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

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

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

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

#rockbox log for 2010-09-26

00:05:20 Quit domonoky1 (Read error: Connection reset by peer)
00:06:49 Join robin0800 [0] (~robin0800@
00:08:07 Join TheSeven [0] (5d84be41@rockbox/developer/TheSeven)
00:21:31 Join s1gma_ [0] (~d.d.derp@
00:22:38 Quit _s1gma (Ping timeout: 276 seconds)
00:22:55 Join insp_ [0] (
00:39:15 Quit Staphylo (Quit: Bye les gens =))
00:43:59 Quit GeekShadow (Quit: The cake is a lie !)
00:46:39CIA-81New commit by 03Buschel (r28165): Add current consumption and battery capacities to iPod nano 2G config file.
00:47:46*TheSeven likes the fact that Buschel steals items from his todo list :)
00:48:16Buschel:-) the battery bench and the discharge curve is still missing though
00:48:21CIA-81r28165 build result: All green
00:48:33*TheSeven committed a discharge curve once
00:49:34TheSevenhm, someone stole that, too
00:50:09Buschelyes, there is one. but to me it seems the included one will reach 0% too early
00:50:27Buschelbtw, any chance to work on LCD_SLEEP?
00:50:40Buschelthis might bring in some more savings...
00:51:37TheSevenah, there it is
00:51:40TheSevenit only moved a bit
00:51:43TheSevenconst unsigned short percent_to_volt_discharge[BATTERY_TYPES_COUNT][11] = { { 3550, 3783, 3830, 3882, 3911, 3949, 3996, 4067, 4148, 4228, 4310 } };
00:51:50TheSevenlooks fine to me
00:53:35TheSeveni tweaked the lowest two values a bit to give a better average accuracy in the nearly-empty area
00:54:56TheSevenshutoff at 3.35V was needed so that the ipod has enough charge left to do a worst-case ftl commit (about 30 seconds of writing to flash)
00:55:05Buscheldon't you think it would be better to even use lower voltages for "dangerous" and "shutoff"? both are >>3.0V which is the lowest voltage used player internally (from what I can see)
00:55:10TheSeventhe flash is powered through an LDO linear regulator at 3.00V
00:55:30TheSevenso we need to still be a bit above of that to make it work in a stable fasion
00:55:31Buschelah, you just typed the answer for the question I was asking
00:56:00Buschelso, then the battery stuff is finished :)
00:56:03TheSeventhe voltage falls rather fast starting at about 3.30V
00:56:17TheSevenat 2.90V the CPU will die
00:56:52TheSeveni don't know if the power manager shuts it off deliberately or if it just fails because the RAM voltage drops too much (it's also configured at 3.00V)
00:57:17TheSeventhe CPU internally runs at 1.00V or 1.05V through a switching regulator
00:57:47Buschelis it possible to use lower voltage when the cpu is clocked down to 48 MHz?
00:58:08TheSevenyes, I managed to get it down to 0.95V but that doesn't save much
00:58:25TheSevenit even seems to run at that voltage at 200MHz
00:58:40Buschelwhat were the savings?
00:58:47Buschel<1 mA?
00:58:47TheSevenmaybe 1mA
00:59:04TheSeventhe problem is that the XTAL oscillator frequency drops noticably below 0.925V on mine
00:59:25TheSeventhis was reported to happen at even higher voltages on some devices
00:59:52TheSeveni once managed to run it at 0.85V at 48MHz, but the real cpu clock was more like 43MHz
01:00:30*TheSeven didn't know that xtal oscillators can behave that badly
01:00:53TheSeventhe problem with frequency-dependent undervolting is that switching the voltage takes way too long
01:01:11TheSevenapparently the cpu frequency change function can be invoked from an IRQ context
01:01:30 Quit insp_ (Quit: ChatZilla 0.9.86 [Firefox 3.6.8/20100722155716])
01:01:36TheSevenand we need to wait for I2C transactions to complete and voltages to stabilize before raising the clock again
01:01:50TheSevenif that could be split into multiple operations, this might help
01:01:59TheSevenbut I can't block IRQs for that long without disrupting playback
01:02:25Buschelso, best would be to either find a lower voltage that also fits for 192MHz or just keep the actual voltage
01:02:49TheSeventhat I2S core is extremely sensitive to FIQ latencies
01:02:50 Quit s1gma_ (Ping timeout: 265 seconds)
01:03:11TheSevenmore than about 20 *microseconds* of latency until the next DMA transfer is started and you'll notice clicking
01:03:48TheSevenso we can't even afford to lock FIQs while waiting for a PLL to lock
01:04:25TheSeven(that's why I'm using an assembly FIQ handler)
01:04:40Buschelthat's why you are using the fastbus-option to force the cpu clock to 48 MHz?
01:04:52 Join Nausicaa [0] (
01:04:57TheSeveni can't stop the pll, because I don't manage to get it running again fast enough
01:05:35TheSevenif we would change the core in a way that allows boosting/unboosting to be split into multiple operations, that might help
01:05:50TheSevenor at least get rid of the calls from IRQ context
01:05:58TheSeventhat would allow us to block at least
01:06:07TheSeventhen we could also change the voltage
01:06:17TheSevenpowering down the unused PLL saves about 2mA IIRC
01:06:40Buscheldo we already do this?
01:06:47soapHas any disassembly been done on the OF?
01:06:57TheSevena bit, but not much of it
01:07:25TheSevenmost of it was reverse engineered from the apple bootloader and diagmode
01:08:02soapI'm curious if any knowledge has been gleaned on how the OF manages the CPU frequency, since it sounds much harder to maximize part-turnoff-savings than on the previous ipods I was curious if the OF even bothered.
01:08:25*TheSeven has no idea
01:08:29soap(not that I have ANY insights, just curiosity)
01:08:51TheSevenone might want to look at their FIQ handler
01:09:02TheSeventhey might have found a better way to deal with the I2S DMA
01:09:26TheSevenauto-reloading DMA didn't work for me for some reason, and they didn't use it in diagmode at all
01:10:37 Quit Highlander (Quit: Quitte)
01:12:42 Quit engwan_ (Ping timeout: 276 seconds)
01:14:51BuschelTheSeven: the LCD driver has lots of waits between each 16 bit write access to the 16bit data register. if I remove one of those waits between 2 write accesses, the speed is increased and i can see no visible impact to the screen.
01:18:21 Quit TheSeven (Ping timeout: 252 seconds)
01:23:23 Join _s1gma [0] (~d.d.derp@
01:23:45 Quit robin0800 (Ping timeout: 276 seconds)
01:29:48soapsecond half of his post might be relevant to you.
01:30:23soap28164 to blame?
01:32:29Buschelsoap: yep. works for me though... will create a flyspray patch. it might be connected to different LCD types...
01:32:43 Join TheSeven [0] (5d84be41@rockbox/developer/TheSeven)
01:33:04TheSevensorry, webchat went dead for some reason
01:33:16TheSevenbut in the meantime i had a quick glance at the OF's FIQ handler:
01:33:19TheSevenSTMFD SP!, {R0-R12,LR}
01:33:39TheSevenapparently they didn't quite realize what kind of advantage an FIQ offers in terms of registers
01:33:52 Quit PaulJam (Ping timeout: 272 seconds)
01:33:59TheSevenso I think they aren't suffering as bad from the latency problem as we are
01:35:13TheSevenactually an FIQ seems to be considered a fatal error
01:37:15CIA-81New commit by 03Buschel (r28166): Roll back r28164 as this change introduced LCD issues on some nano 2G.
01:38:47CIA-81r28166 build result: All green
01:39:35 Join antil33t1 [0] (
01:39:39 Quit antil33t (Disconnected by services)
01:42:54TheSeventheir IRQ handler seems to be doing a context switch
01:43:00TheSevennot very latency-oriented either
01:44:11BuschelTheSeven: You have got a type1 or type0 LCD in your nano 2g?
01:45:02*TheSeven doesn't remember the way they were numbered
01:45:05TheSevenI have the LDS176
01:45:29Buschelsame than I have...
01:46:30TheSevenif you need init code, there's some code in norloader/embios loader that should do the trick, but it's a bit fishy
01:47:11TheSevenalso IIRC there's a patch on flyspray that's dealing with LCD shutdown, don't remember if it was yours or liar's or someone else
01:51:14***Saving seen data "./dancer.seen"
01:53:17*clone4crw wonders if gui_synclist is documented in a way that makes sense to someone who's never used it before.
01:56:06 Quit Zambezi (Read error: Connection reset by peer)
02:00:43 Quit DerPapst (Quit: Leaving.)
02:04:20TheSeventhis is just ridiculous
02:04:52*TheSeven has never seen such an inefficient IRQ handler before
02:14:49 Join Zambezi [0] (~Zulu@
02:15:29*TheSeven doesn't manage to figure out where the handlers for the individual IRQs are
02:16:00 Quit clone4crw (Remote host closed the connection)
02:16:09TheSeveni found the place calling them, but can't find anything that's writing to that table after it is initialized (zeroed) and before it is used
02:18:32 Join NoGare [0] (~barry@unaffiliated/nogare)
02:22:30TheSeventhis crazy thing is dealing with linked lists of structs just to do the IRQ mapping apparently
02:23:48TheSevenapparently that beast is even calling malloc() in the irq handler...
02:39:10 Join Synthbox [0] (~x@
02:40:07 Join kugel [0] (~kugel@rockbox/developer/kugel)
02:44:48 Join kugel_ [0] (
02:45:10 Quit kugel (Disconnected by services)
02:45:15 Nick kugel_ is now known as kugel (
02:45:19 Quit kugel (Changing host)
02:45:19 Join kugel [0] (~kugel@rockbox/developer/kugel)
02:48:30*TheSeven just spotted the timer irq handler, but that's the only one that's registered from known code
02:48:45TheSeventhere are probably more calls to this one in sections not yet marked as code
02:53:03 Quit dfkt (Quit: -= SysReset 2.53=- Sic gorgiamus allos subjectatos nunc.)
02:54:58 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
03:03:30 Quit Buschel (Quit: ChatZilla 0.9.86 [Firefox 3.6.10/20100914125854])
03:07:04TheSevenBuschel (for the logs): re r28161: I don't think this will make a difference at all
03:07:44TheSeventhis is probably in the range of a few microamps
03:08:30TheSevendid you verify that one can still wake up the ipod after powering down with the hold switch locked with that change?
03:10:43TheSevenand re: #define CURRENT_NORMAL 21 /* playback @48MHz clock, backlight off */
03:10:52TheSeventhis appears to be a bit too high to me
03:11:08TheSevenIIRC i had values around 18mA on mine
03:11:54 Quit MethoS- (Remote host closed the connection)
03:12:01TheSevenjust tried and got 40mA! wtf?
