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 2005-01-03

00:00:24[IDC]Dragonhaven't fixed anything yet
00:00:36[IDC]Dragondid you commit something?
00:01:13amiconnNo, I'm not that far yet. Just wading my way through fat.c
00:01:57amiconnBtw, how is the fat test code supposed to be compiled and used? You already did this once...
00:03:27[IDC]Dragoncompiled with make, I guess...
00:03:43 Quit elinenbe (" HydraIRC -> <- IRC has never been so cool")
00:03:50[IDC]Dragonthen you'll run it on an image
00:04:16[IDC]Dragonthere is a script with test cases
00:04:42[IDC]Dragonbut it needs linux forthe full thing, I stripped it down
00:06:06quelsaruktime to sleep
00:07:56jyp_I got "hello world" to run on my gmini ;)
00:09:17[IDC]DragonHey, big congratulations!
00:09:26jyp_heh ;)
00:09:34jyp_I got a question for you though...
00:09:49jyp_I have a wait loop (to see the message)
00:10:07jyp_but that part works only when plugged to DC...
00:10:18jyp_Does it remind you of something ?
00:10:29[IDC]Dragonis it really writing to the LCD graphics?
00:11:25jyp_We've been analysing the archos firmware since a long time ... ;)
00:11:34[IDC]Dragonwhat's happening without DC? I don't get it.
00:12:01jyp_The message is displayed but the gmini crashes 0.1 sec after
00:12:17jyp_wheras I have a wait loop (10secs) when plugged in
00:12:32[IDC]Dragoncrashes or turns off?
00:12:46[IDC]Dragonmaybe you have to actively hold power
00:12:58jyp_crashes (I get a big HD click)
00:13:09jyp_It must be that
00:13:34[IDC]Dragonbesides the click, does it appear off?
00:13:47[IDC]Dragondisplay gone, etc.?
00:13:49jyp_Let me check ;)
00:14:23jyp_Mh, yes ... It shuts off
00:14:53[IDC]Dragonthen I'm sure you have to hold power by some software control
00:15:09jyp_Alright... Thanks for the info :)
00:15:11[IDC]DragonFM and Ondio are the same
00:16:16amiconn...iRiver too, iirc.
00:16:40amiconn[IDC]Dragon: I found a somewhat funny thing I'm not sure about. It seems unnecessary though...
00:17:16amiconnIn fat_opendir, the root dir cluster for FAT32 is calculated: startcluster = sec2cluster(IF_MV2(fat_bpb,) fat_bpb->rootdirsector);
00:17:58amiconnHowever, there is fat_bpb->bpb_rootclus readily available, and this is the only place where sec2cluster() is called...
00:19:20[IDC]Dragonmaybe an oversight
00:19:46[IDC]DragonI vaguely remember that sec2cluster() isn't used excessively...
00:23:47 Quit quelsaruk ("KVIrc 'System Virtue'")
00:25:54 Quit jyp_ (Remote closed the connection)
00:26:45lImbus[IDC]Dragon, amiconn: do you know "Understand for C++" ?
00:26:55lImbusits a software to analyse code
00:27:13lImbusit shows you which function is called how often, and so on.
00:27:16[IDC]Dragonyes, I've used it a few times
00:27:39amiconnI didn't know
00:27:44lImbusit's just because I was finally able to find a version + keygen-combo
00:27:48[IDC]Dragonbut not for rockbox
00:27:48 Join JJ-Demon [0] (
00:28:47***Saving seen data "./dancer.seen"
00:30:40 Join thegeek_ [0] (
00:33:48lImbusamiconn: I can draw nice and biiiig invocation diagrams
00:37:14 Quit thegeek (Read error: 60 (Operation timed out))
00:38:45 Join thegeek [0] (
00:45:46 Quit thegeek_ (Read error: 60 (Operation timed out))
00:48:59 Join jyp_ [0] (
00:58:13amiconnGrrr. I added the root dir offset calculations, but it still doesn't work :(
00:58:55*[IDC]Dragon feels sorry, too
01:02:41amiconnHowever, getting rid of sec2cluster() does work :-/
01:04:43amiconnI was also able to get rid of fat_open_root() for FAT16 support
01:05:22[IDC]Dragonyes, you can "absorb" these lines if you like
01:05:38[IDC]Dragonmy cache split wasn't successful either
01:06:14[IDC]Dragonthe recipie doesn't fail any more, but I still get a blank volume 0
01:06:26amiconnI guess I need to do some splashing (or using gdb on the player - grr)
01:06:32[IDC]Dragonwhen wildly clicking through thedirs
01:07:51 Quit lImbus (" HydraIRC -> <- irc client ownage!")
01:13:57amiconnHmm. Now that is really strange:
01:14:28amiconnWith my modifications, I do no longer get a panic when trying to create a dir. The dir is even created - correctly!
01:14:51amiconnBut still I don't see the root dir contents within rockbox !??!
01:15:02 Join lImbus [0] (
01:19:00[IDC]Dragonamiconn: maybe the 2nd bug kicks in now
01:19:25amiconn??? Do you mean the multivolume problems?
01:20:02[IDC]Dragonor, have you been talking about the 2nd volume's root?
01:20:13amiconn2nd volume's root.
01:20:26amiconnFirst partition is FAT32, second is FAT16
01:22:03amiconnWith the root dir offset being 0, it still works (like current cvs). I quickly tested it on my Ondio (with both volumes), but did not do the fat_test tests
01:22:46[IDC]DragonI also haven't done that recently
01:27:05 Quit Stryke` ("Friends don't let friends listen to Anti-Flag")
01:30:08[IDC]DragonI need to sleep
01:30:36[IDC]Dragonamiconn: feel free to tell logbot on how you do
01:30:42[IDC]Dragonor, cvs ;-)
01:30:48amiconnNite [IDC]Dragon
01:30:59amiconnI think I'm on to something
01:31:01lImbusgood nite
01:31:09[IDC]Dragonnite all!
01:31:14 Quit [IDC]Dragon ()
01:40:54amiconnI have a working version (though not yet very elegant)
01:43:09amiconnIt's still working on Ondio :) :) :)
01:50:42jyp_gratz & good night !
