damaged
18:01:25TorneYou must have been flashing the bootloader :)
18:01:31 Join fkhodkov [0] (
18:01:44Tornethe state it's in now is the state it goes into when it gets a bad flash
18:01:54TornePretty sure nothing else causes that
18:01:57u42poh, i should have said. i used rockbox for ages on it.
18:01:58*TheSeven wonders what that 4MB volume is good for if it doesn't allow to recover
18:01:59u42poh :(
18:02:16TorneTheSeven: It's probably the original DFU supported by the silicon
18:02:32TheSevensounds like we should find a way to use it for our purposes
18:02:38TorneTheSeven: lots of people like to make devices where the actual boot flash is not the kind of device the SoC is intended to be used with
18:03:06TorneI suspect whatever mode that is, it's probably an on-chip rom that may well not have any clue how to write to the actual flash
18:03:18Tornethere's probably another memory somewhere which has teh actual recovery mode in it
18:03:26Tornewhich is selected by shorting those pins
18:04:15 Join krazykit [0] (
18:04:23Tornealternatively it might be a two-stage thing where it expects to have a small recovery image written to it
18:04:32Tornewhich it then boots and presents a better recovery system.
18:05:28Tornethe onchip rom might only be smart enough to let you write an image to RAM
18:05:45TheSevenTorne: aren't those SoCs documented?
18:05:51Torneer, maybe.
18:06:00TorneIt's also possible that the situation is the other way around
18:06:03TheSevenand 4MB sounds like a ramdisk, yes
18:06:08Torneand the *working* recovery mode is the one in the bootrom
18:06:18Torneand the useless one is some stub on the actual firmware
18:06:41Tornewhich doesn't normally get overwritten/corrupted during flashing and thus ends up running by default if the main image is broken
18:06:42TheSevenhow do we know that it's useless?
18:06:48Torneuseless to us at present, then
18:07:18Tornewhat actually are the terminals that get shorted?
18:07:44Torneit's presumably connected to some kind of boot select pseudoregister, or to the flash (and thus shorting it kills the flash's ability to respond to reads)
18:08:06TheSevenhas anyone dumped the content of that 4MB volume?
18:08:16Tornei hope so ;)
18:08:18TheSevenis it zeroed? random? or something static?
18:08:32Tornemaybe i'll take my clipv2 apart sometime and see what i can work out
18:08:36Tornei never use it for anything
18:09:01TheSevenhow stable is that unbricking method?
18:09:12Tornei don't think anyone who's actually tried it has failed
18:09:17TheSevenis there a way to brick it even more, so that this shorting trick doesn't work any more?
18:09:40TheSevenI remember people jtagging these devices because that was the only means of access left
18:10:32Tornehm, maybe.
18:10:41gevaertsTheSeven: it took a while for people to discover those pins
18:11:14Torneso these devices have just one flash, and some portion of it is used as bootflash and is hidden from normal use?
18:11:41TheSevenI remember older sansas (pre-ams?) having some i2c flash as well
18:12:01TheSevenbut that's probably a completely different story
18:12:28Torneso is anyone likely to know what those pins are actually connected to?
18:12:39TheSevenshouldn't be hard to figure that out
18:12:43TheSevenanyway, if someone throws a bricked sansa and a rom image of it at me, I'll see what I can do :)
18:12:55TheSeventhere must be something useful we can do with that 4MB volume mode
18:13:13Tornewell, funman or someone may know more
18:13:17Tornei'm curious now, tbh :)
18:13:28TheSevenit strongly smells like the UMSboot thing i've written for the ipods :)
18:13:48TheSevenjust that mine works on top of FAT16 to make it more convenient
18:16:53 Quit aevin (Ping timeout: 240 seconds)
18:16:53u42pi guess the player needs to be glued together after taking it apart?
18:18:22 Quit [Saint] (Ping timeout: 264 seconds)
18:18:41Tornethe as3525 manual mentions how the internal boot thing works vaguely
18:19:01Torneit doesn't actually explain how the usb boot prommer mode works, only what bits to set to enter it.
18:19:12 Join [Saint] [0] (S_a_i_n_t@
18:19:13Torneso i think my first guess is right, one way around or the other
18:19:28Tornethere's a stub in the NAND which doesn't get touched by a normal flash
18:19:39 Part rasher
18:19:44 Join rasher [0] (~rasher@rockbox/developer/rasher)
18:21:40 Join Xerion [0] (
18:23:39TheSevenTorne: do you know how we are writing firmware?
18:23:55Torneyou mean when we do a normal upgrade?
18:24:03TheSevendo we access the flash directly? or do we update it through the OF somehow?
18:24:05Tornethe existing firmware does it for us
18:24:12Tornedetects presence of file on usb unplug
18:24:33TheSevenhow can we brick them that way then?
18:24:50TheSevenwouldn't the OF do some CRC check or something with the firmware file dropped onto it before flashing it?
18:24:57TorneYou might think
18:24:59TorneI have no idea.
18:25:15 Join leavittx [0] (~lev@
18:25:27TheSevena half-written firmware file certainly is no rare problem
18:25:39TheSevenno way such a thing can brick a device
18:25:52Tornei'm not sure i'd assume that :)
18:26:28TheSevenbtw, is this a raw nand chip or some kind of emmc?
18:26:39TheSeventhe package looks like just nand
18:27:04TheSevenis there some hardware bridge in between, or is there no wear leveling at all?