03:13:12TheSevenaha. usb seems to be the culprit
03:14:11TheSevenafter plugging unplugging usb: backlight off: 17-18mA, backlight on (brightness 24): 23-24mA
03:14:34TheSevenat brightness 46 I get 42-43mA
03:15:33TheSeven(measured at 4.19V battery voltage, 320kbps mp3)
03:17:30TheSevenwith 16 ohms earphones at -1db (clipping limit), it's at 17-31mA
03:18:06TheSeven17-18mA at -20dB
03:18:43TheSevenmeasurements done using r28047M
03:21:34 Quit Judas_PhD (Quit: This is a quitting message)
03:22:15 Quit kugel (Remote host closed the connection)
03:36:48*TheSeven just located the DMA IRQ handler
03:43:59 Join Kitr88 [0] (
03:46:52 Quit Kitar|st (Ping timeout: 240 seconds)
03:48:36 Quit Kitr88 (Ping timeout: 255 seconds)
03:51:18***Saving seen data "./dancer.seen"
03:53:58 Join Kitar|st [0] (
04:01:35 Join mikewkrc [0] (
04:03:51mikewkrcwow somebody spent a really long time putting together the "read this first"
04:14:43mikewkrcbrb rebooting into unbuntu for first try at running a patch/compile of rockbox :) (wishing myself luck :) )
04:14:59 Join Judas_PhD [0] (
04:15:06 Quit mikewkrc ()
04:19:15 Join mikewkrc [0] (
04:21:27 Quit NoGare (Quit: leaving)
04:25:22 Quit pixelma (Disconnected by services)
04:25:23 Quit amiconn (Disconnected by services)
04:25:24 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:25:25 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:25:26 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:25:31 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:25:55 Quit Synthbox (Ping timeout: 240 seconds)
04:39:22mikewkrcThank the computer gods for inventing i7 compiling is sooo fast
04:47:28 Join clone4crw [0] (
04:49:10mikewkrccompiling with a 1.8 system load :)
04:53:10 Join Barahir_ [0] (
04:57:01 Quit Barahir (Ping timeout: 276 seconds)
04:59:10soapyou need faster drives.
05:02:08mikewkrcholy crap it worked
05:02:23mikewkrcinstalled psgroove through rockbox first try :)
05:02:29mikewkrcon my sansa e260
05:04:07 Join xavieran [0] (
05:04:33 Quit xavieran (Remote host closed the connection)
05:04:52 Quit Zambezi (Read error: Connection reset by peer)
05:05:04 Join xavieran [0] (
05:05:57 Quit rvvs89 (Ping timeout: 272 seconds)
05:14:27 Join Zambezi [0] (Zulu@
05:20:32 Quit _s1gma (Quit: Leaving)
05:31:47 Quit ps-auxw (Read error: Operation timed out)
05:36:49 Quit Zambezi (Read error: Connection reset by peer)
05:39:28 Join rvvs89 [0] (rvvs89@pdpc/supporter/base/rvvs89)
05:40:02 Quit Judas_PhD (Quit: This is a quitting message)
05:45:21 Join ps-auxw [0] (
05:51:20***Saving seen data "./dancer.seen"
05:54:32 Join clone4cr1 [0] (
05:57:14 Quit clone4crw (Ping timeout: 255 seconds)
05:58:43 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
05:58:55 Quit clone4cr1 (Ping timeout: 240 seconds)
06:02:06 Join fyre^OS [0] (
06:04:38 Quit fyrestorm (Ping timeout: 276 seconds)
06:17:02 Quit bluebroth3r (Ping timeout: 255 seconds)
06:18:45 Join bluebrother [0] (
06:18:45 Quit bluebrother (Changing host)
06:18:45 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
06:28:09 Join engwan_ [0] (~engwan@
06:48:46 Join Zambezi [0] (Zulu@
07:02:59 Quit sasquatch (Ping timeout: 276 seconds)
07:06:42 Join sasquatch [0] (~username@
07:18:41 Join dys [0] (
07:22:59 Join Horscht [0] (~Horscht@xbmc/user/horscht)
07:25:05 Quit Horschti (Ping timeout: 265 seconds)
07:41:59 Join shai [0] (
07:47:19 Quit anewuser ()
07:51:21***Saving seen data "./dancer.seen"
08:09:55*JdGordon contemplates bringing up the question of making the quickscreen button customisable to be either the QS or hotkey....
08:15:28 Quit slothearn (Quit: Lost terminal)
08:16:19 Quit Nausicaa (Ping timeout: 265 seconds)
08:19:30 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
08:35:41 Join Jaykay [0] (
08:43:36CIA-81New commit by 03jdgordon (r28167): Fix FS #1159 - stack overflow in the skin engine when there is too many nested conditionals
08:45:30CIA-81r28167 build result: All green
08:47:10JdGordonwhy are 3 of those builds such a large green delta?
09:03:37 Join Rob2223 [0] (
09:03:48bertrikI saw some corruption in the wps (default cabbiev2) on my clip+ yesterday, but it only happened when timestretch was enabled. I also saw a comment on the forums from someone who got data aborts with timestretch enabled.
09:03:59bertrikI guess something's dodgy with timestretch.
09:06:17 Quit BHSPitMonkey (Remote host closed the connection)
09:06:35*JdGordon slaps kugel for commiting // comments
09:06:35 Quit Rob2222 (Ping timeout: 240 seconds)
09:06:49 Join Spaceghost1 [0] (~pidgin@unaffiliated/spaceghost)
09:07:00Spaceghost1hello folks
09:07:44Spaceghost1I read here:
09:08:06Spaceghost1that this memory Transcend 32GB 133x CF card worked on a 2g Mini
09:08:38Spaceghost1if that card worked on the mini, probabily that
09:09:40Spaceghost1will work on a 4th generation ipod if I buy a 1.8 to CF adapter? like that:
09:10:03JdGordonit should
09:12:04Spaceghost1what about use a microphone on ipod?
09:12:16Spaceghost1is really hard get one that work very well on it?
09:13:24Spaceghost1if I think give (in others) that use to the iphone, probabily is better that buy another cheap player?
09:14:02 Join Judas_PhD [0] (
09:14:16 Quit Judas_PhD (Client Quit)
09:17:06 Quit AoEKiller ()
09:21:20 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
09:32:19CIA-81New commit by 03jdgordon (r28168): Fix FS #11552 - touches outside of the UI viewport can do unexpected list movements. ...
09:34:05CIA-81r28168 build result: All green
09:45:42 Join Buschel [0] (
09:46:18 Join Judas_PhD [0] (
09:49:29BuschelTheSeven: I measured the 21 mA with headphone at about -30dB, backlight off and playing mpc files. in this state the debug screen showed 20-21 mA. this also perfectly matches a runtime test I have done −− result was ~19h.
09:51:09BuschelTheSeven: and of course this was measured after a clean restart.
09:51:22***Saving seen data "./dancer.seen"
09:56:47BuschelTheSeven: from a psychological point of view it is imho the better choice to perform better in real life than the pure numbers say :-)
09:58:15 Quit Guest934 (Ping timeout: 264 seconds)
10:10:24BuschelTheSeven: just retested -> still 20-21 mA. only when using 950mV for the CPU I get 18-20 mA.
10:11:55 Quit scorche (Read error: Connection reset by peer)
10:15:36 Join kugel [0] (~kugel@rockbox/developer/kugel)
10:15:38 Join PaulJam [0] (
10:16:56 Join scorche [0] (~scorche@rockbox/administrator/scorche)
10:18:23 Join n1s [0] (~n1s@rockbox/developer/n1s)
10:20:44 Quit sasquatch (Ping timeout: 276 seconds)
10:29:53 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
10:33:05 Join sasquatch [0] (
10:43:25pixelmabertrik: the screen corruption thing is a bit irregular for me (e.g. on my own Ondio WPS I usually don't see corruption but sometimes a few bitmaps are missing but can reappear after reboots or reloads of the theme, got a few unnecessary pixels only once)
10:44:10 Join ender` [0] (
10:50:15 Join Spaceghost [0] (~pidgin@unaffiliated/spaceghost)
10:53:03 Quit Spaceghost1 (Ping timeout: 272 seconds)
10:54:00bertrikpixelma, I noticed that timestretch uses buffer_alloc to get memory and I guess the skin stuff does too. Maybe we could add some debugging to that.
10:54:34 Quit scorche (Read error: Connection reset by peer)
10:55:20bertrikLike adding magic markers / ids around the allocated memory blocks and checking them regularly
10:55:55n1sbertrik: you mean like the canary value we use on stacks?
10:56:05n1sto check for overflow
10:56:44gevaertssounds like a good idea
10:57:43n1syeah, memory corruption is no fun to debug
10:58:09bertrikMaybe a buffer_alloc is silently failing, resulting in corruption later on
10:58:54JdGordonI dont think so
10:58:57JdGordonI mean, I doubt it
10:59:08bertrikI see there is currently no protection/check against too much memory allocated.
11:00:55n1syeah, buffer_alloc just increments the audiobuf pointer with the requested size with no checks whatsoever
11:01:21 Quit amiconn (Remote host closed the connection)
11:01:22 Quit pixelma (Remote host closed the connection)
11:01:34 Join pixelma [0] (quassel@rockbox/staff/pixelma)
11:01:47 Join amiconn [0] (quassel@rockbox/developer/amiconn)
11:03:45n1sa simple if (audiobuf+size > audiobufend) return NULL; should do no?
11:04:03n1sand some checking in the places that do the allocations of course
11:04:25JdGordonwell yes and no
11:04:44JdGordonisnt the problem that alot of things wont fail gracefully if it doesnt get the ram
11:04:58 Quit bertrik (Read error: Connection timed out)
11:05:27n1sthat is buggy then and should be fixed, i think allocating memory that doesn't exist (or even memory allocated by something else) is worse
11:05:43gevaertsJdGordon: sure, that too. But if buffer_alloc() doesn't even tell them, then...
11:05:47 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
11:05:51n1siirc the codecbuffer is at the end of the audiobuffer so if we alloc too much we could overwrite that
11:06:20JdGordonand also the order is close enough to random to causing problems... a bunch of non-essential buffers could be being allocated (and work) before a big essential buffer
11:07:37n1sthat really has nothing to do with proper eror checking, but sounds like it needs a think-through
11:08:11kugelsome parts don't even use buffer_alloc, but modify audiobuf directly
11:08:40JdGordonkugel: not after audio is init-ed though...?
11:08:59bertrikn1s, yes I meant some kind of canary value, plus maybe some extra info (block size) so we can parse the alloc buf
11:09:33 Quit Spaceghost (Read error: Connection reset by peer)
11:09:37gevaertsYes, we'd need to keep more information of course
11:10:13 Join DerPapst [0] (
11:10:20JdGordonbertrik: we could do that with logf just as easily though... if there is no protection then timestrecth+skin buffer could probably be naughty, should be just a matter of checking the audio buffer size int he debug screen though
11:12:14pixelman1s: codecbuffer is unlikely because I see problems on my hwcodec target too
11:12:19gevaertsExcept that logf isn't any good in official builds
11:12:49JdGordonof course, but that sounds like a waste in regular builds also
11:13:17gevaertsHow many buffer_alloc()s does a regular running system do?