01:50:45 Quit jyp_ ("Leaving")
02:28:51***Saving seen data "./dancer.seen"
02:39:47 Quit lImbus (" HydraIRC -> <- IRC has never been so cool")
02:51:31 Join matsl [0] (
02:56:18 Quit matsl ("Leaving")
03:29:16amiconn@[IDC]Dragon & other log readers:
03:30:19amiconn(1) I'm pretty sure my FAT16 changes do work correctly, but I'll commit them only after the next daily build got built i.e. in the morning
03:32:09amiconn(2) I have to agree that the fat file access is non-atomic.
03:32:14amiconnfat.c itself does not contain calls to yield(), however, it calls ata_read_sectors() and ata_write_sectors(), which may yield if they have to wait for the medium.
03:32:59amiconnThis certainly will cause trouble with the fat cache if 2 files are accessed by different threads at the same time.
03:34:03amiconnThis problem will be more likely with MMC (Ondio), because there is more latency.
03:34:23amiconnThe problem is independent from multivolume support.
03:34:43amiconn*now* it's really time to sleep.
03:34:49amiconnNite all
03:35:00 Part amiconn
04:03:50 Join thegeek_ [0] (
04:03:50 Quit thegeek (Read error: 54 (Connection reset by peer))
04:28:53***Saving seen data "./dancer.seen"
05:31:28 Join midk_ [0] (
05:31:45 Quit midk_ (Client Quit)
06:28:57***Saving seen data "./dancer.seen"
07:43:44 Join digitall [0] (
07:47:41digitallAnybody around to answer a question which is not in the FAQ etc?
07:50:36digitallI have a Archos 6000 Player and am very happy with it ( Since installing Rockbox ) :-)
07:51:29digitallBut after I installed the 2.3 release as archos.mod, the backlight has been acting oddly
07:52:10digitallThe 2.4 release has not solved this and back dating to the 2.2 release did not either
07:53:10digitallSetting the backlight to off is OK, but when I set it to on, it flickers
07:54:34digitallIn fact, it appears to be active only when a key is held depressed i.e. the program is handling the keypress interrupt service routine
07:55:31digitalli.e. the register controlling the backlight was being continuously cleared in the main loop
07:55:43digitallat least this was my sumise
07:56:10digitallBut this has not been corrected by the revert to 2.2 which had the backlight working correctly
07:57:16digitallAt this point, I'll go and look to the source, but I thought that it might save time to speak to someone as I can't seem to find this in the DOCS or bugzilla
07:58:10digitallI'm wondering where the config is stored and whether this is in flash or RAM?
08:24:03 Quit midk (Read error: 54 (Connection reset by peer))
08:24:51digitallI'm currently wearing the metaphorical paper bag ;-(
08:25:32digitallApparently with the A6K, if you disassemble them to say replace a dud HDD as I did a week back
08:27:06digitallThe connection between the LCD backlight and the main board can become loose
08:27:33digitallBy pressing on the front just above the circular button, the problem has disappeared
08:28:33digitallProbably the return for the backlight is via the metal and was not making firm contact resulting in the sporadic behaviour observed
08:28:59***Saving seen data "./dancer.seen"
08:29:18digitallSorry for the bother. Problem with being a hardware engineer is the tendancy to blame the software...
08:30:24digitallA cautionary tale on preconceived notions when approaching a problem
08:31:38digitallMight want to add this to the FAQ on Disassembly and Reassembly of A6K Player in case others blame backlight problems on S/W
08:32:27digitallThanks for the time. Merry <Insert Winter Festival> and A Happy New Year!
08:33:16 Quit digitall ("ChatZilla 0.9.61 [Mozilla rv:1.7.5/20041219]")
08:37:02 Join [IDC]Dragon [0] (
08:37:37[IDC]Dragonoh, digitall is gone
08:38:16dwihno[IDC]Dragon: I see the multi-volume support going live! Congrats!
08:38:41[IDC]Dragonthanks, but we some issues left to clear
08:38:56[IDC]Dragon.. we have some ...
08:39:55dwihnomajor issues?
08:40:01 Join amiconn [0] (
08:40:09[IDC]Dragonyou could say so
08:40:23[IDC]DragonHi Jens "insomnia"
08:40:30amiconnGood morning
08:40:38dwihnomajor issues being?
08:41:04amiconn[IDC]Dragon: It seems you checked the log. What do you think concerning the non-atomic issue?
08:41:04[IDC]Dragonthe internal memory getting eclipsed sometimes
08:41:22[IDC]Dragonamiconn: sounds like to worry about
08:41:32amiconnBtw, I did not observe this yet with my changed fat16 root handling
08:41:47amiconn(The disappearing root I mean)
08:42:07dwihno[IDC]Dragon: whoa. strange matters
08:42:11[IDC]Dragondid you change anything more than the offset?
08:42:32amiconnMaybe I simply did not yet try it hard enough
08:43:10[IDC]Dragonany idea on how to make the f/s thread-safe?
08:43:28[IDC]DragonI started thinking, but didn't reach a conclusion
08:43:58[IDC]Dragonexcept for plastering the code with mutex
08:45:54amiconnWe need to check which places are non-atomic, i.e. calling the ata sector functions, and don't trust local variables derived from the fat cache afterwards.
08:46:58[IDC]Dragonthat's a generic, plastering solution
08:46:59amiconnI think the only other option would be a mutex used in those places
08:47:39[IDC]Dragonyes, I prefer a smart placed mutex
08:48:16[IDC]Dragonit would need to be held as long as working on a cached fat sector
08:49:56[IDC]Dragonbut that's many places, here we go plastering again
08:50:39[IDC]Dragonan "uncache" function would help
08:52:12amiconnBtw, my FAT16 changes do also fix the handling of files in the root dir (the dircluster value in the fat_file structure was unsigned...)
08:52:48[IDC]Dragonfix files in the root?
08:54:27[IDC]Dragondircluster value?
08:54:38 Join amiconn_ [0] (
08:54:52 Quit amiconn (Nick collision from services.)