18:27:52Torneit's an eMMC
18:28:02Torneor at least has the interface of one
18:28:56Tornenot all eMMC chips are equal :)
18:29:05Tornesome have magic hidden boot areas and the like
18:29:16TheSevenwhy does an emmc have that many pins?
18:29:41Tornemaybe there's an interface inbetween somewhere
18:29:47Torneor maybe the SoC itself has that logic in.
18:30:05Tornethe datasheet is terrible
18:31:20Torneask someone who actually knows about the port :)
18:32:34Torneso yeah. the part is a regular ONFI NAND part
18:32:37 Join LambdaCalculus37 [0] (~3f74f70d@rockbox/staff/LambdaCalculus37)
18:32:44Tornebut we definately drive it using an SD interface
18:34:25 Join Stummi [0] (~Stummi@rockbox/developer/Stummi)
18:35:06TorneI would guess that Sandisk's ASIC has one of their own flash controllers onboard
18:35:43TheSevendo they have some asic on board? i thought it was just some standard soc from ams?
18:36:00TheSevenor do you mean a different chip?
18:36:01TorneThe SoC is the ASCI
18:36:08Torneit's not an actual AMS part
18:36:17Torneit'll be based on the AMS soft IP
18:36:25Torneso, customised any way sandisk feel like
18:36:30 Join Sochiro_ [0] (~Sochiro@
18:36:37TheSevenhm, so a similar mess like those Samsung S5L87xx things
18:36:48 Quit MagusG (Read error: Connection reset by peer)
18:36:49Tornelarge vendors like that never actually use the *parts* from SoC vendors
18:36:51 Join MagusG [0] (
18:37:00Tornethey just make their own ASIC from the silicon IP
18:37:19Torneso yeah, that's a *third* possible source for the alternate recovery/flash mode
18:37:42Tornethe chip might actually have *two* bootroms in it, one from the AMS design and one added by Sandisk
18:38:35Tornein which case yeah. the AMS manual makes no reference to being able to boot directly from SD/MMC.
18:39:23Torneso it wouldn't be surprising if it also can't write to SD/MMC in its USB prommer mode
18:39:27TheSevendo you happen to have a link to that manual around?
18:39:48TheSevenTorne: I've rarely ever seen DFU modes really write to flash
18:39:58TorneYah, but it's not DFU
18:40:02TheSevenmost of them operate on ramdisks that are then booted and do the actual flashing operation
18:40:06TorneDFU is a standard protocol
18:40:27TheSevenyeah, but it's for a similar purpose in a similar situation, so it will probably use a similar approach
18:40:43TheSevenyou usually don't trust the flash access layer at that stage
18:40:45Tornethe tiny amount of text in the AMS manual suggests that it is literally a flashprogrammer
18:40:53Tornei.e. it's for manufacturing, not recovery
18:41:02TheSevenif you decide to use a different flash chip, you might have to make new asic masks
18:41:16TheSevenmanufacturers usually don't like that :)
18:41:18Torne is the thing i'm looking at
18:41:39TorneWell, there *are* versions of the clip/etc that are different ASICs, it seems
18:41:49 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
18:41:53Tornenote the wiki: teh sd device identifies itself differently on different devices
18:42:05Tornewhich suggests that the silicon that's presenting the NAND as if it were SD is not the same
18:42:34Tornealso i think you have entirely too rational assumptions about the motivations of ASIC designers :0
18:44:37 Quit sideral (Quit: Leaving.)
18:45:37TheSevensounds like we should have a look at a dump of that on-chip rom
18:45:46TheSeventhat will most certainly provide some more detailed information
18:47:23TorneI don't know about the AMS but normally they are undumpable
18:47:38*TheSeven doubts that
18:47:46TorneMost SoCs use their bootrom as the anchor for their secure boot chain of trust
18:47:51TheSeveneven those apple SoC's roms are dumpable!
18:48:14Tornereally? that's odd.
18:48:16TheSevenaccessing the ROM doesn't break the chain of trust if it's implemented correctly
18:48:30TorneTheSeven: ..and? :)
18:49:01TheSevenis there even a way to hide a ROM away?
18:49:12Torneanyway, most SoCs i've worked with not only is it impossible to dump the rom, but even the *interface8 to the rom is super confidential and secret and isn't in the regular docs, requiring special contracts to read
18:49:26TorneOf course there is
18:49:42TheSevenwell, maybe by setting some kill-bit in the SoC that removes access to it once it's finished
18:49:46Tornehack the silicon that implements the bus
18:50:10TheSevenapple actually had that kind of thing for the hardware AES keys, but they didn't bother using it :)
18:50:27TorneThe GBA only lets you read from ROM when the program counter is inside the ROM :)
18:50:36Tornethat one was entertaining
18:50:59Torneon modern ARMs the bootrom is normally only accessible in secure mode
18:51:09Tornewhich is a much easier and more reliable way of doing it
18:51:09TheSevenyou can't do that kind of hack without modifying the ARM core, and sandisk doesn't have a license for that
18:51:20TorneTheSeven: You can if you have an ancient ARM with no cache
18:51:33Torneyou can have the rom itself detect the isntruction fetches to know where the PC is :)
18:52:09TheSevenare people designing that kind of thing masochists?