11:13:51n1spixelma: ah
11:14:00JdGordonif audio playback works at all then it is almost certainly not the actuall alloc() which is causing problems
11:14:18JdGordongevaerts: hundreds if dircache is enabled
11:14:33gevaertsWe're talking about a few lines of code, and a few bytes (8? 16?) of metadata (length+magic value to check) in before and after each block I think
11:14:35JdGordonotherwsie I'd guess a dozen or so
11:15:26gevaertsMight still be a good idea to implement, even if it's not added to default builds
11:15:55gevaertsEven if the issues we're talking about now probably aren't related, some issues in the past have definitely been
11:19:24JdGordon*cough* proper buflib
11:19:33bertrikyeah, we could make this extra administration optional (#ifdef it) if people object to the few extra bytes it takes
11:19:56bertrikwe could also store some kind of thread id with the blocks to know who allocated it
11:20:30JdGordona string would be more useful, just about all would be on the main thread
11:20:38bertrikand have a debug screen with stats
11:22:29gevaertsI think having buffer_alloc() check if there's enough room and panic if not should be there
11:22:54JdGordonpixelma: check the info screen, whats the audio buffer size say?
11:23:01gevaertsOr return NULL and hope that callers check
11:23:14 Quit Buschel (Quit: ChatZilla 0.9.86 [Firefox 3.6.10/20100914125854])
11:23:22bertrikgevaerts, yes that's a good start with very limited impact
11:24:14kugelJdGordon: IIRC dircache doesn't call buffer_alloc
11:24:19*JdGordon is charging his v1 clip to prove this isnt the problem
11:24:22pixelmaJdGordon: audio buffer?
11:24:34kugeland not hundreds of times anyway
11:24:51JdGordonpixelma: the "Buffer" line
11:24:53kugelI think there're under 10 buffer_alloc calls in total
11:25:37pixelmaJdGordon: on my Ondio? It says "1.22 MB", not sure what this is going to tell though
11:26:17JdGordonthat says that buffer_alloc(0 isnt running out of room
11:27:12JdGordon$ find -iname "*.c" | xargs grep "buffer_alloc(" | wc -l -> 64
11:27:36bertrikthat also includes a lot of skin_buffer_alloc's I think
11:28:59kugeloh I was wrong, dircache buffer_allocs a lot
11:33:01JdGordontagcache is also a big user
11:33:10 Join Giova [0] (~giovanni@
11:36:17GiovaJdGordon: Thanks for r28168, scrolling the menu via soft keys now working really good.
11:37:58Giovabut what about r28146? It looks terrible on my onda
11:40:06Giovascrolling the menu in absolute mode, made the selection bar disappear, maybe on Android it can be used, I don't like it on vx777
11:40:50kugelI find showing a selection bar during scrolling looks terrible
11:43:37Giovawhat about give to the user an option for that?
11:44:15bertrikgevaerts, n1s : something like this ?
11:44:20 Join scorche [0] (~scorche@rockbox/administrator/scorche)
11:44:49gevaertslooks like a good start
11:45:10JdGordonbut that isnt the problem!
11:45:25JdGordonand my damn clip isnt charging so i cant prove it :/
11:45:31gevaertsJdGordon: who cares? :)
11:45:41gevaertsIt might be a problem at some point!
11:45:50 Join MethoS- [0] (~clemens@
11:46:08kugelbertrik: you could use %zu for size_t in format strings
11:48:06kugelmaybe timestretch is overflowing its buffer
11:48:26bluebrotheris there a reason doesn't use −−disable-nls?
11:48:30JdGordonthat sounds more likely
11:48:35*gevaerts is looking at adding overflow checking
11:48:50gevaertsbluebrother: n1s might object to that :)
11:49:03gevaertsOh, *l*
11:49:28bluebrothergevaerts: he might have a problem with my mac then. I get binutils building only with −−disable-nls ;-)
11:50:09bluebrotherbut do we want localization for the cross compiler anyway?
11:51:10 Quit Giova (Read error: Connection reset by peer)
11:51:25***Saving seen data "./dancer.seen"
11:52:50pixelmaJdGordon: I'm getting a "I09: CPUAdrEr"2 with r28168 on my Ondio with the "offending" theme, looking up the address after lunch
11:53:17pixelmaa simpler one wrt nested conditionals works
11:54:40gevaertsIs there a reason why buffer_init() handleds the alignment, and not attributes or the linker?
11:55:33 Join user890104 [0] (
11:57:14pixelmaJdGordon: seems to be in the scroll_engine
11:57:32bluebrotheryay, eabi toolchain working on OS X 10.6 :)
11:58:52*bluebrother wonders if he should commit his changes to or post a patch to get others try it first since he's using MacPorts quite a bit
12:00:08 Join Giova [0] (~giovanni@
12:01:35pixelmabluebrother: there was someone in the forums a few days ago with MacOS 10.6, maybe you can ask there?
12:01:51bluebrotherpixelma: yeah, that was the reason for looking into it again :)
12:02:09bluebrotherguess I'll post a patch and post the id to that thread
12:06:14gevaertsWhere is the code that checks for stack overflow?
12:06:39gevaertsah, thread.c
12:10:39 Join evilnick [0] (~Evilnick@rockbox/staff/evilnick)
12:12:05 Quit Judas_PhD (Quit: This is a quitting message)
12:13:35gevaertsThat still needs some #ifdefs
12:13:45gevaertsIt's also untested
12:15:00*bertrik is looking at too many things at the same time
12:15:06 Join Judas_PhD [0] (
12:15:09gevaertsI think it doesn't just "waste" some memory, it's probably also too slow to enable all the time
12:18:31gevaertsOh, and of course if an interrupt handler overwrites memory, it will blame the wrong thread
12:18:49bertrikclip+ with both tagcache and timestretch enabled fails in various strange ways
12:19:23 Join Jerom [0] (~heidi@
12:24:11JdGordondoes it fully boot though?
12:25:11bertrikit wouldn't boot with a microsd inserted
12:29:42gevaertsYay, my patch makes my c250 not boot :)
12:31:28JdGordondon't thread switches need to be really *really* fast :D
12:31:51gevaertsyes, that's why this can't be enabled by defaukt
12:32:06JdGordonor apparently at all? :)
12:32:09gevaertsThe not booting thing is that it claims buffer corruption
12:32:30bertrikgevaerts, on my clip+ I get a "tagcache corrupted buffer 300BC680 start" on boot :)
12:32:44CIA-81New commit by 03kugel (r28169): Android: Exclude the main binary from make zip.
12:33:01gevaertsbertrik: yes, same symptoms :)
12:33:04*gevaerts looks into it
12:33:15gevaertsMaybe logf is better than panicf to get started
12:33:46JdGordonwhat needs to be done to add the sdl app to the build table?
12:34:25CIA-81r28169 build result: All green
12:38:06gevaertsJdGordon: Zagor did some work on that recently. I suspect he may have a plan in mind
12:38:11 Quit user890104 (Ping timeout: 272 seconds)
12:39:01JdGordonok, I noticed android is in the table
12:42:33 Join [sko] [0] (~sko]
12:43:02 Join teru [0] (
12:43:03 Quit [sko] (Client Quit)
12:43:30 Join [sko] [0] (~sko]
12:44:30 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
12:45:07gevaertshm, I can't remember now...
12:45:14gevaertslogf dumps are in reverse order, right?
12:45:44gevaertsno, they can't be
12:47:25JdGordonpanic from audio_reset_buffer with svn...
12:47:59JdGordonhow do i reset the settings on boot on the clip?
12:50:40JdGordonwith timestrecth, cuesheet, lastfm all enabled i still have 800KB buffer
12:50:55JdGordonsetting both limit settings to max crashes it thouhg
12:51:13JdGordonand database enabled also
12:51:44kugel <−− pixel accurate list scrolling
12:51:49kugelandroid builds updated
12:51:53JdGordondata abort starting playback though
12:52:50soapIf I were to create a patch which /clearly/ will never be committed, yet wanted to share it with others who might find it useful, could I host it on FS, or should I host it on my site?
12:53:43JdGordonkugel: are you purposly reverting my first fix commit from a few hours ago with that diff?
12:54:04JdGordonremoving the code under /* make sure it is inside the UI viewport */
12:54:06kugelno, it's not reverted
12:54:33kugelaction_get_touchscreen_press_in_vp() is called
12:54:47JdGordonah right, my bad
12:55:51bertriksoap, I probably wouldn't mind having it on FS, if it's a kind of "research" patch for example
12:56:41pixelmarecently I _sometimes_ get a "STKOV Main" on my M5 when skipping, going to file browser or changing volume or something like this from the WPS
12:56:51kugelsoap: we only want patches on fs that seek inclusion
12:56:58soapNothing that defensible. bertrik.
12:57:21AlexPsoap: It has been said in the past that flyspray is just for things working towards being committed
12:57:30AlexPWe've been closing those that never will be
12:58:11terucould anyone give me a comment about FS #6321? i'm not sure what to do about it.
12:59:12AlexPteru: I think a universal image viewer plugin would be a good thing
12:59:25AlexPTalk to wodz also, he has been working on a png viewer
12:59:59kugelteru: you don't need rb->cpucache_invalidate() if you use lc_open
13:00:14AlexPFS #11641 I think
13:00:43gevaertsDefine BUFFER_ALLOC_DEBUG to use it
13:00:49AlexPteru: Yeah, FS #11641
13:00:49pixelmateru: are you the one recently working a lot on the text viewer?
13:01:47teruAlexP: thanks, i saw it.
13:01:53AlexPok, cool
13:02:08bertrikgevaerts, I'll try it right away
13:03:44JdGordonkugel: not bad... its not using the full screen width apparently
13:04:09kugelwhat do you mean?
13:04:26terukugel: I'll remove it then, thanks.
13:04:29JdGordonI dont think you actually compiled it for 480x800 :)
13:05:18kugeloh yes, 320x800 :\
13:05:54kugelbecause tools/configure doesn't work properly
13:06:44bertrikgevaerts, it boots and plays OK with your patch and timestretch enabled
13:07:00terupixelma: i think i don't touch the text viewer recently?
13:07:28bertrikwhen I also enable dircache, it still boots but I do see a WPS corruption
13:08:00pixelmateru: I didn't mean "recently" as "in the last week(s)" or so, more like "the latest changes to it were yours"
13:09:11bertrikgevaerts, never mind, I forgot the BUFFER_ALLOC_DEBUG
13:11:15kugelJdGordon: updated
13:12:29pixelmateru: if so, I wondered about the blank line at the bottom even with page numbers turned off. It's wasteful on small displays
13:12:59kugeluchida worked on it
13:13:35pixelmayeah, I wasn't sure who it was (which is why I asked before)
13:15:15JdGordonandroid really does need the hardware buttons to work :/
13:15:36kugelthey should work, but they don't. the code is in svn
13:16:34JdGordonalso kinetic scrolling
13:16:40JdGordonor whatever the heck its called
13:16:47kugelwe're working on that :)
13:17:00JdGordonoh back button works
13:17:10JdGordonup/down does but not very often
13:17:44JdGordonhmm, not consistantly though :)
13:22:23 Join utanapischti [0] (
13:25:43kugelJdGordon: like it yea?
13:26:19 Quit sasquatch (Ping timeout: 272 seconds)
13:27:17 Join dfkt [0] (dfkt@unaffiliated/dfkt)
13:27:29 Quit bmbl (Quit: Bye!)
13:38:32 Join Staphylo [0] (
13:39:29CIA-81New commit by 03bertrik (r28170): RDA5802 tuner: fix small bug in rda5802_init (writing too much data)
13:39:53 Quit GeekShadow (Ping timeout: 252 seconds)
13:40:30 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
13:41:25CIA-81r28170 build result: All green
13:43:37CIA-81New commit by 03bluebrother (r28171): Improve some trace messages.