08:54:53 Nick amiconn_ is now known as amiconn (
08:55:49amiconnfile->dircluster was unsigned, so it would not have pointed to the correct place in case the file was located in the root dir
08:56:20[IDC]Dragonor, only after casting
08:56:47[IDC]Dragondid we have problems with root files?
08:57:37*[IDC]Dragon is curious for a commit
08:58:05amiconnThe struct was defined that way. I did not observe problems myself, maybe this struct member is not often used.
08:58:27[IDC]Dragonperhaps it was used in a signed calculation
09:01:14amiconnPerhaps someone could do these test_fat tests...
09:01:21 Join Zagor [242] (
09:01:30[IDC]Dragonthat someone?
09:01:42[IDC]DragonHi Zagor, welcome back
09:01:42Zagorhi all
09:03:16[IDC]Dragonwe're messing around with your fat code again...
09:05:19Zagorok, what's the issue
09:06:17[IDC]Dragonwe think we have problems because the file system is not thread safe
09:06:39[IDC]Dragonintroduced by ata_read/write_sector yielding
09:08:00Zagorhmm, looking
09:08:02[IDC]Dragonwe broke it, not you ;-)
09:08:19[IDC]Dragondid you read the log?
09:08:35Zagornot yet. which day should I start with?
09:09:07[IDC]Dragontoday, basically
09:11:16[IDC]Dragonthe HD ata code does yield, too, if I read that correct
09:11:42[IDC]Dragonso perhaps the problem was lurking around all the time
09:12:27[IDC]Dragonthen I wonder why we had no real world issues
09:13:57Zagorgotta go a bit. bbl.
09:16:14[IDC]Dragonamiconn: your commit didn't change my eclipsing
09:16:15amiconn[IDC]Dragon: Yes the HD ata code does yield too
09:16:48amiconnOkay, so it seems I simply didn't try hard enough :-/
09:17:30amiconnHowever, it definitely does fix the large FAT16 partition access, where the root dir does not start on a cluster boundary :)
09:17:50amiconnPlus, it saves quite some code :)
09:18:06[IDC]Dragonyes, that's nice
09:18:19amiconn...even a bit for FAT32 only platform (~100 bytes)
09:19:19[IDC]Dragona nice fix, it's just unrelated to the eclipse problem
09:19:41amiconnI think this has to do with the fs not being thread safe
09:20:02[IDC]Dragonthere is a mutex about the sector read itself
09:20:20[IDC]Dragonor write
09:21:20[IDC]Dragonlet me think again on when the problem can happen. We only work on one cached sector in a local function
09:22:28[IDC]Dragonso, if thread A wants to change the fat and reads a sector in the cache,
09:22:41[IDC]Dragonthat code may yield to thread B
09:23:14[IDC]Dragonworking on a different fat sector being mapped to the same cache location
09:24:07[IDC]Dragonit just won't enter the ata code, because of the mutex
09:25:42[IDC]Dragononly after the mutex unlock, and more importont after the next "free" yield
09:26:10[IDC]Dragonwhich is not happening before the function ends
09:26:32[IDC]Dragonbecause we have no preempting multitasking
09:29:31Zagorthe thread doesn't have to want to change the fat, does it? merely accessing files outside the cached fat will risk corrupting the cache for the other thread?
09:30:04[IDC]Dragonah, yes
09:30:33Zagorhowever i agree a mutex is the right solution
09:30:43amiconnBut the function call from thread B may then use some cache-related values it precalculated before its ata call (where it was blocked by the mutex) after ths call. They may no longer be valid...
09:31:01Zagoryes, a fat mutex is needed.
09:33:05[IDC]Dragonmaybe it can be a bit more simple/relaxed, taking into account that the ata code is mutexed, we only work on a single sector, and don't yield meanwhile
09:34:08Zagorif we don't yield in the sector reads, we won't yield at all. and that will make the whole interface very sluggish during buffer refills.
09:34:42amiconn[IDC]Dragon: Imho these scenarios don't explain why the root dir vanishes on the Ondio. The root dir is the only dir *not* using the fat with FAT16...
09:35:02[IDC]DragonI'm not argueing against yielding
09:35:32[IDC]Dragonplus, the vanishing scenario is read-only
09:39:26amiconnDoes this happen with cvs code, or with your 2 separate caches?
09:40:07[IDC]Dragona bit different with 2 caches, but still
10:10:00 Part Zagor
10:14:15 Join Zagor [242] (
10:21:31 Join UnjustlyB [0] (Hydronic@rescomp-04-61236.Stanford.EDU)
10:21:38UnjustlyBHey anyone here?
10:22:26UnjustlyBSo according to the FAQ, there is no way to incorporate support for winamp-like audio plugins in rockbox?
10:22:41Zagornot on the archos hardware, no
10:22:51UnjustlyBdamn. why is it impossible?
10:23:13Zagorthe hardware handles mp3 data directly, so we never get to touch and refine the decoded audio stream
10:23:44UnjustlyBbut it seems like you guys have kinda altered the bass and treble features..
10:24:01Zagorno, we just make full use of the features already available in the hardware
10:24:43UnjustlyBand as far as sound quality goes, you guys can only change bass and treble?
10:25:05UnjustlyBit seems like winamp-like plugins really only play with those two settings
10:25:48UnjustlyBwhat about the newer archos's.. can you guys add support for plugins on those?
10:26:27Zagorpossibly, it depends on the hardware. but we don't have rockbox for those. there's a group working on a port for gmini, but it's very early still.
10:28:24 Quit [IDC]Dragon (Read error: 110 (Connection timed out))
10:29:02***Saving seen data "./dancer.seen"
10:30:17 Join [IDC]Dragon [0] (
10:30:31[IDC]Dragonnow I get identical, simple reproduceability, regardless of 1 or 2 fat caches
10:30:58[IDC]DragonI'm beginning to believe tisis a more simple, plain bug
10:31:11[IDC]Dragons/tisis/this is
10:42:54UnjustlyBZagor: Adding new sound effects requires reprogramming the MAS chip, and we can't do that. The MAS chip is programmable, but we have no access to the chip documentation.
10:43:09UnjustlyBis this why you guys cant add support for an audio plugin?
10:44:10Zagoryes. also, reprogramming the mas removes the mp3 decoding capability, which of course is not very desireable.