18:52:28TorneI suspect so, but Nintendo did it anyway
18:52:42TorneAnd then forgot to bounds check the MIDI key number to frequency conversion table
18:52:47TheSevenwell, DAP OS developers are lazy, we know that
18:52:59Torneso you can dump teh whole rom in a mangled roundabout fashion by asking for unusually high/negative MIDI key numbers
18:53:04TheSevenso the OF will probably running in the secure world's supervisor mode all the time (at least apple did that)
18:53:14TorneAS3525 is ARM9, nos ecure mode anyway
18:53:16TheSevenI would be very surprised if we can't get at that rom somehow
18:53:35Torneonly ARM1176 and later has secure mode.
18:53:54TorneTheSeven: There is another option, actually ;)
18:54:04TorneThe rom might not be *in the physical memory map* when it's not selected as the boot source
18:54:16Tornethe internal/external boot switch might actually just disconnect the rom from the bus entirely.
18:54:29Torneand if itnernal boot doesn't go anywhere useful that you can run code in, there's then no way to dump it
18:54:36Tornewithout anyone making any deliberate attempt to make it hard.
18:55:24Tornesome of these boot roms work by overriding what would normally be at physical address 0
18:55:26TheSeveni'd assume that it always boots from the internal rom
18:55:41Torneyou're *probably* wrong.
18:55:48TorneIt has a switch for it
18:56:05Tornebooting from external would require something attached that was sufficiently memorylike to be used by a GPMC
18:56:17Tornebut whatever mystical device it is that's doing the nand->sd remap might include one
18:56:28Tornesince that's a pretty standard thing to do for diskonchip type stuff.
18:56:49Tornethe internal rom doesn't know how to boot from SD either, so unless sandisk modified it it wouldn't be any use
18:56:53*TheSeven suspects that the internal rom is just reading the OF through the SD interface
18:57:07TorneWell, the docs don't lsit that as a boot option
18:57:19Torneso, sandisk would've had to have modified it probably
18:57:20TheSeveni strongly assume that they modified it in one way or another, if they did their own SoC anyway
18:58:07Tornei seriously think it'sm ore likely that the internal boot rom is just ignored entirely and whatever block implements teh SD interface to flash also offers itself as a bootable memory device
18:58:19Tornewhcih probably just reads sector 0 from the flash into a boot memory
18:58:37Tornethere's loads of boot flash devices that are actualyl NAND and just load the first sector of the NAND into a temporary memory to fake being real mapped memory
18:58:50Torneso that a "dumb" processor can boot off them directly withotu needing a smart bootrom
18:59:16 Join ajb_oe [0] (
18:59:44Tornepeople who make ASICs arrange this stuff based on "what copypastable blocks do we already have"
18:59:48Tornenot what's actually a clean design
18:59:49 Part ajb_oe
19:00:20Torneanyway. maybe someone knows already.
19:00:38Torneso it's probably easier to check if they do rather than speculating :)
19:00:52TorneKnowing what the recovery thing is connected to would be a good hint
19:01:06Torneif it's connected to one of the boot select pins documented in the as3525 manual then that's a big hint :0
19:01:17saratogaTheSeven: i believe shorting those pins actually powers down teh NAND chip, and the disk you see is the rom on the SOC going into some kind of failsafe mode
19:01:27saratogawhen it can't init the NAND chip
19:01:35saratogathats why you short it, then quickly unshort it as it boots
19:01:38Tornesaratoga: yes, but that could be more than one choice of what "rom" is doing it
19:03:13saratogaanyway, given how simple the SOC is, i suspect you can just read out the ROM over jtag if you really want
19:03:51Torneyes, someone with jtag set up could just look and see what it does.
19:05:33 Join pamaury [0] (
19:05:34 Quit pamaury (Changing host)
19:05:34 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
19:08:02saratogaIn the case that a USB connection is present and either an update button is pressed or there is no bootable device, the USB promer is started (see Figure 7 Boot decision between normal boot and USB boot promerā€¯ for details). The USB boot promer allows update of the firmware by using an USB mass storage class device. This update can be used either for initial programming (factory programming) or as mechanism for an in-field firmwa
19:08:08saratogasounds like thats the mode we end up in by shorting those pins
19:08:10*TheSeven has a jtag adapter, but no sansa
19:08:28Tornesaratoga: Except the bootrom described there doesn't know how to boot from SD/MMC
19:08:38Torneso it's very unlikely that it knows how to write to SD/MMC either :)
19:09:04Torneso either there's an external rom which is the real boot device, or the internal rom runs something written by sandisk instead and thus the manual doesn't apply
19:09:28Torneor something much weirder is going on
19:10:02saratogathey customized the chip, so they probably updated the bootloader to work with the SD controller
19:10:06saratogathey'd have to right?
19:10:14Torneyou just turn it off
19:10:22Torneand have the SD interface also be a boot flash
19:10:39Tornebecause that's probably a piece of silicon IP you already have lying around
19:10:43saratogahow would the bootloader get code off the SD interface if it didn't know how to talk to it?