13:45:30CIA-81r28171 build result: All green
13:47:27 Quit teru (Ping timeout: 276 seconds)
13:48:47 Join teru [0] (
13:49:26amiconnTheSeven: Hmm, where do we have CPU frequency changes from IRQ context? This can't be in general code, as then it would fail on coldfire
13:51:26***Saving seen data "./dancer.seen"
13:53:58 Nick bzed_ is now known as bzed (
13:56:45bertrikhmmm, now the WPS corruption problems on my clip+ seems to correlate with having tagcache enabled or not
13:57:14 Join Lear [0] (chatzilla@rockbox/developer/lear)
13:57:20 Join saratoga [0] (9803c22e@gateway/web/freenode/ip.
13:57:22JdGordonwhat sort of corruption?
13:59:01CIA-81New commit by 03learman (r28172): Make sure get_lif_token_value isn't inlined, as it would defeat the purpose of introducing that function. E.g., gcc 3.4.6 for m68k is keen on inlining ...
13:59:02bertrikthe outer part of the progress bar is not showing, only the filling part
13:59:08TheSevenamiconn: I don't know for sure where it came from, but something apparently called set_cpu_frequency from an IRQ handler
13:59:33TheSevenat least I ran into massive trouble when trying to access I2C (which uses IRQs itself) from there
13:59:35bertrikwhen it does show correctly, it seems as if the progress bar is written a few times, first without the outer part, then with the outer part
14:00:18bertrikthe s5l8700 i2c drive rno longer works on the meizus
14:00:39CIA-81r28172 build result: All green
14:00:47amiconnTheSeven: Well, on coldfirecpu_set_frequency can take up to 10ms (2ms typical) due to pll relock
14:01:02TheSevenbertrik: interesting... IIRC I reverted that to almost what you were doing before after the IRQ-based one caused trouble on the nanos
14:01:09amiconnHence it must not be called from interrupt context
14:01:10 Join _s1gma [0] (~d.d.derp@
14:01:18JdGordonbertrik: that's not what I'd call corruption, but ok
14:01:25TheSevenamiconn: wouldn't IRQs be locked out during a PLL relock anyway?
14:01:33amiconnNo, why?
14:01:51amiconnThe cpu is running at base clock during that time
14:02:06TheSevenon the nano2g it might be running at like 32khz during that
14:02:09amiconnLocking out interrupts for such a long time would be bad
14:02:24amiconnHmm, that may b e more problematic
14:02:32amiconnColdfire base clock is 11.2896 MHz
14:05:45CIA-81New commit by 03gevaerts (r28173): Add optional (define BUFFER_ALLOC_DEBUG to enable it) code to check for code overflowing buffer_alloc()-allocated buffers. ...
14:06:27 Quit MethoS- (Remote host closed the connection)
14:07:31saratogagevaerts: you should accidentally enable that in an SVN build for a week or so and see how many people report panics ;)
14:08:00CIA-81r28173 build result: All green
14:08:02kugelthe panic is enabled as I see it
14:09:03 Quit ender` (Quit: Little expense had been spared to create the impression that no expense had been spared.)
14:09:04saratogaonly if BUFFER_ALLOC_DEBUG is enabled?
14:09:17 Quit [sko] (Quit: Leaving.)
14:09:20gevaertsThe panic that checks if there's enough room is always enabled
14:09:31n1sthat's good imo
14:09:43JdGordongevaerts: can you play with macros and get the calling function/line into the buffer header? I'll whip up a nice debug screen if you do :)
14:09:50gevaertsThe extra metadata, buffer guards, and code that walks the buffers to check for overflow at every thread switch isn't
14:10:16gevaertsJdGordon: That would be nice, but it's tricky...
14:10:27gevaertsAs in, I have no idea how to do it :)
14:10:32JdGordonit really doesnt need to check every thread switch, really only when coming off the main thread
14:11:29gevaertsMaybe, but even then I think it's too expensive for a regular build
14:12:00 Join [sko] [0] (~sko]
14:12:46gevaertsThe way I see it, this is something you enable whenever you're hunting a bug with interesting effects
14:13:18*JdGordon would also like to use host malloc instead of buffer_alloc() for app builds to not need to reboot to enable stuff
14:13:43gevaertsYes, that would make sense
14:13:58gevaertsAnd easy to do, actually
14:14:07JdGordonespecially when it is pretty much impossible to reboot the android app currently
14:14:12gevaertsJust call malloc() from buffer_alloc()
14:14:33JdGordonwell yes, but that doesnt make the settings magically start working :)
14:14:41kugelgevaerts: and who fixes the calling code that expects the need to reboot?
14:15:01gevaertskugel: yes, but JdGordon didn't ask for that :)
14:15:35JdGordonAlso, it would be nice if we came up with a good way to do host integration for things that arnt possible on DAPs... i.e logging and AA downloading
14:17:04gevaertsWhat is the block starting at dircache.c line 534 trying to do?
14:17:25*JdGordon likes guessing games :)
14:18:36TheSevenbertrik: do you have any idea what could have made I2C fail on meizu?
14:18:45gevaertsoh, I think I see...
14:19:12TheSevenas I said, I reverted most of my changes (and used the polling version again) after we had some trouble with I2C interrupts getting lost on the nano2g
14:19:15JdGordongevaerts: I think those targets actually save the dircache cache as a file so it doesnt need to rescan
14:19:18bertrikTheSeven, no, I just reverted to an earlier version to make it work again
14:19:38JdGordonso that makes sure the file actually has a chance of being usable because the pointer is the same as last boot?
14:19:43gevaertsJdGordon: yes, and this code checks if it would be loaded at the same address
14:19:54bertrikthe meizus are not even close to normal rockbox use, so don't worry to much about it
14:20:11 Join ender` [0] (
14:20:21TheSevennevertheless I'd like to know *why* it breaks, as this doesn't seem to work 100% right on nano2g either
14:20:23 Join wodz [0] (
14:20:43JdGordonby the way, IIRC pondlife made a patch ages ago to panic if anything called buffer_alloc after audio was initialsed, might be worth looking at with this fiddling
14:20:43gevaertsJdGordon: that will break (as in, throw away the file every time) with BUFFER_ALLOC_DEBUG defined
14:20:58gevaertsUnless I add a 0-size check
14:21:12JdGordonprobably not a big deal
14:21:14wodzamiconn: ping
14:21:46TheSevenbertrik: if I use the interrupt-based driver, it locks up when doing things like changing volume, while the interrupt-based driver I use in emBIOS seems to work just fine, even though it isn't used heavily
14:22:25TheSeveninterestingly I can't remember seeing it lock up when switching the backlight on or off, so it might be somehow related to the codec which I'm not using at all in emBIOS
14:22:48gevaertsmaybe not. On the other hand, adding this check (only in the debug case anyway...) isn't much work either...
14:23:35bertrikTheSeven, maybe we're really losing interrupts somewhere, or incorrectly clearing them
14:23:45 Join s1gma_ [0] (~d.d.derp@
14:24:33TheSevenit happened reproducibly when changing volume, but apparently the backlight was working fine
14:24:49JdGordonOK, I'm looking at doing host malloc... is there any point doing the eisting buffer alloc if malloc() returns NULL?
14:25:49 Quit _s1gma (Ping timeout: 240 seconds)
14:25:50 Join fdinel [0] (
14:26:50wodzamiconn: greylib uses 129 shades - does it means that the input value should be within the bounds or 0-255 is rescaled internally?
14:27:05amiconnThe latter
14:27:32wodzand 0 is white right?
14:27:41amiconnInput is plain standard 8 bit greyscale data.
14:27:44CIA-81New commit by 03gevaerts (r28174): If BUFFER_ALLOC_DEBUG is defined, make buffer_alloc() not actually allocate anything if size==0, so code that uses buffer_alloc(0) to find out what ...
14:28:20amiconnThe greylib applies gamma adjustment, lcd linearity adjustment, and rescales to 0..128
14:29:11CIA-81r28174 build result: All green
14:29:19amiconnIt also handles necessary white<->black flipping, even dynamically on iPod G1/G2
14:29:28 Quit liar (Ping timeout: 264 seconds)
14:29:30JdGordonhaha! simply putting the malloc() call into buffer_alloc() made the sdl app segfault!
14:29:39JdGordonwe may just have a problem here!
14:29:59amiconnThe iPod G1/G2 LCD is inverted when backlight is on
14:30:15gevaertskugel: the easy way to "fix" that would be to just always allocate at boot (without checking the setting), and hope that the OS we use does lazy allocation
14:30:30gevaertsJdGordon: any backtrace?
14:30:33JdGordonis 0xfffffffff08f1010 a probable usable retval from malloc?
14:30:38JdGordon64bit os obviously
14:31:31kugelgevaerts: I'm not convinced that's an easy way :)
14:31:42wodzamiconn: can I assume that background in greylib is 255?
14:31:58gevaertsJdGordon: I get very different pointers in a trivial test program
14:32:06amiconnwodz: Assume in what context?
14:32:19gevaertsJdGordon: what OS is this?
14:32:43JdGordonlinux, and yeah something is not right (This might actually be a real skin bug!)
14:32:55wodzamiconn: I want to support transparency in png to some degree so I have to know what is background
14:33:00JdGordonmalloc is returning sane values but skin_alloc_element() is breaking
14:33:54gevaertsSomething storing pointers in a type that can't handle them?
14:34:06JdGordon1320960, 0xfffffffff08f1010 alloced that looks odd... all the other allocs are in 0x122d4c0 area
14:34:31amiconnDefault background (after grey_init() ) is white (255). It can be changed and queried as with any other implementation of the rockbox gfx api
14:34:45 Join _s1gma [0] (~d.d.derp@
14:34:46amiconngrey_set_background() / grey_get_background()
14:35:07JdGordon1080 retval->type = UNKNOWN;
14:35:07 Join notlistening [0] (
14:35:08JdGordon(gdb) p retval
14:35:08JdGordon$1 = (struct skin_element *) 0xfffffffff08f1010
14:35:39 Quit s1gma_ (Ping timeout: 255 seconds)
14:36:26gevaertsJdGordon: no segfault here
14:36:38amiconnOf course this won't help if there's a backdrop, but fortunately (in this case) the greylib doesn't support backdrops
14:36:52notlisteningkugel do you want bugs reporting on the andoid app..?
14:36:55JdGordongevaerts: 64bit os?
14:37:03gevaertsJdGordon: yes, but a sim, not the app
14:37:07kugelnotlistening: no
14:37:16notlisteningok ;)
14:37:36kugelyou can mention them here if you want, if you're lucky someone will do something about it, but not on flyspray please
14:37:48JdGordonunsigned char* should be safe no?
14:38:29JdGordonbuffer_front = (void *)(((unsigned long)buffer_front + 3) & ~3); <- not entirely safe?
14:39:17gevaertsintptr_t sounds better
14:39:30amiconn'long' isn't guaranteed to be large enough to hold a pointer
14:39:51amiconnOn linux x86_64 it should work though, but it will break on win64
14:39:53JdGordonI dot think that is the problem though, that would cause the crash the 2nd time not first time, I'll fix it anyway
14:39:56bertrikis the cast to unsigned long (or anything else) really needed anyway?