10:44:59UnjustlyBclearly i am a programming nub. you guys cant reprogram it but leave the mp3 decoding capabilites intact?
10:45:14UnjustlyBor reprogram it with *new* mp3 decoding capabilities
10:46:01Zagorno, iirc the function of the chip is to either run the internal mp3 decoder code or the custom code. not both.
10:46:53UnjustlyBso you guys cant reprogram it with a custom code?
10:47:13Zagornot if we want to use the mp3 codec
10:47:23UnjustlyByou cant import your own?
10:47:37Zagorno, the area for the custom code is very small
10:48:08UnjustlyBwell thanks
10:48:17Zagornp, good night
10:48:21 Quit UnjustlyB ("ircN 7.27 + 7.0 for mIRC (2002/01/10 00.00)")
11:01:31 Quit [IDC]Dragon (Read error: 110 (Connection timed out))
11:11:44 Join [IDC]Dragon [0] (
11:17:11 Join pfavr [0] (
11:29:18pfavrThanks for the very nice Christmas present
12:09:24 Join ashridah [0] (
12:29:04***Saving seen data "./dancer.seen"
12:34:09 Join lImbus [0] (
12:58:36[IDC]Dragonamiconn, Zagor: a test version with no fat caching at all shows the same symptom
12:59:34[IDC]Dragonso it must be something else
13:03:24Zagorok. interesting
13:08:55 Join quelsaruk [0] (~kvirc@
13:09:35[IDC]Dragonnot really interesting. maybe very dull to debug...
13:12:28[IDC]DragonI should disable multivolume for the public while this is ongoing
13:18:21quelsarukhmm.. Offtopic question (as always)... any french here right now?
13:33:31quelsarukanyway, normal working hours in a european company are...¿? (Spanish companies are different)
13:36:33BagderI'd say it differs all over
13:37:07quelsarukit's to know when can i call archos france...
13:37:26quelsaruktomorrow expires the warranty and i *want* a new HD
13:37:37BagderI'd say they should work right now
13:37:53quelsarukuntil what hour?
13:38:13Bagder~18:00 or so
13:38:26Bagderjust my guess
13:38:54quelsaruki'm going to call them then :)
13:39:11 Nick quelsaruk is now known as Quel|away (~kvirc@
13:39:12 Join Lynx_ [0] (HydraIRC@
13:40:04*Bagder beeps
13:42:42ashridahheh. not even sure where i thought i was typing. this was the only terminal i had open.
13:51:50 Quit pfavr ("ChatZilla 0.9.61 [Mozilla rv:1.7.3/20041007]")
14:11:36[IDC]Dragonhey, I just saved 200 bytes in dir.c
14:12:59crash_zzzodersomaybe a dump question, but is this important?
14:13:15 Nick crash_zzzoderso is now known as crash_ (
14:13:21[IDC]Dragonmeaning, use a pointer instead of the same array reference again and again
14:13:49Zagor[IDC]Dragon: i would have thought the compiler did that already
14:13:56Zagorcrash_: is what important?
14:14:12[IDC]Dragoncrash_: 200 bytes code is some amount for us
14:14:14crash_is it important to replace this occurrences in the source code
14:14:45crash_thats why i ask, thought this would be "intelligently" solved by the compiler ;)
14:15:13[IDC]DragonZagor: perhaps it doesn't becaude opendirs[] is global
14:16:05[IDC]Dragonand modified god-knows-where from a compilers' view
14:16:47[IDC]Dragondo we have more spots like this?
14:16:58Zagorwell it is static, so the compiler would know. but it's still a reasonable theory.
14:17:31Zagortry moving the declaration inside opendir() and see if it makes a difference
14:20:09 Quit Quel|away (Read error: 60 (Operation timed out))
14:21:46[IDC]Dragonno difference
14:22:05[IDC]Dragontried both local static and automatic
14:22:14[IDC]Dragongcc sucks
14:22:31Bagderwe better write our own! :-)
14:22:52[IDC]Dragonfeel free to file a patch
14:23:32crash_Zagor: btw i never had the chance, so i'm wanted to say a big thank all you guys for rockbox. i'm with this project since about june 2002 and you've really done great work :)
14:24:02crash_hope you guys can imagine what you have reached
14:24:21Zagorworld domination? ;)
14:24:31crash_not yet
14:24:44crash_but i'm sure this will come
14:25:22crash_you know if linus had any luck in progressing with the iriver ?
14:25:45ashridahhow's progress on the iriver port been recently? slow/medium/blazing? any new milestones since basic lcd drawing?
14:26:04Zagorthe boot loader is nearing
14:26:12crash_great :)
14:26:29Zagorashridah: we've got basically everything working now, including flashing
14:26:41crash_hope i can help sometime with giving info about h3xx
14:29:08***Saving seen data "./dancer.seen"
14:29:26[IDC]Dragonanother 50 bytes for replacing just the search access in dir.c:84 with a pointer
14:29:50[IDC]Dragonamazing, arrays of structs seem expensive to address
14:30:52JJ-Demonanyone here with a fast internet connection that can test something?
14:31:23ZagorJJ-Demon: sure
14:31:34JJ-Demoni think we have a broken router
14:31:41JJ-Demonneed someone other then where i live to test though
14:31:45JJ-Demontell me if u get crappy speed
14:31:53Zagori do
14:31:53gromit`maybe i can test
14:32:03gromit`wget ?
14:32:22Zagoraverage ~14kb/s
14:32:33Bagderme too
14:32:35gromit`same for me
14:32:56JJ-Demonthis jsut started 10 mins ago
14:33:01*JJ-Demon calls theplanet
14:33:33crash_speed goes from 7k to 25k with 100mbit connect here in germany
14:33:59JJ-Demonyea normal its full speed
14:34:07JJ-Demonits on all of our servers on our router
14:34:13crash_what does fullspeed mean for you?
14:34:32JJ-Demonnormal net speed you get over your connection?