19:10:53TorneYou don't use the bootloader
19:11:02TorneYou just have the chip boot directly from whatever is at physical address 0
19:11:11Tornethe onchip bootrom is optional
19:11:18TheSevensaratoga: using some steppingstone-like construct, that loads the first SD block into some internal RAM and emulates that as an external ROM to boot from
19:11:44TheSeventhe code in that first block will then setup the SD controller and possibly SDRAM and load more code
19:11:45 Quit linuxguy3 (Ping timeout: 264 seconds)
19:11:57saratogawell either way that construct is going to have to end up on the SOC's internal memory unless their SD controller is intelligent enough to memory map things into the CPU without any asisistance
19:12:10Tornesaratoga: Yes, it probably is
19:12:17TorneThere are loads of NAND flash controllers that are that smart
19:12:20TorneIt's a common block :)
19:12:47Torneand it's often easier to just copypaste soemthing liek that into your design rather than go anywhere near touching the existing boot logic in a SoC design you licensed
19:13:01TorneIt's not really an SD controller, remember
19:13:07TorneThe SoC alerady has an SD controller
19:13:13 Join linuxguy3 [0] (
19:13:20saratogaso you think the code we end up in for recovery mode is just the start of the NAND?
19:13:27Tornethe extra block of logic that must be there somewhere is implementing an SD card interface using a NAND flash
19:13:40Torneit could trivially also implement a bootflash interface.
19:14:01TorneI think *one* of the two modes you can boot into when it's bricked is just the start of the NAND.
19:14:10TorneMaybe it'st he one you can recover from, maybe it's the one that doesn't seem to be usable
19:14:49TorneSo yes, if shorting the pins just cripples the NAND then maybe *not* shorting the pins gives you whatever's at the start of the NAND and that's a bit of logic that is either nonfunctional or we just don't know how to talk to it right
19:15:01Tornewhereas if you cripple the nand you get something that's actually in a ROM somewhere
19:15:29Torneanyway. my point was there's a very large number of ways this could be implemented and even really illogical seeming ones are quite plausible/likely
19:15:34*TheSeven wonders if anyone has tried just slapping arbitrary data at that 4MB volume
19:15:44Torneso if we want to know someone should actually experiment
19:15:53TorneTheSeven: i would hope so
19:16:19TorneJTAG is probably the best way to try this if possible, since you can just trap reset and walk throught he entire boot process
19:17:18Tornebut a handy shortcut to narrow down the possibilities would be if someone knows what the pins-to-be-shorted are actually connected to
19:17:21TheSevenwell, I've managed to do so on locked-down devices that don't even have working jtag, so with jtag it should be piece of cake
19:21:16 Quit DerPapst (Quit: Leaving.)
19:25:04 Quit tchan (Quit: WeeChat 0.3.4)
19:25:14 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
19:26:12 Quit LambdaCalculus37 (Quit: back to work!)
19:29:40saratogaTheSeven: you should probably try and get an e200v2 then since its the easiest to jtag
19:30:01 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
19:30:07TheSeveneasiest in terms of connecting?
19:30:09 Join Keripo [0] (
19:35:16saratogathe solder points are fairly large
19:35:21saratogai think the m200v4 is similar
19:35:28saratogathe clip has a tiny board with tiny, tiny points
19:38:54TheSevenwell, I've managed to solder inside an ipod nano 2g, that can't be much worse
19:43:21saratogaare you in the EU or US? I could probably just send you my old e200v2
19:43:57 Join stoffel [0] (
19:45:52***Saving seen data "./dancer.seen"
19:47:09 Quit linuxguy3 (Ping timeout: 264 seconds)
19:49:01 Join linuxguy3 [0] (
19:55:22 Join DerPapst [0] (
19:56:38[Saint]TheSeven: You're quite soldered a pad in the Nano2G...jtag in this case can't be much worse, if at all.
19:57:12[Saint]just, more of them.
19:59:14TheSevensaratoga: germany
19:59:37TheSeven[Saint]: swapping the NAND certainly *is* worse though
19:59:55[Saint]Ok, yess..indeed. ;)
20:00:25[Saint]And I'll quite happily tell anyone that mentions that very thing to avoid it at all costs if they're considering it :P
20:05:00 Join Zagor [0] (~bjst@rockbox/developer/Zagor)
20:05:27 Quit Sochiro_ (Ping timeout: 240 seconds)
20:05:41 Quit stoffel (Remote host closed the connection)
20:14:18 Join kevku [0] (~kevku@2001:7d0:0:f9af:feed:feed:feed:feed)
20:17:16 Join thomasjfox [0] (~thomasjfo@rockbox/developer/thomasjfox)
20:21:40 Quit Zarggg (Quit: Zarggg)
20:28:30 Nick kugel_ is now known as kugel (~kugel@rockbox/developer/kugel)
20:29:34 Join sideral [0] (~sideral@rockbox/developer/sideral)
20:34:06 Quit saratoga (Ping timeout: 272 seconds)
20:38:41 Quit benedikt93 (Ping timeout: 246 seconds)
20:41:04 Join TheLemonMan [0] (
20:41:41 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
20:42:58 Quit Keripo (Quit: Leaving.)
20:44:20gevaertskugel: do I guess right when I suspect that make apk needs make and make zip first mainly because it needs the zip and make zip doesn't force a build?
20:45:23kugelthe reason is that make zip forces to be unpacked on start up, very annoying during development
20:46:00kugelmake apk only makes the binary and re-uses the existing zip
20:47:36gevaertsI know what it does :)
20:47:59gevaertsI just want to understand what would need to be done to get the build system to build it
20:48:27gevaerts(step one: make sure it compiler :)
20:48:32kugelmake zip doesn't force a build because it has no dependencies. if it had could work without make zip
20:49:46kugelI suppose we could have a new make zipapk which always makes zip (or repurpose make apk, but still retain the current behavior with another target)
20:49:54kugelis the apk the problem, or the 3 steps?