14:40:40amiconnEver tried logical operations on pointers? The compiler doesn't allow this
14:40:44JdGordonyeah, that wasnt it
14:40:50notlisteningkugel I will try and get some useful information that will help out more than just it stops playing :D
14:43:10bertrikamiconn, ah ok, I was confused because it seemed as if the cast was needed to do +3
14:43:19 Join user890104 [0] (
14:43:22kugelJdGordon: do you call malloc for each skin element or do you obtain the buffer with a single big malloc?
14:43:32JdGordonsingle buffer still
14:43:49JdGordon(that will eventually be changed also, but one thing at a time)
14:44:15amiconnWell, you could cast outside the inner ( ), but if you do this with anything else than [unsigned] char* you will confuse people, probably even yourself
14:44:31kugelsingle buffer is probably easier (also we don't need to drift away too much from dap-rockbox), unless you want to mess with freeing all elements
14:44:59bertrikamiconn, no, that would not confuse me at least :)
14:45:00amiconn(also void* when using gcc - other compilers don't allow pointer arithmetics on void*)
14:45:05JdGordonthat code is there already, but yes, its low on the priorities
14:47:28 Quit _s1gma (Ping timeout: 264 seconds)
14:47:43 Join _s1gma [0] (~d.d.derp@
14:48:10n1sdon't we have a macro for aligning pointers?
14:48:22gevaertsJdGordon: did you include stdlib.h?
14:48:55gevaertsIf not, C assumes that a function returns int, which messes with the pointer that malloc() returns
14:49:17JdGordongood call :)
14:49:42JdGordon1320960, 0x7ffff08f1010 alloced <- still a wierd looking region, but ok :)
14:50:32kugelJdGordon: btw, I think you should fix up size, not the pointer
14:50:45gevaertsYes, I was going to suggest that too
14:50:48bertrikyeah, I wondered about that too in buffer.c
14:51:15gevaertsIf you round the size up, the pointer will stay aligned
14:51:25gevaertsAnd size is never that big, so you're fine by definition :)
14:51:36JdGordonnot my code :)
14:51:47gevaertsI actually changed that in buffer.c
14:52:10kugelJdGordon: really? I seem to remember the same lines in the old skin engine :)
14:52:17kugelbefore bieber :)
14:52:24JdGordonthat line copy+paste from buffer.c
14:52:37gevaertsbuffer.c doesn't do that!
14:52:46JdGordonit did when I copied it :p
14:52:52gevaerts(since a half an hour or so anyway)
14:52:52JdGordonI'll change that now
14:56:05CIA-81New commit by 03jdgordon (r28175): fiddle with the alloc requested size instead of the buffer pointer to keep the buffer 32bit aligned
14:56:41teru(FS #6321) I want opinion about "exclude entries from viewers.config for the targets which does not support that file type". currently, it is png for non-color targets.
14:57:07 Quit notlistening (Remote host closed the connection)
14:57:27gevaertsJdGordon: you'll have more fun. Try to find out why dircache.c messes with audiobuf directly...
14:57:34wodzteru: png support is on the way
14:57:49CIA-81r28175 build result: All green
14:57:53n1steru: sounds like a good idea
14:57:54JdGordongevaerts: oh SHIT! you serious? :'(
14:58:04gevaertsJdGordon: that's what it looks like anyway
14:58:12bertrikteru, png can also be greyscale, right?
14:58:16n1ssurely there are other filetypes not supported on all targets
14:58:45gevaertsrolo too, but I don't think that matters much here
14:58:56AlexPGive wodz a few minutes and png will be on greyscale too :)
14:59:14JdGordonwe can disable dircache on hosted though :)
14:59:14teruwodz: yeah, nice work. but same situation could happen in the future. for example when someone add gif viewer only for color target.
14:59:24gevaertsah, good call :)
14:59:30gevaertsWe probably should, anyway
14:59:37gevaertsIf it's not already
14:59:37kugelAlexP: actually, I think he managed greyscale already and is now adding greylib support (i.e. even nicer :) )
14:59:51kugelit is disabled
15:00:48AlexPkugel: Yeah, it already works
15:00:57gevaertsJdGordon: ok, dircache is fine. Now fix tagcache.c which does that too :)
15:00:59AlexPI was using it last night :)
15:01:15kugelthe host does caching, and it's not reasonable to cache the entire filesystem on a real machine
15:01:43kugelignoring the general breakage with recursive symlinks :)
15:02:09JdGordonarg, tagcache is sort of needed though
15:02:31kugeltagcache works as far as I can tell
15:03:19gevaertskugel: yes, but you didn't rip out buffer_alloc() :)
15:03:31gevaertsJdGordon: looks like an easy fix there though
15:05:47n1sbtw, having the properties plugin in the viewers list makes no sense
15:07:27JdGordongevaerts: did you change dircache.c to not be naughty?
15:07:42gevaertsIf it's disabled anyway, who cares?
15:08:06JdGordonIdeally nothing should use audiobuf directly
15:08:31gevaertsI'm trying to find out why that tagcache audiobuf abuse is safe (I assume it is, there aren't enough crash reports for it not to be...)
15:08:55JdGordonbecause it happens early probably? /me hasnt had a look yet
15:09:17gevaertsThat's what I'm looking for. From what I can see, it can happen at any time...
15:09:58GiovaJdGordon: One of the latest build has fixed the bug #11626
15:10:17JdGordon(to both above messages :p )
15:11:45terun1s: I'm not sure how to implement the idea. in my patch, image file types are associated to imageviewer.rock and imageviewer is build for all bitmap targets. so all image file types are included to viewers.config and it causes the issue.
15:11:47Leargevaerts: Only during tagcache init, as far as I can see...
15:12:04n1steru: aha
15:12:11gevaertsLear: oh, right
15:12:28JdGordonyeah 4417 seems scary, but probably safe? untill we get preemtive scheduling
15:12:29*gevaerts missed the #ifdef __PCTOOL__
15:13:23n1sbut tagcache.c should call buffer_alloc and not do this directly anyway, right?
15:13:24gevaertsI assume commit() does file IO, which could yield as far as I know
15:13:43kugelteru: just preproces viewers.config?
15:14:11gevaertsoh, and it sleeps, which definitely yields
15:14:48*TheSeven knows why he doesn't like cooperative multitasking
15:14:49gevaertsSo yes, that bit could well be responsible for weird hard to reproduce issies
15:14:55Leargevaerts: But the main thread waits for the initialized flag or something to be set (hence the "in foreground" comment).
15:15:07gevaertsLear: but do other threads?
15:15:21gevaertsWell, other threads shouldn't be buffer_alloc()ing anyway I guess
15:16:27JdGordonthe buffer isnt required after the commit so there must be a better way to do this
15:16:40JdGordonwhich would also mean not neeing a reboot to commit
15:17:12gevaertsIt could steal the plugin buffer :)
15:17:38JdGordonany ideas how much it needs?
15:17:40*gevaerts finds dircache_steal_buffer()...
15:18:23gevaertsIt grabs the entire audio buffer basically
15:18:39kugelJdGordon: it needs exactly 32*1024*1024
15:18:53gevaertskugel: only if __PCTOOL__ is defined :)
15:18:56kugeloh, that's the pctool case
15:19:06gevaertsSo it's "safe" in the sense that no other allocations can mess up its idea of the world
15:19:17kugelwell, I stronly suspect it won't need more on the dap
15:19:17n1s32MB would be kind of hard on most targets ;)
15:19:44kugelerr, that's 32*M*B, oops (/me thought K)
15:21:36*JdGordon notes there is no locking on the plugin buffer also
15:23:05n1sthat's why it's so much fun hunting bugs, you usually dig up a whole bunch of bad things...
15:23:06kugelthat tempbuf code is strongly suspicious in any way
15:23:21kugelwhat if the audio buffer isn't big enough? it could overwrite codec or plugin buffer
15:24:05JdGordonI'm replacing it with plugin_get_buffer() and removing the dircache steal
15:25:02kugelhopefully it won't clash with skin engine which also uses the plugin buffer?
15:25:18gevaertsJdGordon: that may well be too small...
15:25:30*gevaerts doesn't like that code either
15:25:53JdGordonkugel: Lear said the main thread waits for init so it should be safe, but yes a proper fix is also needed
15:25:55*TheSeven suggests just throwing everything away and rewriting it from scratch
15:26:17JdGordongevaerts: I'm working on the assumption that it doesnt care how much it has, but the more the merrier
15:26:17kugelI see calls to the strange code from the tagcache thread so it's not only at init
15:26:52 Join anewuser [0] (anewuser@unaffiliated/anewuser)
15:27:19 Quit BlakeJohnson86 (Read error: Connection reset by peer)
15:27:40JdGordonwhich strange code?
15:27:44n1sJdGordon: the size is checked in tagcache.c:2435 and it bails out if there is too little buffer left
15:27:48 Join BlakeJohnson86 [0] (
15:28:43kugelJdGordon: the tempbuf allocation one
15:28:49n1sthe build_index comment seems to be lying, returning 0 or < 0 has the same result
15:29:11n1sah, no it doesn't i can't read
15:31:25n1shmm, or well, the only difference is that the commit_delayed field is set which is only checked in one place and results in a "reboot please" splash and nothing else afaics
15:32:17gevaertsCan't we make the bootloader commit the database? That would fix everything!
15:32:18*gevaerts hides
15:33:08 Join domonoky1 [0] (
15:33:08 Quit Horscht (Ping timeout: 240 seconds)
15:33:45 Quit domonoky (Ping timeout: 252 seconds)
15:34:54n1sfor some things it would be nice to be able to alloc a largeish buffer for a short time and then free it again, we need a proper malloc!
15:35:07n1swith special magic of course
15:35:50 Join FlynDice [0] (~FlynDice@
15:36:02gevaertsWe could have a buffer_borrow() function with a mutex so buffer_alloc() will block until buffer_return() is called
15:36:03n1sso it would primarily evict the already consumed data in the audio buffer and if that wasn't enough the data furthest in the future and else, fail
15:36:33kugelgevaerts: that's not a bad idea
15:37:20gevaertsIt would handle this case anyway
15:38:32JdGordonand using that for the actual audio buffer would make things a bit safe also
15:39:27 Quit n1s (Quit: Lmnar)
15:39:45gevaertsOr for USB
15:40:32*TheSeven just found something in the nano2g that looks suspiciously like memory2memory dma
15:40:35gevaertsAlthough that would need more work I guess
15:40:42 Join Horscht [0] (
15:40:42 Quit Horscht (Changing host)
15:40:42 Join Horscht [0] (~Horscht@xbmc/user/horscht)
15:41:20 Quit GeekShadow (Ping timeout: 240 seconds)
15:42:35TheSevennano2g OF*
15:42:45gevaertsBasically unify audio_get_buffer() and buffer_alloc() a bit
15:44:09JdGordonand to overengineer it, allow it to be non-blocking (so fail if it cant get the buffer) and requested_size so more than one thing can borrow a block
15:44:37gevaertsWouldn't you get fragmentation then?
15:44:47TheSevenJdGordon: What about just implementing a proper malloc?
15:44:58JdGordonwell that would be the obvious solution :p
15:45:07gevaertsHow? No MMU, remember...