14:34:36gromit`same on a 100Mb in france so
14:34:50crash_this depends on the server
14:34:55JJ-Demonsorry i meant to say
14:35:02JJ-Demonnormally its full speed*
14:35:36JJ-Demontry ti again
14:35:52gromit`i have 8.5Mbps on
14:35:59crash_goes from 200 - 300k
14:36:03crash_should be ok ?
14:36:23gromit`64k on
14:36:41crash_364k @ top
14:36:56JJ-Demonthat sounds better?
14:37:10gromit`14:38:02 (7.93 MB/s) - « 1200k.test.2 » sauvegardé [1239192/1239192]
14:37:20gromit`ya better :)
14:37:26gromit`but maybe cache :)
14:37:38JJ-Demondamn they are fast... musta flicked the wrong switch somewhere earlier
14:37:43crash_no cache here, and all the time about 300k ;)
14:38:32Zagori still only get 200kb/s
14:38:59gromit`14:37:44 (218.06 KB/s) - `1200k.test' saved [1239192/1239192]
14:39:01gromit`at home
14:39:04crash_aehm a difference from 14k average to 200k should be ok, diesnt it?
14:39:18Zagordepends on what you are paying for...
14:39:36crash_yeah for sure, just reminded of the difference
14:39:38JJ-Demonwhats your home connection?
14:39:44crash_and that he thought of a crappy router
14:40:11JJ-Demoneveryone else im asking is getting full spee dnow
14:40:16crash_@home dsl (1024kbit) / work: direct connect to backbone of 100mbit
14:40:16JJ-Demonfor their connections
14:40:36crash_JJ-Demon: where is this server located ?
14:40:39gromit`no full speed for me
14:40:54JJ-Demonat theplanet
14:40:57JJ-Demonin texas
14:41:17crash_yeah i see, should have done trqaceroute before asking ;)
14:42:16crash_tested again, seems like the file is too small for reaching line limit here
14:43:05JJ-Demonover local network
14:43:11JJ-Demonit cant go past 9.45M/s
14:43:16JJ-Demoncaus eof the filesize
14:43:47JJ-Demonumm test one of the others too
14:44:51Zagorsame thing, never reaches above 200kb/s. takes 6 seconds to download on my 100Mbit connection
14:45:14crash_same with the other
14:45:17crash_350k top
14:45:20gromit`230kbps at 15mbit
14:45:42ZagorI reach 3800+kb/s on a 2MB file on another server
14:46:11JJ-Demonstrange we pay enough for these issues not to arrise lol
14:46:23JJ-Demonbetter send in a ticket too about it
14:47:07crash_dont think this will help. you maybe have to see, that our tests all are done from europe
14:47:34ashridahi've got a site in the US. gimme a few
14:47:46ashridahnot sure how well it's connected, but it should be reasonable
14:48:29ashridahjust got 600Kbits/s
14:48:48ashridahnot kb
14:49:20ashridahyeah, seems to peak out at that.
14:53:14JJ-Demoni get 19Mb/s uploading from the server to a server at
14:53:57[IDC]Dragonmore on gcc opt: just tried pointers in fat_init(), no change there
14:54:15JJ-Demontnx guys :)
14:54:31[IDC]Dragonthis is also a static array
14:55:06[IDC]Dragonbut with a power-of-2 size, that may be the difference
14:58:59Zagorthat would mean the structs aren't padded, which would be quite bad
15:00:16Zagoruh, forget that
15:00:20[IDC]Dragonpadded to the next 2^n?
15:09:42 Join midk [0] (
15:13:53 Quit ashridah ("sleep")
15:14:41 Join einhirn [0] (
15:15:55 Quit JJ-Demon ("—I-n-v-i-s-i-o-n— 2.0 Build 3515")
15:34:23 Join c0g0 [0] (
15:34:51c0g0ja har en ratt latt fraga om de e nan som har lust att svara...
15:35:21lImbusuh ? english ?
15:35:22Lynx_he is speaking in tongues ;-)
15:35:29c0g0sorry :D
15:35:44c0g0ive got a question for u guys, i wonder if u have any information about the archos 480
15:36:11c0g0does it exist other firmwares to it as to the jukebox?
15:37:17c0g0or will there be any?
15:37:25Zagorpossibly linav works on it, but it's not a complete firmware replacement like rockbox
15:37:47c0g0oh thx man
15:39:50c0g0hmm i also have a pretty hard question, im goin to singapore in 25 days, and im gonna by an media player. i dont really know wich one i should buy cause i want to be able to listen to music and i want to view videos and i want a big screen, but if it possible i want to play games on it also
15:40:48c0g0i heard that you can update the firmware on some iriver pmp100 series so u can play nes emulator on it, and i wonder if u know if it is the same with the archos av400 series?
15:41:29Zagorno idea. we're not much into the video units.
15:41:56c0g0aight... thx man
15:41:56c0g0ill ask in linav ;D
15:43:11Zagorbtw, their channel topic is "linux on av3xx | does not work on av4xx"
15:43:53c0g0ya i saw ;/
15:44:35[IDC]Dragonc0g0: how about the new iriver then?
15:44:51c0g0ya i wonder about that one...
15:45:19[IDC]Dragonif the hardware is similar, maybe one day there'll be rockbox on it
15:45:21c0g0but the negative thing with it is that it has only 40 gig and 3.5 inch instead of 3.8 i think
15:45:40[IDC]Dragonand we need somebody to find out... ;-)
15:46:22[IDC]DragonI was playing with it in a shop, way cool display and handling
15:46:24c0g0hehe ;D
15:46:35Zagordoes it really play video?
15:46:47c0g0ya all mpeg4
15:46:47[IDC]Dragonlooked pretty nice, yes
15:46:49Zagorthe h300, i mean. not the pmp.
15:47:08c0g0i think so but its a smaller screen
15:47:09[IDC]Dragonoh, sorry, I thought you meant the newer one
15:47:11Zagor[IDC]Dragon: or are you also talking about the pmp?
15:47:27crash_the h3xx plays videos just fine
15:47:34[IDC]Dragondon't know how exactley it's called
15:47:36crash_though only 10 fps
15:47:43Zagorcrash_: ok
15:47:55[IDC]Dragonbut the one with the "big" screen, gameboy look
15:47:56c0g0crash_ but its smaller screen right?