20:50:11gevaertsthe three steps, mainly. I think
20:50:33gevaerts"Normal" builds run make and make zip, and then upload the zip
20:50:49gevaertsWe could add a special case to the script of course...
20:53:08 Quit pamaury (Ping timeout: 272 seconds)
20:53:42 Join L-Strife89 [0] (~Strife89@
20:54:20gevaertskugel: by the way, are you aware that it currently doesn't build?
20:54:22u42pi am this close to ordering a new clip+, dammit, why did i rush things in the morning...
20:54:31kugelgevaerts: no
20:54:44gevaertskugel: it seems to be missing sys_poweroff()
20:55:14Zagorgevaerts: my modified rockbuild version has a more generic build command line
20:55:38gevaertsZagor: ah, so we could do make && make zip && make apk?
20:55:57gevaertsok, then we probably should migrate
20:55:58Tornewhy doesn't make zip depend on having done a build, anyway?
20:56:08Tornethat seems bad
20:56:25gevaertsTorne: because that makes building with broken source impossible :)
20:56:56*Torne is personally a fan of makefiles which guarantee exactly correct output all the time :)
20:56:58gevaertsBasically because it's annoying that make zip doesn't work while you're trying to get plugins to build on a new port
21:02:09 Join Zarggg [0] (
21:04:22gevaertsthomasjfox: you broke the android build with r29467
21:04:37 Quit fyrestorm (Quit: Ur skills' fireproof like a wooden panel -- U got feds talking leet on your IRC channel!)
21:05:15gevaertsandroid doesn't do sys_poweroff()
21:05:53thomasjfoxgevaerts: Ups. Is there an alternative or should we just add a stub for now?
21:06:06gevaertsno idea. Discuss with kugel I'd say :)
21:06:19thomasjfoxkugel to the rescue!
21:07:32kugeldoes the sleep timer work now?
21:08:16thomasjfoxkugel: Still needs the power thread to work properly. Just did a quick hack in my local tree yesterday
21:08:42thomasjfoxkugel: Though I'm still thinking what code we would put in the dedicated power thread
21:10:26kugelstubbing sys_poweroff is fine
21:11:24*gevaerts doesn't like things that work by coincidence...
21:12:12kugelthomasjfox: you can also move the sys_poweroff() in powermgmt.c down so it's seen by the app ports
21:12:18gevaertsRight now arm and android seem to be able to coexist because our toolchain is prefixed with arm-elf-eabi while android has arm-eabi
21:12:31gevaertsThat sounds *just* a bit too fragile to me
21:12:40thomasjfoxkugel: That's what I wanted to try ;)
21:13:10*thomasjfox just installed meego as my SD card arrived today. New target? ;)
21:13:19kugelgevaerts: android has arm-linux-androideabi-
21:13:33gevaertsI looked in the wrong directory then...
21:13:39TheSevensounds safer :)
21:13:39kugelarm-eabi is from the old ndk, that's obsolete
21:14:17gevaertsThat's fine then
21:14:45kugelandroid-linux-androideabi- is almost upstream gcc (with a few patches, including the android target) backported from 4.5/6
21:14:47 Join Stephen___ [0] (~S@
21:15:33gevaertsI wonder if we could just have the build network *build* the android bits (i.e. the standard make), without uploading. That would only require one more architecture string in the server, and of course build clients exporting ANDROID_NDK_PATH
21:15:51gevaertsAt least we'd get notice on breakages then
21:16:18 Quit L-Strife89 (Quit: Standby)
21:16:37 Quit benedikt93 (Quit: Bye ;))
21:19:50TheSevengevaerts: what about doing the signing stuff on the build server?
21:20:09gevaertsTheSeven: that would be out of scope :)
21:21:57gevaertsI'm just thinking that one of the reasons that we don't have android autobuilt is that we seem to always want to cover the entire chain from source to market in one go. I'm basically proposing to break it down into simple steps
21:24:23u42pTheSeven: unless my clip+ magically starts working again in the next few days, you will receive it ;-}
21:24:44TheSevenu42p: where are you located?
21:24:55TheSeventhat sounds like a plan then :)
21:25:02u42pyeah, just ordered a new one
21:27:14 Quit [fred] (Remote host closed the connection)
21:30:51 Join [fred] [0] (
21:31:33*thomasjfox is looking forward to android in the build system
21:34:23 Join pamaury [0] (
21:34:23 Quit pamaury (Changing host)
21:34:23 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
21:34:41CIA-70New commit by thomasjfox (r29485): Expose sys_poweroff() and cancel_shutdown() to RaaA. Hopefully fixes android build
21:36:10CIA-70r29485 build result: 15 errors, 3 warnings (thomasjfox committed)
21:37:58 Join Sochiro_ [0] (~Sochiro@
21:45:53***Saving seen data "./dancer.seen"
21:50:03 Quit Stummi (Quit: Bye!)