15:45:22gevaertsi.e. you can implement malloc(), but it won't work well
15:45:32TheSeventhe aim should of course be to make sure that nothing stays allocated for too long
15:46:07JdGordonthe requested_sie thing was me being silly, but not blocking would be useful
15:46:26TheSeventhe audio buffer should just float around in the empty space
15:46:51Jaykayrockbox does work with sdhc cards of all sizes, right? (i plan to buy a 8gb one)
15:47:00gevaertsJdGordon: for cases like doing whatever you need to do, or splash for audio stop or reboot, you mean?
15:47:29JdGordonor trying other potential buffers like the plugin buffer if audio one isnt available
15:47:39JdGordonwhich would be unlikely if this isnt abused
15:48:00TheSevenactually I'd like to get rid of the plugin buffer as well :D
15:48:06JdGordonJaykay: yes
15:48:12Jaykayok, thanks :)
15:48:44gevaertsI wouldn't mind seeing the buflib system that was once proposed, and position-independent plugins
15:48:50TheSevenor we could at least use the plugin buffer for audio cache as long as no plugin is running
15:49:08gevaertsThat doesn't mean we can't improve things a bit now though :)
15:49:33*FlynDice finishes 30 minutes furiously searching for prebuilt android .apk availability before deciding to punt and request handy link from someone smarter here....
15:50:09kugel :)
15:51:01FlynDicekugel: Thanks, of course I coulda rolled my own by this time...
15:51:28***Saving seen data "./dancer.seen"
15:51:43bertrikDoes anyone here have a nano 1g and an ipod fm remote?
15:52:28 Quit BlakeJohnson86 (Read error: Connection reset by peer)
15:58:15Jaykaywiki SansaE200Port : "Flash driver DONE - This can be substantially improved by using DMA and sleeping/yielding appropriately" and "Not all plugins have been adapted to the Sansa's screen layout yet." Are these still valid?
16:00:06FlynDicekugel: Is the AndroidPort hidden on the wiki main page on purpose or has it just not been added yet?
16:00:22JdGordonwell replacing those nasty allocs in tagcache with plugin_get_buffer works
16:01:24gevaertsJdGordon: with big databases?
16:01:36JdGordon3000+ files, e200 sim
16:01:57JdGordonif it works on target is another story
16:02:23gevaertsI doubt if it will work with 50000+ files
16:03:11kugelFlynDice: no special reason
16:03:56FlynDiceOk, I'll try to add it.
16:04:43gevaertshm, does the database actually work with large amounts of files on Archoses, or the 2MB clips?
16:05:14JdGordongevaerts: in that case it would be nice if someone reworked it so work with smaller buffers\
16:06:26JdGordonI'm not entirely sure buffer_borrow() would solve tagcaches nasty allocing anyway
16:09:35gevaertsCould someone who has a Clip (and can find it...) try putting lots of files on it (e.g. the HVSC), and see if it can initialise the dayabase with that?
16:09:40 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
16:11:05JdGordoncould someone make a legal mp3/whatever with no audio and a script to copy it 100000 times?
16:11:23gevaertsJdGordon: just use the HVSC :)
16:14:57 Quit fdinel (Read error: Connection reset by peer)
16:16:45 Quit GeekShadow (Ping timeout: 240 seconds)
16:22:23kugelFlynDice: write just AndroidPort, the wiki software recognizes it as a wiki word and makes the link
16:26:03FlynDiceSeems I always have to make things way too complicated....
16:29:04 Quit PaulJam (Ping timeout: 255 seconds)
16:31:13 Quit Lear (Quit: ChatZilla 0.9.86 [Firefox 3.6.10/20100914125854])
16:37:20 Join Evilnick_ [0] (
16:38:53 Join Evilnick__ [0] (
16:40:30 Quit evilnick (Ping timeout: 240 seconds)
16:41:51 Quit Evilnick_ (Ping timeout: 252 seconds)
16:52:33 Join BlakeJohnson86 [0] (
16:53:15 Join Nausicaa [0] (
16:55:29 Quit tchan (Quit: WeeChat 0.3.3-dev)
16:55:46wodzehh simple transparency mode of PNG is a pain in 16bit depth
16:56:26 Join clone4crw [0] (
16:56:56 Quit clone4crw (Remote host closed the connection)
17:01:37 Quit teru (Quit: Quit)
17:01:46 Join clone4crw [0] (
17:06:56 Quit [sko] (Quit: Leaving.)
17:07:33 Quit Jerom (Read error: Connection reset by peer)
17:08:20 Join Jerom [0] (~jerome@
17:10:46 Quit JdGordon (Ping timeout: 276 seconds)
17:28:08 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
17:30:00 Join mirak_ [0] (
17:40:25 Quit edboyer93 ()
17:41:40 Join webguest45 [0] (
17:41:50 Quit webguest45 (Client Quit)
17:42:23wodzhow fancy should be our png decoder? I think we should support all color modes (including alpha channel). But what to do with 1)simple transparency 2) background chunk? This two adds much complexity to decoder
17:44:38pixelmawodz: did you try at least try compiling of the imageviewer (with the greylib png support) for sh with the small plugin buffer?
17:51:30***Saving seen data "./dancer.seen"
17:52:27wodzpixelma - it compiles for ondioFM 2MB
17:53:26 Quit mirak_ (Read error: Connection reset by peer)
17:54:41wodzsmall png files works ok in ondioFM sim (2MB)
17:54:43 Join mirak_ [0] (
17:58:09 Join jimbo [0] (
17:59:56jimbohey guys, had the error message about the arm-elf-eabi-gcc command not found, ran to download it
18:00:04jimbohow long will that take to complete?
18:00:25gevaertsjimbo: that depends on your system
18:00:46jimborunning vmware
18:01:14jimbobut my system is a dual core 2ghz, with 4gb ram
18:01:24gevaertsIt takes about 15 minutes on my laptop running linux, up to ten hours or so on cygwin on older systems
18:01:27jimboi'm seeing a ton of text
18:01:34jimboits scrolling too fast to read any of it
18:01:41gevaertsThat's good :)
18:01:50gevaertsI'd say half an hour or so
18:02:08 Join s1gma_ [0] (~d.d.derp@
18:05:15 Quit _s1gma (Ping timeout: 240 seconds)
18:05:37bluebrotherjimbo: doesn't download the compiler. It download its source and compiles it, which is why it takes a notable time to finish.
18:11:00jimboi'm trying to compile now, it says that this client is too old to work with working copy
18:11:19jimbopleae download new subversion or something
18:11:26 Quit s1gma_ (Quit: Leaving)
18:11:50jimbonow it seems to be compiling it
18:12:04jimbogetting lots of warnings, etc
18:12:16bluebrotherhow did you check out the Rockbox source? This indicates that you checked out with a different (i.e. newer) client than you have running right now
18:12:34jimboi used svn to download it this morning
18:12:55jimboran a patch on it, which seemed to work
18:13:08bluebrotherwell, but how? If you see this message then you definitely used a different version of svn (or a different client) when checking out
18:13:13 Join _s1gma [0] (~d.d.derp@
18:13:31bluebrotherpatching the sources won't make a difference here.
18:14:15jimbowell when I got that error message I ran rockboxdev and entered just "e" when asked what flag to use
18:14:23*amiconn suspects that *writing* to 0x0.... area on PP actually writes to IRAM as well
18:14:30jimboit did its thing, and now i'm building
18:14:39*amiconn will test this theory soon doesn't use svn at all
18:15:06jimbothen i'm not sure whats wrong :S
18:15:18jimboits still doing smoething though
18:16:10jimboi'll let it continue and see where i'm at
18:16:21amiconnHmm, or maybe not
18:20:32jimboyep got the subversion error again so i'm guessing it didn't work
18:20:40jimboshould i get teh source again and start from scratch?
18:24:53jimbohow did I get an older svn then?
18:25:01jimbois there a way to upgrade it?
18:25:04TheSevensvn up -r whatever
18:30:01pixelmawodz: I don't think sim tells much there, I think I'm going to try. What was the FS# again?
18:32:00wodzFS #11641 - try v1 patch
18:32:09jimboinside the rockbox directory, typed in svn up, it said this client is too old...
18:32:31jimbothis client is too old to work with working copy '.'
18:32:43wodzsim tells something - bigger images don't load with error of out of memory from decoder
18:32:55TheSevenso you're using a working copy that has already been touched by a newer svn client with an older client
18:33:06TheSeventhere was an incompatible working copy format change some time ago
18:33:34TheSeveneither re-checkout it completely and don't touch it with newer SVN clients any more, or update *all* of your svn clients that might touch it
18:34:30pixelmawodz: was there something to pay attention with svn?
18:36:18wodzpixelma: ah yes v1 patch do not apply cleanly to svn checkout ($id$ issue)
18:40:19pixelmaif it was only one failed hunk to correct manually... :\
18:41:29gevaertspixelma: if you open the files that fail (the originals, not the patch), and change the "$Id:.....$" bits to "$Id$", the patch applies
18:41:43 Quit FlynDice (Remote host closed the connection)
18:42:42pixelmawell, it's the same as applying the things in the .rej
18:46:05 Quit CGL (Remote host closed the connection)
18:48:08 Quit Giova (Quit: Sto andando via)
18:48:27pixelmaor not :( no. fun.
18:51:34 Join m0ar [0] (
18:52:39m0arMy clip+ has been acting strange the last couple of days, instead of skipping to the next track when pressing >>| it just skips 0.001 sec forward in the song. I have tried to locate the setting, but had no luck. What might I have touched?
18:53:48kugelhow did you measure the 0.001s?
18:54:02m0arMy mind os powerful.
18:54:16m0arkugel: Just estimating, it's prolly like 0.2 or sumting
18:54:41kugeluse real words please
18:55:08m0arkugel: Oh well, If I keep it pressed down it searches the track forward as usuam
18:55:33m0arkugel: but a swift click does not change the track, it just skips an extremely small bit forward
18:55:36m0arGet it?
18:55:58kugelperhaps you have enabled "prevent track skipping"
18:56:55wodzpixelma: when I finish transparency handling I'll provide patch witch will apply cleanly to svn
18:56:59m0arAh damn, I even recall doing that when you mention it
18:57:00m0arkugel: ^
18:57:19m0arkugel: Never really understood what that setting did, but i feel so retarded right now.. Thanks mate
18:57:57kugelit causes some confusion, you're not the first one
18:58:56m0arkugel: Good ;)
18:59:18m0arIt's there to prevent accidental trackchanges for things like audiobooks i guess?
19:01:42 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
19:01:57m0arJust wondering, what is there to solve before the clip+ can be called stable?
19:04:17gevaertsPossibly USB, but also that it's not a good idea (as we noticed with the clip) to promote a target to stable when we're not doing a new release
19:05:25m0argevaerts: You mean filetransferring while connected via usb? What do you mean with the last thing?