15:48:18crash_and resolution of 220x176
15:48:31c0g0hmm wich one is the most hacked media player yet?
15:48:37c0g0does anyone know?
15:49:02Zagorcrash_: is that the h300?
15:49:26crash_but the "screenshot" is f*** up :/
15:49:41crash_the picture is really clear and looks great
15:49:49c0g0ohh nice
15:49:52crash_unfortunatly you cant get this with a cam
15:50:18crash_hope someone of rockbox will handle video someday ;)
15:50:39[IDC]Dragonwe do...
15:50:49crash_already ?
15:50:58crash_i thought this is something lower prio ?
15:51:12[IDC]Dragonrockboy has a video player
15:51:30crash_i know, but dont have that much information about the capabilities
15:51:34c0g0hmm wich one?
15:51:50crash_cause my 6000 jukebox isnt capable of playing video
15:51:55Zagorfor archos recorder only
15:52:05[IDC]Dragon112*64 "greenscale"
15:52:27Zagorcrash_: nobody thought the records could either :)
15:53:12crash_so there is hope i will someday watch video delivered by rockbox on my h320 :)
15:53:34crash_one step closer to world domination
15:53:39c0g0anyone have any suggestion of wich video player is the best?
15:54:02crash_c0g0: you could go for a win ce powered pda ?!
15:54:27crash_c0g0: but this wouldnt be my idea of having portable video ;)
15:54:40c0g0hmm... hehe thats what i want...
15:54:43c0g0the vidoe..
15:56:10c0g0i know pretty much about the archos and the iriver now i think, and its between the pmp140 or the av480, but if there is some other player that u have any suggestion of?
15:58:32crash_i think you have found the main competetors
15:59:51c0g0i think ill go for the archos anyways...
16:00:06c0g0but i really hope its goin to get hacked
16:00:13c0g0if it hasnt already
16:00:21[IDC]Dragonvery unlikely
16:00:43c0g0that it will be or that it is?
16:01:11c0g0ohh.. why do u think that?
16:01:20crash_btw is it right to say, that the future of rockbox is going to iriver?
16:01:54 Join quelsaruk [0] (~kvirc@
16:01:55[IDC]DragonArchos lately is "sealing" their products with real encryption
16:01:57quelsarukhi again
16:02:20Zagorcrash_: iriver is the next target hardware, so the immediate future will be there.
16:02:38crash_this is what i meant
16:03:02crash_as is said before im watching rockbox since mid 2002 and always saw, that the multimedia devices
16:03:05Zagor[IDC]Dragon: is that confirmed for av400 too?
16:03:06crash_coudln be supported
16:03:27[IDC]DragonZagor: no, just an extrapolation
16:03:32Zagorcrash_: the problem with supporting them always was lack of public documentation.
16:03:38c0g0ohh.. so if i want to be sure it will be homebrew software i should go for the iriver pmp series?
16:03:56crash_iriver seems to code same shitty firmware like archos in early times, but as [IDC]Dragon said that archos is scrambling their stuff
16:04:14Zagorc0g0: is anyone working on firmware for the pmp?
16:04:19crash_c0g0: dont think anyone will give you any guarantee
16:04:23c0g0i dunno?
16:04:32c0g0i thought u might know ;P
16:04:41[IDC]Dragonthe av500 could be "open" again, if itfeatures linux
16:05:00c0g0ya, i wonder about the reldate?
16:05:29crash_c0g0: the rockbox guy work on the "early" archos devices and h1xx (and hopefully h3xx), all other devices arent really planned if i followed correctly ;)
16:05:31Zagorthe fact is rockbox is the only howebrewn mp3 player firmware in existence, and will probably stay so for quite some time. there have been various linux ports, but no other complete firmware replacements.
16:05:55crash_[IDC]Dragon: dont u think this is only for a stable basis, and wont be that easily cracked ?
16:07:16c0g0hmm... this is making me really hard to make a disition >;D
16:08:10[IDC]Dragonin any case, it runs with texas chips again, for which there's no open compiler
16:08:34crash_[IDC]Dragon: you speak of the archos linux powered devices?
16:09:12crash_c0g0: i would recommend buying the device you like the most right know and sell it later if there is something you can use with other firmware...
16:09:56Zagorthat sounds like a good strategy to me
16:10:26[IDC]DragonI'd check what's inside the iriver and if similar to our coldfire, get that
16:10:38[IDC]Dragon(a developer opinion)
16:13:28c0g0ya maybe thats the best way...
16:13:51 Join echs [0] (
16:13:52c0g0its pretty similar
16:14:16c0g0but thou it has only 40 gig and a 3.5"
16:14:26c0g0agianst 80 gig and 3.8
16:14:48[IDC]Dragon1.8" drive vs. 2.5"
16:15:09c0g0and videoplayback with640x480 against 704x480
16:15:13[IDC]Dragondepends on wheter you want it lightweight or high-capacity
16:15:25c0g0prefer high capacity
16:15:48[IDC]Dragonthen get what *you* need
16:16:25c0g0he... ya.. maybe it will be something about the linav also?
16:16:41c0g0or some other guys maybe will look into it
16:16:57Zagorbtw, the pmp runs microsoft windows "portable media center".
16:17:46c0g0u know what the av4xx runs?
16:17:55Zagorcustom archos software
16:17:56echscustom archos os?
16:18:10[IDC]Dragonfrom the specs, Archos is better
16:18:20echstheres also the pmp-100 from iriver running linux..
16:18:38[IDC]Dragononly 20g extra weight for a 2.5" drive, amazing
16:19:03c0g0ya it is..
16:19:31crash_hi echs
16:19:33Zagoroh, my mistake. it's the iriver PMC that runs windows.
16:19:55c0g0ohh ok
16:20:01echsoh hi crash :)
16:20:10[IDC]DragonArchos is even smaller
16:21:12crash_so i would conclude, including my former stragedy: now buy archos, in future go for iriver with rockbox ;)
16:21:31c0g0maybe something like that ?D
16:21:40Zagoryou may also wish to consider that the av400 is the third generation archos video player while the pmp is iriver's first attempt.
16:22:05echsi thought it would be too much effort to make a similar os like the one from iriver for the pmps?