21:51:03 Quit u42p (Quit: Leaving)
21:52:49 Join dfkt|x [0] (
21:52:49 Quit dfkt|x (Changing host)
21:52:49 Join dfkt|x [0] (~dfkt@unaffiliated/dfkt)
21:57:07 Quit Zarggg (Quit: Zarggg)
22:01:39 Quit Stephen___ (Read error: Connection reset by peer)
22:02:09 Quit dfkt|x (Remote host closed the connection)
22:05:08 Quit tchan (Quit: WeeChat 0.3.4)
22:06:57CIA-70New commit by thomasjfox (r29486): Fix red
22:09:01 Join bertrik [0] (
22:09:01 Quit bertrik (Changing host)
22:09:01 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
22:10:07 Quit [Saint] (Disconnected by services)
22:10:09 Join S_a_i_n_t [0] (S_a_i_n_t@
22:11:33 Quit Sochiro_ (Ping timeout: 264 seconds)
22:12:01CIA-70r29486 build result: 15 errors, 3 warnings (thomasjfox committed)
22:19:38CIA-70New commit by thomasjfox (r29487): Fix red - 2nd try. Use same ifdef style as in firmware/drivers/pcf50606.c
22:19:47 Join mystica555_ [0] (
22:22:02 Join sampattuzzi [0] (
22:24:31CIA-70r29487 build result: All green
22:24:31 Quit factor (Read error: Connection reset by peer)
22:26:11 Join factor [0] (~factor@
22:26:55 Quit Rob2223 (Read error: Connection reset by peer)
22:27:09 Join Rob2222 [0] (
22:27:54 Quit leavittx (Ping timeout: 250 seconds)
22:30:40 Join Keripo [0] (
22:41:44 Join aevin [0] (eivindsy@unaffiliated/aevin)
22:44:42 Join Jerom [0] (~jerome@
22:45:07 Join gamefreak264 [0] (~KING@unaffiliated/gamefreak264)
22:46:19 Quit maraz (Ping timeout: 240 seconds)
22:46:28 Join mudd1 [0] (
22:46:56gamefreak264Is one likely to get better or worse battery life on the 5.5G iPod on rockbox firmware than default?
22:47:15 Join maraz [0] (
22:51:32 Quit evilnick_B (Ping timeout: 272 seconds)
22:52:02 Quit pamaury (Remote host closed the connection)
22:53:07 Quit bicyclerepairman (Ping timeout: 240 seconds)
22:53:36 Join bicyclerepairman [0] (~Sam@
22:56:13Lloreangamefreak264: I think we expect better battery life now, assuming similar use patterns.
22:56:35LloreanSome people switch to Rockbox but also take advantage of lossless files, or caption backlights, or other things that end up using a lot more battery life.
22:56:58gamefreak264Llorean: Well, I was mostly intending to playback oggs and MP3 files
22:57:42gamefreak264up until a month or so ago I was using my Sansa Clip on default firmware for that, but now that it's lost my library is mostly unplayable due to it being ogg
22:58:36gamefreak264I thought that my old iPod on rockbox would be a good substitute
22:59:09n1sit should play mp3 and ogg vorbis efficiently
22:59:13gamefreak264Llorean: Also, thanks for the swift response
22:59:40n1sbattery time with aac would probably be slightly worse
23:00:11gamefreak264By the way, is there any way to play back audible files in Rockbox or would I have to dual boot or something?
23:00:26gamefreak264I'm assuming it's propritary so it's a no-go
23:00:28LloreanDual boot.
23:00:36LloreanIt's not just proprietary, it's encrypted and we can't do DRM.
23:02:19gamefreak264" Very stable. Is lacking support for the video decoder chip. " Does this mean video playback would suck up extra battery life or be hiccup-y or what?
23:02:25 Join stripwax [0] (
23:02:55LloreanProbably a bit of both
23:03:05LloreanAnd we can't do AVC video, only MPEG1/2 like other Rockbox players
23:04:18thomasjfoxgevaerts: Where would we add Meego/Pandora packaging specific files in rockbox? I don't want to pollute the top level source directory
23:04:52 Quit Keripo (Quit: Leaving.)
23:05:14gamefreak264thomasjfox: What's the benefit of running rockbox on something like a Pandora over say mplayer or cmus or whatever?
23:05:24 Quit bertrik (Quit: :tiuQ)
23:06:15gevaertsgamefreak264: that's an interesting question, but let me ask another one: what's the advantage of running mplayer or cmus or whatever over say rockbox? :)
23:06:18thomasjfoxgamefreak264: All the features that make rockbox great on other platforms, too: Support for 20+ codecs, gapless, crossfade, browse by folders or database, highly themable and so on
23:06:37thomasjfoxgevaerts: how about "packaging"?
23:06:54 Quit stripwax (Client Quit)
23:07:16gamefreak264thomasjfox: I don't actually have any experience with rockbox first hand, I've kind of been following the project from a distance for a few years, but never actually ried it
23:07:28thomasjfoxgevaerts: I also thought we could move the "debian" directory in there and have a "make maemo" copy it from packaging/maemo to debian/ during the build process
23:07:59gevaertsthomasjfox: making directories in the root during build is *ugly*...
23:08:06gevaertsthomasjfox: I really don't know enough about meego and pandora packaging systems to have a serious idea
23:08:26thomasjfoxgevaerts: All of them need some platform specific support file
23:08:38thomasjfoxgevaerts: Some kind of "how do I packages this" file
23:08:59gevaertsthomasjfox: well yes, but is it one file, or several, and does the packaging system care where they are?