19:07:00gevaertsm0ar: basically that declaring a target "stable" should be done at the same time that we release 3.7, or 3.8, or whatever. If it's done in between, various scripts and bits of the website get confused
19:07:45gevaertsThere are now e.g. various links to a 3.6 build for the Clip, which doesn't exist, because the Clip wasn't stable yet when 3.6 was released
19:08:36m0arAh, thank you
19:09:01gevaertsSo we either have to fix those, or just not declare targets stable at random times
19:12:11 Join sasquatch [0] (
19:15:31 Quit utanapischti (Ping timeout: 240 seconds)
19:18:37 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
19:24:03 Quit anewuser ()
19:24:07 Quit mirak_ (Read error: Connection reset by peer)
19:25:24 Join mirak_ [0] (
19:33:47 Join pamaury [0] (
19:33:47 Quit pamaury (Changing host)
19:33:47 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
19:33:58 Quit jimbo ()
19:42:16 Quit mirak_ (Read error: Connection reset by peer)
19:43:16 Join mirak_ [0] (
19:44:21 Quit Jerom (Read error: Connection reset by peer)
19:44:34 Join Jerom [0] (~jerome@
19:47:13 Join BoomerET [0] (
19:47:43BoomerETI have an iPod Nano 2nd Gen I'd like to install rockbox on... The installer said it was successful, but I still can't seem to use it...
19:48:16BoomerETI hold down the play button to turn it off, and the screen goes black, but if I press menu, it comes on immediatly, not at all like it's booting
19:48:38BoomerETI've never used a nano before, I'm used to my 5th Gen iPod
19:48:57BoomerETand yes, I have the manual open in front of me.
19:51:34***Saving seen data "./dancer.seen"
19:51:57 Join Jerom1 [0] (~jerome@
19:51:57 Quit Jerom (Read error: Connection reset by peer)
19:52:28BoomerETGot it working, had to install the bootloader (even though it told me it was already installed)
19:53:55TheSeveneven if the bootloader was installed before (but the apple firmware booted through it), the ipod won't reboot completely unless the firmware partition is touched again (by reinstalling the bootloader)
19:54:12TheSeventhose things don't power off completely when holding play
19:54:16 Join Evilnick_ [0] (
19:54:19BoomerETI see that now ;)
19:54:34TheSevenholding menu+select for 5 seconds will force it to reboot
19:55:17BoomerETI'm good to go, appreciate the help.
19:55:29BoomerETI can switch back and forth between both now.
19:56:01 Quit Evilnick__ (Ping timeout: 276 seconds)
20:07:19 Join Evilnick__ [0] (
20:09:41 Quit MagusG (Ping timeout: 264 seconds)
20:10:17 Quit Evilnick__ (Read error: Connection reset by peer)
20:11:01 Quit Evilnick_ (Ping timeout: 245 seconds)
20:17:02 Join stooo [0] (
20:17:16 Join evilnick [0] (
20:17:56 Quit stooo (Remote host closed the connection)
20:19:04 Join stooo [0] (
20:19:05 Quit stooo (Client Quit)
20:21:26 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
20:29:52 Quit froggyman (Quit: Bye)
20:30:13 Quit Staphylo (Quit: Bye les gens =))
20:36:40 Quit mirak_ (Read error: Connection reset by peer)
20:37:37 Join mirak_ [0] (
20:38:13 Quit elcan (Read error: Connection reset by peer)
20:38:58 Join elcan [0] (
20:40:24 Quit Judas_PhD (Quit: This is a quitting message)
20:40:51 Quit wodz (Quit: Leaving)
20:42:42 Join Benjamin_ [0] (
20:43:11 Quit elcan (Ping timeout: 240 seconds)
20:43:12Benjamin_Could anyone explain to me why once i've installed rock box 3.6 on ipod nano 1g
20:43:35Benjamin_why the /.rockbox/ folder doesnt show up?
20:44:06 Join elcan [0] (
20:44:11gevaertsIn the rockbox file browser you mean?
20:44:30gevaertsYou can change the "show files" setting to show hidden files too
20:44:50gevaertsMost people don't need to see .rockbox :)
20:44:54Benjamin_im on a mac (new convert) so how do you do it:)?
20:45:10Benjamin_haha i want to edit some stuff in it
20:45:29gevaertsWait, you mean you don't see it on the mac?
20:46:02Benjamin_yeah, ive got it mounted
20:46:13Benjamin_but it only shows...
20:46:17gevaertsIf so, you need to tell your mac to show hidden files, something with which anothe mac user might probably be able to help.
20:46:24Benjamin_calendar contacts etc..
20:46:24*gevaerts isn't a mac user
20:46:50Benjamin_right ill have a look, google can be that user to help me haha
20:47:28xnyhpsBenjamin_: You can press command-shift G when at the root of your device, and enter .rockbox to go there.
20:48:01Benjamin_xnyhps thats fab
20:48:05Benjamin_did the trick cheers guys:D
20:52:51 Quit Benjamin_ (Quit: Benjamin_)
20:52:56 Quit bmbl (Quit: Bye!)
20:58:29 Quit GeekShadow (Quit: The cake is a lie !)
21:02:09 Join anewuser [0] (anewuser@unaffiliated/anewuser)
21:05:31 Quit BoomerET ()
21:14:26 Join Evilnick_ [0] (
21:17:12 Join Judas_PhD [0] (
21:17:51 Quit evilnick (Ping timeout: 240 seconds)
21:19:45 Join TheSphinX^ [0] (
21:23:44 Join plerohero [0] (
21:23:48 Quit dys (Read error: Connection reset by peer)
21:24:08pleroheroI have a simple question...
21:24:44pleroherowhen a dynamic playlist (not saved) reaches its end it's deleted?
21:26:41pleroheroI usually play one track and then enqueue more tracks... but when the playlist reach its end its erased.
21:26:48clone4crwI think it's erased only when another one is created
21:27:17pleroheroits automatically deleted
21:27:57pleroheroI use file browser to add a directory
21:27:57clone4crwAnd you want to be kept?
21:28:16clone4crwon the main menu playlists > save current playlist
21:28:43pleroheroi dont want it to be physically saved
21:29:27pleroheroi just want to have a persistant playlist lets say
21:30:04pleroherojust like amarok or foobar playlist
21:30:13pleroheroyou add your songs to the playlist
21:30:25pleroheroit plays them in the order you want
21:30:42pleroheroif it ends its not destroyed
21:31:11pleroheronor you have to store it manuallly every time
21:31:46clone4crwoh i see what you mean.
21:31:46gevaertsIt doesn't look deleted to me...
21:32:00gevaertsplerohero: when it reaches the end, there's still Playlists->Current Playlist
21:32:59gevaertsok, that doesn't survive reboot
21:33:32clone4crwno, i think he's going to have to save it to disk.
21:34:37clone4crwSo, I created a playlist of some random songs, and hit stop, and it persists in Current Playlist and can resume it
21:35:05bertrikthis behaviour of rockbox confuses me a bit too
21:35:44gevaertsIt's in a very vague way related to not saving pitch and timestretch settings
21:36:11gevaertsI'd like not using the player for half an hour be the same as shutting it down and booting it again in half an hour
21:36:28clone4crwAnd that playlist is still there after I reboot. huh.
21:36:49gevaertsclone4crw: yes, it's a playlist that *ended* that behaves differently
21:37:17clone4crwOH. I have mine set to repeat all. That's why.
21:37:28gevaertsYes, it never ends then :)
21:38:25pleroheroi thought it was deleted
21:38:40clone4crwOK, so I set repeat to off, and I still see the playlist in Current Playlist
21:38:42pleroheroits in the Playlist->Current playlist
21:39:08gevaertsclone4crw: yes, until you reboot. (unless you just stopped it, which is different)
21:39:42gevaertsA playlist that didn't reach its end is *always* persisted
21:40:24clone4crwYeah, I see now. Hmmm... Think it should be an option to always persist unless a new one is started?
21:40:28 Join Horschti [0] (~Horscht@xbmc/user/horscht)
21:40:43gevaertsclone4crw: actually, I see no good reason to clear it
21:40:59gevaertsSo I don't see the need for an option. I just think current behaviour is wrong :)
21:41:11pleroherothanks for clearing this up! bye (i think it should be documented in the manual somehow)
21:41:42pleroheroif i want it stored ermanently i'll use repeat all
21:42:30pleroheroi think the behaviour of amarok or foobar playlist would be better.
21:43:07 Join [sko] [0] (~sko]
21:43:14*gevaerts has no idea how those behave
21:43:40clone4crwI don't use either, but MediaMonkey always persists unless you start a new playlist
21:43:54pleroherothat's their behaviour
21:44:11 Quit Nausicaa (Ping timeout: 265 seconds)
21:44:31gevaertsWhat does "always persists" mean *exactly*? Is the only difference that we don't keep it after reboot?
21:44:47 Quit Horscht (Ping timeout: 272 seconds)
21:44:51clone4crwPretty much
21:44:58pleroherothat's one point
21:45:01pleroheroanother on
21:45:06pleroherois that
21:45:12pleroheroone is that
21:45:22pleroheroif my playlists reaches the end
21:45:22 Join slothearn [0] (
21:45:28pleroheroand i click play
21:45:31AlexPplease don't use the enter key as punctuation
21:45:33*TheSeven wonders if plerohero's keyboard has a spacebar
21:46:25AlexPgevaerts: I agree - I've never understood why the end of a playlist means it disappears
21:46:42pleroheroif my playlists reaches the end and i click play it should not respond "nothing to resume"
21:46:43gevaertsAlexP: well, technically it doesn't. It just doesn't get saved :)
21:47:06AlexPgevaerts: Ignore what I say and just listen to what I mean )
21:47:08bertrikplerohero, you mean you expected it to be stopped and rewound?
21:47:13pleroheroI'm not a native english speaker so ignore my mistakes please..
21:47:37AlexPbertrik: It should restart IMO
21:47:46*gevaerts disagrees
21:48:02AlexPHow do I play it again if I want to?
21:48:20slothearnit should stop at the last track if repeat isn't enabled.
21:48:26slothearnbut replay the last track if play is pressed
21:48:28plerohero@alexp you go to playlists->view current playlist
21:48:44AlexPSeems a bit of a long way round
21:48:51gevaertsIf I listen up to the end of a playlist, and then I fall asleep, and the next day I restart the player, and the start screen is set to WPS, should it restart the playlist? If so, why do need repeat modes?
21:49:17AlexPTo restart automatically
21:49:19pleroherobeacuase i'm lazy :-p
21:49:38AlexP(why there are repeat modes that is)
21:49:38gevaertsSo people who *don't* want to listen to the same thing again aren't allowed to be lazy?
21:49:43pleroherobecause i'm lazy :-p
21:49:59AlexPgevaerts: How does it impact you if you don't want to listen again?
21:50:26gevaertsAlexP: I have to stop playback and leave the WPS first
21:50:41AlexPOK, I see that if you have it set to WPS on start
21:50:52gevaertsWhich is about as many keypresses as you need to go to current playlist and start again :)
21:51:05gevaerts(provided it gets saved)
21:51:21AlexPI don't actually mind if play restarts it or not, it just seems that the method of restarting is a bit baroque
21:51:32AlexPAnd yes, the real issue is the not saving it
21:51:35***Saving seen data "./dancer.seen"
21:51:40slothearnhere's a question, why do I sometimes have to use multiple keypresses on an iPod 5G?
21:51:55slothearndoesn't always work the first time. I've noticed this :O
21:52:31slothearnit's also very intolerant of accidental presses. like the menu/select reset combo
21:52:44slothearnthat sometimes needs to be like 2-3 times to get it to work.
21:52:59slothearnI'm assuming because my fingers slide ever so slightly and register as scrolls.