16:22:19echsthe pmp is just too expensive..
16:23:19c0g0okay guys..
16:23:24c0g0thanks for the help!
16:23:29c0g0igotta go
16:23:30c0g0cu !
16:23:33 Quit c0g0 ()
16:27:31quelsaruk[IDC]Dragon: do you have your car radio/rockbox interface somewhere in the web?
16:29:11***Saving seen data "./dancer.seen"
16:29:31[IDC]Dragona plugin, yes
16:30:15[IDC]Dragonwhat do you mean with "interface"?
16:31:22 Join methangas [0] (
16:31:30quelsarukwell... you need a microcontroller or something to understand what the radio is saying and explain to rockbox, isn't it?
16:32:32[IDC]Dragonthe "microcontroller" is the Archos CPU
16:33:44quelsarukso is not like the gnunilink or the vwcdpic project
16:34:10[IDC]Dragonsee links to my postings, on the bottom of
16:34:25[IDC]Dragonall you need is a cable
16:34:52quelsaruki wanted something like this last page
16:52:37 Join Lynx0 [0] (HydraIRC@
17:03:30 Quit Lynx_ (Nick collision from services.)
17:04:11 Nick Lynx0 is now known as lynx_ (HydraIRC@
17:07:37 Nick quelsaruk is now known as Quel|away (~kvirc@
17:12:29 Part Zagor
17:14:45 Join webguest93 [0] (
17:14:53 Quit webguest93 (Client Quit)
17:17:17 Join mecraw_ [0] (~lmarlow@
17:52:34 Part lImbus
18:04:40 Quit lynx_ (" The IRC Client of the Gods! -> <- HydraIRC")
18:07:52 Join jyp [0] (
18:12:46 Join zeekoe [0] (
18:17:34 Join einhirn_ [0] (
18:17:46 Quit einhirn_ (Client Quit)
18:25:49jypWorking on the gmini lcd driver...
18:26:06jypBasically lcd-recorder.c works out of the box on gmini
18:26:23jypexcept it is reversed left <-> right
18:26:27crash_jyp: i just saw a gmini in live, is there any progress status published on
18:26:43jypI think not
18:27:10jyp(this is gmini 100 & 200 series)
18:27:56crash_shouldnt someone do this?
18:29:03jypthe gmini team has a website of its own... Though it is down now
18:29:11crash_ah ok
18:29:14***Saving seen data "./dancer.seen"
18:29:24crash_porting it to rockbox would make sense, doesnt it?
18:29:55crash_it seems like you can edit the wiki if you register, so i have some spare time and would like to help if there is interest
18:31:04 Join zeekoe2 [0] (
18:31:29crash_just tell me how i can help
18:31:36 Quit zeekoe (Nick collision from services.)
18:31:42 Nick zeekoe2 is now known as zeekoe (
18:31:57jypthe truth is, the project leader and hoster of the website hasn't shown up since 20th december
18:32:07zeekoehm.. just like avos...
18:32:44jypI've done quite a lot of progress since though
18:32:49crash_if webspace is the problem, this is easily solved, but i think main focus should ly on making it public
18:32:51jypso the project is live
18:33:15crash_just like me, showing somone rockbox and seeing, that there is nothing on rockbox,org
18:33:45jypYes... The thing is the situation is somewhat embarrassing...
18:34:53crash_so should we speak with zagor or someone other responsible if we could put it into the wiki?
18:34:57jypBasically I don't like to make a political decision with Strath (leader) away
18:35:29crash_hmm ok although i think putting it to the wiki isnt a "political decision" ;)
18:35:35jypStrath has started the gmini project, I mean
18:36:12jypAlso, we have quite a detailed wiki already... Not not accessible :/
18:36:29 Quit mecraw_ (Read error: 113 (No route to host))
18:36:36jypnot not -> But not
18:36:51crash_this is really a shame :/
18:37:04crash_and u dont have a idea where strath is?
18:37:27gromit`no :/
18:37:34jypWe *hope* he's on hollidays... But he didn't say
18:37:54crash_jyp: to get it clear, your port to the gmini is rockbox and not just an extension or something?
18:38:03jypI had expected him to show up today
18:38:11 Quit methangas (" HydraIRC rocks! -> <-")
18:38:14zeekoewiki really is accessible
18:38:24jypThat's what I plan
18:38:25zeekoejust register, and you can edit what you lole
18:39:00jypis the wiki I'm talking about
18:39:33zeekoe\kick jyp >#gmini ;-)
18:40:05crash_jyp: at first site i think is is a bit too technical for the average user?
18:40:10zeekoeand this does not work?
18:40:27jypMm, I can't access it from here...
18:40:42crash_i dont have problems accessing the wiki
18:40:52jypalright then
18:41:08gromit`ha ?
18:41:15gromit`crash : you can access gmemu wiki ?
18:41:29gromit`i can't !
18:41:43gromit`and ?
18:41:55zeekoeworks too
18:42:14gromit`it is down from my job and from my home
18:42:36jypfrom belgium/skynet too ...
18:43:03amiconnWorks from here (Germany/T-Online)
18:43:10gromit`france/free and france backbone
18:43:24zeekoedoes that work?
18:43:58*zeekoe @holland
18:44:06crash_the site of strath is accessibale without a problem
18:44:34crash_gromit`: no problem with that
18:45:00zeekoeit works through too
18:45:15zeekoeyour provider blocked it :P
18:46:02jypAnyways, crash_, what to you want to do ?
18:46:12jypReport on the progress?
18:46:22jypIf so feel free to do it ;)
18:46:24 Quit Bagder (
18:46:24 Quit ZiRo`5 (
18:46:24 Quit crash_ (
18:46:43zeekoedon't say that again ;)
18:47:33NJoinBagder [0] (
18:47:33NJoinZiRo`5 [0] (
18:47:33NJoincrash_ [0] (
18:47:38zeekoehuh? i'm at zelazny too...
18:47:53jypinternet = hardly any service :/
18:48:35zeekoeoh, it's niven
18:48:42zeekoehey, but internet is cheap
18:48:56 Quit ZiRo`5 (
18:48:56 Quit Bagder (
18:48:56 Quit crash_ (
18:49:35NJoinBagder [0] (
18:49:35NJoinZiRo`5 [0] (
18:49:35NJoincrash_ [0] (
18:50:49jyphello, Huston?