23:09:06thomasjfoxgamefreak264: There are packages for android or maemo (Nokia N900) if that is any help
23:09:32thomasjfoxgevaerts: for the Pandora it's maybe four files
23:09:48thomasjfoxgamefreak264: So you could try it without flashing anything
23:09:52gamefreak264thomasjfox: I was actually considering putting it on my 5th gen ipod, don't know if you saw that
23:09:56thomasjfoxgevaerts: Meego will be about the same
23:10:06gamefreak264Is it just in the android market or do I need to look elsewhere?
23:10:15thomasjfoxgamefreak264: I never tried rockbox on real hardware :o)
23:11:18gamefreak264thomasjfox: I thought you were a developer! :O
23:11:40gamefreak264I actually was inspired to come here after I saw your part message on #openpandora
23:11:45gamefreak264and your little cloak
23:12:04gevaertsgamefreak264: developers develop. If you want people who actually *use* it, you need users :)
23:12:29gamefreak264Personally, I tend to only want to develop stuff that I use myself :P
23:12:39thomasjfoxgamefreak264: Well, I have a .pnd for it ;)
23:12:49gevaertsSeriously though, thomasjfox only works on Rockbox as an Application, not on regular DAPs
23:12:58gamefreak264thomasjfox: I'm still waiting for my Pandora :(
23:12:58gevaertsSo far :)
23:13:14gamefreak264I ordered back in January of 2009 I think
23:14:07thomasjfoxgamefreak264: Waiting for it is part of the joy ;) Just put rockbox on your ipod then
23:14:27 Quit kronflux (Quit: Leaving)
23:14:28gamefreak264thomasjfox: it just makes me depressed, personally
23:14:49gamefreak264thomasjfox: Do you have your unit yet?
23:15:17thomasjfoxgamefreak264: Yes. But please stay on topic (=rockbox) as the channel logs are read by the developers
23:15:36gamefreak264Alright, will do. Didn't mean to get offtopic.
23:16:47thomasjfoxkugel: Are somewhat current android images available somewhere?
23:17:21kugelyou mean rockbox builds?
23:17:29 Quit TheLemonMan (Quit: free(me))
23:18:05gamefreak264My android phone is in pretty much dire need of space, I think I'll just put it straight ontop the ipod
23:19:03thomasjfoxkugel: Yes
23:19:24thomasjfoxkugel: Or are the files in the wiki updated from time to time?
23:19:36kugelthere's some more or less outdated ones on the forum
23:20:00gamefreak264Would I benefit much feature wise from running r29487 or would I likely be better off running a stable build?
23:20:42gevaertsgamefreak264: since the latest stable release is only two days old, I'd say pick 3.8
23:20:49LloreanI'm not sure our latest release *is* that stable, given the number of bugs reported since it came out. Not that a newer build is likely to be too much better at the moment, since it just was released.
23:21:05gevaertsWe haven't had time yet to implement lots of new stuff
23:21:56gamefreak264Llorean: Any nasty bugs I'm likely to encounter?
23:22:20gamefreak264Anyway, I'm going with 3.8
23:22:30 Join stripwax [0] (
23:22:41 Quit S_a_i_n_t (Ping timeout: 252 seconds)
23:22:59n1smostly the usual crashes with files that have odd metadata is my guess
23:23:01Lloreangamefreak264: Nothing harmful that I've heard about. And given that it's generally individuals reporting it without others saying "hey, me too" they may not be 3.8 issues at all, but rather the kinds of problems people occasionally run into.
23:23:19LloreanSince a new release brings a lot of updates, a lot of people can have the chance for one-off issues.
23:23:36gevaertsNew releases also can bring new users, with new corrupted filesystems :)
23:23:51n1sand new exotcily broken files :)
23:23:57 Nick linuxstb_ is now known as linuxstb (~linuxstb@rockbox/developer/linuxstb)
23:23:58 Join faemir [0] (~tarq@
23:24:14gevaertsAnd new dodgy USB cables
23:24:41Lloreangevaerts: I wish they'd try the RC and find the bugs *then*.
23:24:44gamefreak264What motivates you guys to do this? It seems like a dreadful hobby to me
23:24:54faemirHey all, I was wondering whether I can tell how loud I'm playing my music in terms of dbA?
23:25:02gamefreak264Anyway, thanks for devoting your time to making my life more convenient :P
23:25:17stripwaxLlorean - to be honest, I didn't see a lot of advertisement for the RC.
23:25:40Lloreanstripwax: True.
23:25:50Lloreanstripwax: I'm hoping for next time we can put the RC in RBUtil somehow
23:25:56stripwaxfaemir - in rockbox, the volume is in terms of dB where 0 is line levels. the actual loudness depends on your headphones (which obviously rockbox cannot know about)
23:26:09faemirstripwax, I guessed as much, is there a way to measure it?
23:26:37faemirI mean, -25 is good and -20 is loud for me, which suggests they are rather loud
23:26:53faemirI think
23:27:39stripwax-20 is almost 4x the loudness of -25. (sortof)
23:28:16thomasjfoxgamefreak264: When you've used rockbox for some days, you'll know ;)
23:28:31 Join [Saint] [0] (S_a_i_n_t@
23:28:47faemirstripwax, the headphones say 96db @1khz if that helps?
23:29:12sideralAlexP: Where do we store .map files for the 3.8 builds? I'd like to investigate one of the 3.8 DB commit crashes reported here last night
23:29:44Lloreanfaemir: That's noise. There's really no way to know outside of getting something that actually measures loudness.