21:53:08gevaertsslothearn: I know there has been an issue with the first keypress after leaving a plugin, but I think that's fixed
21:53:20gevaertsApart from that, I'm not aware of any issues
21:53:29bertrikgevaerts, no, it's not fixed, but we do know what causes it
21:53:30slothearngevaerts: I thought it was just my iPod tbh, but it works fine in the official firmware
21:54:01domonoky1slothearn: the reset combo isnt controlled by rockbox, its a hardware feature..
21:54:05AlexPgevaerts: Would it be difficult to save the dynamic playlist? I'm assuming not very
21:54:17slothearndomonoky1: that's quite odd then :O
21:54:19 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
21:54:30gevaertsAlexP: I'd estimate three days to understand the code, and then change one line :)
21:54:43AlexPheh :)
21:55:06gevaertsI mean, it *is* saved normally
21:56:23AlexPyeah, just not when finished
21:57:03plerohero@alexp it's saved but it is moved to playlists->current playlist
21:57:37slothearnTwitter in IRC :-D fun
21:57:39gevaertsplerohero: it's *not* saved. It's gone after reboot
21:57:46pleroherothere are two behaviours hereoyt saving playlists after reboot and moving the current playlist when it reach its end
21:57:47AlexPplerohero: You don't need the @
21:58:28gevaertsEither that or the restore sees the end of playlist and then forgets the lot
21:58:34slothearn"al[tab]" message
21:58:52bertrikargh, so possibly someone at some point added extra code for the current behaviour :)
21:59:16gevaertsbertrik: possibly :)
21:59:28gevaertsBut I can't find out who it was unless I find where that code is :)
21:59:40*kugel also doesn't understand the people that prefer the most annoying splash & destroyed playlist over just restarting the playlist
22:00:03AlexPWell, I'm not too bothered about how to resume it, the main thing is saving a finished playlist over reboot
22:00:08gevaertskugel: I *do* understand that!
22:00:16gevaertsDon't you dare change *that* part!
22:00:44kugelI'm fine with repeat all so don't worry :)
22:00:51 Quit Evilnick_ (Ping timeout: 240 seconds)
22:01:04*gevaerts breathes again
22:01:50kugelbut I still think it doesnt make sense
22:02:40pleroherowhether behaviour changes or not I think it should be documented. I always thought that my current playlist was deleted....
22:03:09AlexPplerohero: Patches are welcome to the manual
22:09:37clone4crwgevaerts: by the way, I saw your feedback. I'm working on some new BMPs right now, and the manual. If I'm lucky I'll have both done by tonight
22:14:29 Quit plerohero (Quit: CGI:IRC)
22:18:08gevaertsclone4crw: great. I'll see that tomorrow then I guess
22:18:22*gevaerts finds the way the reload on boot is handled \☺/
22:25:06 Join crow [0] (
22:25:28crowHi, Whats status of Ipod Nano 2g (4gb model)? i see on wiki post from 2009.
22:28:12bertrikDoes loading a bookmark erase the current dynamic playlist?
22:28:30TheSevencrow: some minor issues left, but no blockers
22:28:42gevaertsbertrik: yes
22:28:54TheSevenready for everyday use, concerning battery runtime it's slightly worse than the apple firmware
22:30:18 Join Figa [0] (
22:32:06FigaHi I have problem with convbdf I have problem with this utility. I know that I am in rockbox room. I hope that it does not matter. Is there any who could help me please?
22:33:38gevaertsAlexP: I have it working :)
22:33:56AlexPgevaerts: Ace :)
22:34:08AlexPWas it one line? :P
22:34:23 Join krabador [0] (
22:34:47 Quit lestatar (Read error: Connection reset by peer)
22:35:03crowTheSeven so i can try 3.6 on it? first to flash bootloader?
22:35:27gevaertsTo just get the "view current playlist" bit to work, it was half a line
22:35:59gevaertsTo *also* get "save current playlist" to work, *without* doing "view current playlist" first, there's another line
22:36:11gevaertsI'm now seeing if I can now remove lines elsewhere :)
22:36:41AlexPhaha :)
22:37:29gevaertshm, maybe better not
22:37:38 Quit TheSeven (Ping timeout: 252 seconds)
22:38:25pixelmahur... flyspray notification mails seem to start working for me again. At least I got one now, hopefully it stays that way
22:39:18pixelmaworking I mean, not keep "staying" at one mail ;)
22:42:27 Quit BHSPitMonkey (Read error: Operation timed out)
22:42:58 Join lestatar [0] (
22:43:50gevaertsAlexP: FS #11644
22:44:29 Join nihilanth [0] (
22:45:46AlexPgevaerts: Coolio! I don't imagine it really needs testing?
22:46:11gevaertsAlexP: mainly review by people who understand this code I think
22:46:21AlexPyeah, that's what I thought :)
22:47:20 Part Figa
22:53:43crowi just finished flashing bootloader and rockbox.. but how to restart ipod nano 2g?
22:55:04crowah got it nice rock box is here :)
22:56:27 Quit Jaykay (Ping timeout: 252 seconds)
22:58:32pixelmahow likely is something overflowing being the cause of the screen corruption problem and could an "stkov main" when doing something in the WPS be related?
22:59:02gevaertssomething overflowing could basically cause anything
22:59:16pixelmathe latter happens on my M5 now from time to time while skipping tracks, leaving or entering the WPS, asjusting volume or such
22:59:58pixelmais there something I could test to help debugging this?
23:00:45 Quit nihilanth (Read error: Connection reset by peer)
23:01:55gevaertsprobably. Now just to think what... :)
23:04:11gevaertspixelma: the easiest first test would probably to try a larger stack
23:04:21pixelmamaybe "catch mem accesses" can give me some addresses but that's not guaranteed
23:05:14pixelmagevaerts: where and how? :)
23:07:30gevaertspixelma: for ondio, firmware/target/sh/archos/ Change the 0x2000 to 0x4000
23:07:45gevaertsFor m5 unfortunately IRAM seems to be nearly full, so it's not that easy
23:08:26pixelmathe "stkov main" is on my M5
23:08:46gevaertsyes, I know
23:08:49 Join Heis [0] (
23:09:02pixelmaon the Ondio I still have trouble with "my" WPS though
23:10:30gevaertsAnother thing you could try is adding -DBUFFER_ALLOC_DEBUG to EXTRA_DEFINES in the makefile, to see if something overflows a buffer_alloc()ed buffer
23:10:33 Join nihilanth [0] (~zach@tuxhacker/Nihilanth)
23:10:43gevaerts(with a recent checkout. I only added that today...)
23:10:55 Quit krabador (Ping timeout: 272 seconds)
23:10:59 Part Heis
23:11:51nihilanthMy ipod 5g is stuck in the rockbox bootloader with a checksum error, I already tried hard resetting and disk mode but neither seem to work
23:11:58nihilanthis there anything else i should try? or should i just wait until the battery dies
23:13:01AlexPreset is a hardware thing, just keep trying
23:13:27CIA-81New commit by 03bertrik (r28176): Warn about erasing dynamic playlist when loading bookmark - FS #10482 by Tuomas Airaksinen
23:14:22nihilanthit's only supposed to take ~5 seconds right? should I try buttons again every so often or just keep them held down?
23:14:48gevaertsnihilanth: did you switch hold on and off? Sometimes that makes a difference
23:15:11nihilanthgevaerts, yeah I've done that a few times
23:15:18CIA-81r28176 build result: All green
23:15:22nihilanthit's being real stubborn
23:15:39Tornehold it for 30 seconds or more
23:16:17pixelmagevaerts: ok, I'll try the first on my Ondio and the latter on the M5
23:16:34nihilanthalright, i'll keep at it. thanks for the encouragement :P
23:16:37Torneit resets eventually, if you hold it long enough, every case we've heard of :)
23:16:57gevaertspixelma: should give you a much bigger main stack on m5, at the cost of less IRAM for plugins/codecs (so there may be some performance problems with some codecs). I'd use that for a while, and check the stack usage in the debug menu every now and then
23:17:17gevaertsThe big thing will be to find out what's using all this stack though
23:18:08 Quit TheSphinX^ (Quit: XChat)
23:19:56 Quit feisar- (Ping timeout: 276 seconds)
23:20:14 Join MethoS- [0] (~clemens@
23:21:53 Join fdinel [0] (
23:24:02pixelmaI remember checking the stack usage when investigating the nested conditionals problem with Lear and he made me check the numbers on the Ondio and it didn't seem too much
23:24:49 Quit crow (Remote host closed the connection)
23:25:03gevaertsIt's too much onm5 though
23:25:41*gevaerts would like to see 3.7, but this sort of issue really is a release blocker...
23:26:01 Quit ender` (Quit: The latest survey shows that 3 out of 4 people make up 75% of the world's population.)
23:27:26 Quit [sko] (Read error: Connection reset by peer)
23:27:31 Join [AndrewR] [0] (
23:27:57[AndrewR]hello, I'm trying to write a plugin, and I have rockboxui compiling.. I added my plugin to SOURCES and CATEGORIES, but it's not showing up in the menu
23:28:52[AndrewR]anything else I need to do?
23:29:36gevaertsYou're trying this with a simulator build?
23:29:42gevaertsDid you run make install?
23:30:09[AndrewR]um at some point I think I told it to build simulator, yeah
23:30:22[AndrewR]if I rm rockboxui it rebuilds it with make
23:30:53[AndrewR]the simulator runs, my plugin's just not in the apps list
23:31:16gevaertsDoes your plugin get built?
23:31:25[AndrewR]it does, I see it compiling
23:31:28[AndrewR]to a .rock file
23:31:28pixelmagevaerts: I should really try an unaltered build from after yesterday's fixes first though, I guess - just remembered that I didn't update the M5 today...
23:31:45[AndrewR]I can try make install..
23:31:47gevaertspixelma: ah, yes. Might be useful :)
23:32:02pixelmajust the other two
23:32:27gevaerts[AndrewR]: the sim doesn't get any plugins if you don't do make install
23:32:41[AndrewR]ah, it shows up now, thanks!
23:33:07 Join feisar- [0] (
23:33:12[AndrewR]so I should be doing make install each time?
23:33:25gevaertsOr copy it by hand
23:34:44[AndrewR]ah, ok. thanks
23:37:32 Quit pamaury (Remote host closed the connection)
23:40:45pixelmagevaerts: I could still run with the debug option, you mentioned, for a while. What will I see if it catches some error?
23:40:56gevaertsa panic
23:41:08*gevaerts looks up the actual text
23:41:28 Part domonoky1
23:41:53gevaerts"%s corrupted buffer %x start" or "%s corrupted %x end", with %s being the thread name, and %d being the address of the buffer
23:42:17pixelmaI'll try that then
23:42:46 Join Loto [0] (~ctrlproxy@xbmc/user/Loto)
23:42:50 Part Loto
23:42:51 Join Loto [0] (~ctrlproxy@xbmc/user/Loto)
23:43:01gevaertsThat one might also slow down things a bit :)
23:43:15gevaertsNothing like debugging options to improve performance
23:48:13 Quit m0ar (Quit: leaving)
23:51:36***Saving seen data "./dancer.seen"
23:55:20 Quit dfkt (Quit: -= SysReset 2.53=- Sic gorgiamus allos subjectatos nunc.)
23:57:49 Quit kugel (Remote host closed the connection)

Previous day | Next day