23:30:24AlexPsideral: I don't know where Zagor put them (or if he even took them actually). One mo, I'll get you another link to it
23:30:55faemirLlorean, thought so, well I'll just keep it low and hope for the best I guess, I listen to about an hour a day so I'm sure I couldnt be damaging my ears toooo much
23:31:43gamefreak264Are there any sorts of things you think I should know as a newcomer to rockbox about to do his first install? I'm using the rockbox utility and not the manual install method, by the way.
23:31:49LloreanAt -25 I wouldn't imagine an hour a day is doing any real damage.
23:32:01stripwaxfaemir - to be honest, if it feels loud, it probably is, so turn it down to a level that is quiet enough to still hear fine. -32 works fine for me, sometimes quieter. i have pretty sensitive headphones though.
23:32:21*Llorean listends at -30 to -70 depending on the time of day.
23:32:48AlexPsideral: check PM
23:32:57faemirstripwax, yeah mine are pretty amazing at isolation, but I reckon I damaged my hearing pretty badly a few years back on the school bus with a cd player on max to just hear through the racket :(
23:34:06AlexPZagor: Did you take the elf and map zips from the debug folder for the release? If not, they are still where the builds were, is it possible to stick them on somewhere?
23:34:46faemirLlorean, yeah if I fall asleep to music I'll have it way down in the 70s, but I don't like rolling and damaging/tangling my earphones
23:35:05 Quit sampattuzzi (Remote host closed the connection)
23:35:24 Join ReT [0] (
23:35:29ReThey good afternoon folks
23:36:09stripwaxReT - depends on your timezone :)
23:36:13ReTI just got an iTrip FM transmitter for my iPod video but it dosen't work on rockbox, which is a major bummer because rockbox is awesome
23:36:37stripwaxReT - with Accessory Power enabled?
23:37:11ReTErr good question
23:37:32sideralstripwax, AlexP: Strange, stripwax reported a 3.8 bug for iPod 5G, but the link you sent me doesn't seem to contain a build for a 5G iPod. Does that target go by another name?
23:37:36stripwaxReT -
23:37:47stripwaxsideral - ipodvideo
23:37:53sideralAh, OK
23:39:35ReTStripwax: Thanks man I will try that out
23:40:25ReTStripwax: I would be super bummed if I couldn't get it to work heh ;)
23:40:58faemirstripwax, scrap that, fitted in different size inners and even -30 sounds loud
23:41:23stripwaxfaemir - :-)
23:41:45faemircrazy how much that counts for, -20 was painfully loud suddenly
23:43:26ZagorAlexP: no, I didn't take them. I guess we could put them in a debug dir.
23:43:49AlexPZagor: It might be useful for some, and everyone else will never see them
23:45:57***Saving seen data "./dancer.seen"
23:46:04 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
23:47:26 Join wodz [0] (
23:48:17wodzn1s: genlang profiling ->">
23:48:52n1swodz: i didn't look into it because i try not to toutch perl unless i have to :)
23:49:24ReTStripwax: Still not working. Accessory power is on and the itrip appears to be working but no FM modulation
23:49:32wodzI understood you want to speed up genlang
23:49:43stripwaxReT - you might also need to turn on Line Out!
23:49:52 Quit gamefreak264 (Quit: I QUIT!)
23:50:00ReTStripwax: That is on
23:50:33stripwaxReT - sorry I can't help more :-( I don't have any ipod accessories personally
23:50:58[Saint]If accessory power and line out are both on, I would expect it to work.
23:51:08n1swodz: i'd like it to be faster and have thought abotu it but i'm not good enough at perl and don't want to rewrite the whole thing in c
23:52:26wodzn1s: from profiling it seems that clever rewritten regexp could help
23:53:19sideralstripwax: Re FS #11976: The 4e??? address you reported is in tagcache_search, which AFAICT isn't used in DB commit. So either that address is wrong, or the DB wasn't committing
23:53:21 Quit komputes (Remote host closed the connection)
23:54:03ReTOh well
23:54:17n1swodz: yes
23:55:05stripwaxsideral - does Auto Update use that function?
23:55:32stripwaxsideral - if not, maybe some bad interaction between whatever was happening with my menu interaction and whatever the database was doing at that moment in time?
23:55:41stripwaxcould it be WPS perhaps?
23:55:59sideralstripwax: It's likely used in DB generation −− checking...
23:56:10stripwaxoh. well in that case... that makes sense
23:56:15sideralBTW, the path name in your metadata log looks odd.
23:56:22 Join Keripo [0] (
23:56:24stripwaxit would have had to have happened at pretty much the very last step of db generation though.
23:57:03stripwaxsideral - the path name of which file(s)?
23:57:19stripwaxoh, the last ones. yes, I agree. that seems odd.
23:57:30sideralHave you tried removing this directory?
23:57:35stripwaxin file browser the path just says The Magic Numbers. I have no idea what that weird stuff is.
23:58:11stripwaxsideral - no, I haven't tried removing it. The database built successfully when I just used Initialise Database from the menu.
23:58:48stripwaxhang on a minute
23:58:51sideralstripwax: Interesting taste of music BTW :)
23:58:57stripwaxthat stuff at the end of the metadata log is my *playlist*

Previous day | Next day