18:50:59jypdo you copy ?
18:52:00jypSo, crash, what is your plan ?
18:52:26 Nick zeekoe is now known as Houston (
18:52:31 Nick Houston is now known as zeekoe (
18:53:20zeekoeseems they're still floating around in cyberspace
18:53:53jypThey didn't know they were entering... the twilight zone...
18:55:00zeekoeis that a song?
18:57:40zeekoehm.. fun... i only knew of the name 'twilight zone' by the cd's with illegal software being around a few years ago
18:58:34crash_what a netsplit ;)
18:58:46crash_is a good question
18:59:20jypyour plan?
18:59:22crash_i recently read a post/mail/something of one of the developers, that the wiki should be used for documentation and that too technical stuff shouldnt be there
18:59:56jypthen we'll keep a separate wiki for the tech stuff
18:59:58crash_so i would like to talk with zagor or someone other responsible for the wiki if its ok to put your wiki completly over to rockbox
19:00:03crash_dont know
19:00:08crash_this question is open
19:00:13crash_i dont like this idea
19:00:39crash_but i can imagine this already has been discussed
19:00:43jypI don't have any definite opinion
19:00:52jypFlatly, I don't care ;)
19:01:17crash_i think you can separate the technical stuff quite easy with setting no links ;)
19:01:37crash_but we should discuss this with some of the rockbox guys ;)
19:01:55crash_so who is responsible / has overview over the rockboxwiki ?
19:02:01zeekoeor just a "Warining - the following links are for approved service personnel only"
19:03:16crash_sounds like a sentence fvrom matrix: you can step over this line but we dont take any responsabilty for any knowledage you may gain ;)
19:03:25zeekoebjorn stenberg can reset passwords
19:03:41zeekoei think that's zagor
19:05:54jypAnyways... I was going to ask where I can find the technical domumentation for the recorder lcd
19:06:32[IDC]Dragonin the datasheet section
19:06:57jypchecking ...
19:11:37[IDC]Dragonamiconn: you're probably not really around?
19:12:19amiconnI am here
19:12:55amiconnDo you have some tests for me, e.g. checking on the player with 2 FAT16 partitions?
19:13:07[IDC]Dragonmaybe I've been seing ghosts, but now I can't reproduce the problem without cache
19:13:22[IDC]Dragonbut well with cache
19:14:04[IDC]Dragonsome debug out tells it tries to access out of range sectors #s if the problem approaches
19:14:22amiconnStrange. You reported that the problem is independent from the caching earlier!?
19:14:42[IDC]Dragonthat's the "seeing ghosts" part
19:14:43amiconnHow do you get the debug output?
19:14:54[IDC]Dragonredefineing DEBUGF
19:15:17[IDC]Dragonlocally, to use my pseudo-splash
19:15:39amiconnLet's see if this is reproducable with 2x FAT16 on the player. Then I could try to use gdb (not that this would be fun...)
19:16:09[IDC]DragonI know
19:16:38[IDC]Dragonbetter wait a bit before really gdb-ing
19:16:45amiconnI wonder why out-of-range accesses don't trigger the panic, or are they read-only?
19:16:59[IDC]Dragonread, yes
19:17:59[IDC]Dragonthe fat cache probably sends it into the forest
19:19:00amiconnLuckily the player is flashed, so it's possible to start with 2 FAT16 partitions.
19:19:43[IDC]Dragonah, yes, a side problem
19:20:22amiconnNot a big one though, this would also be possible with archos in flash (with a total of 3 partitions)
19:21:20[IDC]DragonI need to meditate a bit more in the fat cache code
19:21:41amiconnBtw, I'll check the various partition types soon. It seems we don't allow all fat 16 variants (< 32 MB, > 32 MB and LBA)
19:22:14[IDC]Dragonforget <32 MB
19:22:29amiconnWhy? Someone may dig out an old MMC...
19:22:33[IDC]Dragonbesides, I don't know about such variants?
19:24:01[IDC]Dragonfirst I need to go grocery shopping...
19:24:03amiconn[IDC]Dragon: FAT16 <32 MB is type 0x04, FAT16 >32MB is type 0x06, and FAT16 with LBA addressing (located above 8 GB) is type 0x0E
19:24:22[IDC]Dragonlater, gotta run
19:24:25 Quit [IDC]Dragon ("CGI:IRC")
19:24:51 Join mecraw_ [0] (~lmarlow@
20:29:15***Saving seen data "./dancer.seen"
20:38:55jypWe just had a "rockbox gmini" project meeting ... :)
20:39:50jypOur CVS server has been down for two weeks and the person hosting it hasn't given news since then...
20:40:22jypSo, since we plan to "merge" with rockbox anyways... would you provide us with cvs hosting ?
20:58:10amiconnjyp: I'm afraid only LinusN, Zagor or Bagder are able to answer this request.
20:58:43jypThanks for the answer ;)
20:58:44Quel|awayyou can always get cvs hosting from sourceforge
20:59:13jypBut it mage sense to regroup
20:59:16jypjust asking
21:00:35 Join mecraw__ [0] (~lmarlow@
21:03:07 Quit mecraw_ (Read error: 60 (Operation timed out))
21:04:24Quel|awayof course, i just pointed out another option
21:05:29Quel|awayamiconn: i finally contacted archos and they told me to send them the box, they pay the shipment :)
21:05:58Quel|awayi suppose they will change the drive and send mine to seagate, as seagate gives 3 year warranty
21:12:24 Join [IDC]Dragon [0] (
21:13:20[IDC]Dragonhi again
21:17:54amiconnDo you have a reproducable method to trigger the vanishing root dir?
21:18:17[IDC]Dragonthe one I told yesterday
21:23:32amiconnHmm. I didn't manage to make the root dir completely vanish again since I changed the FAT16 handling.
21:23:40amiconn(on Ondio, that is)
21:24:33amiconnThe only effect that I am able to trigger is that the root dir is displayed incomplete, when flipping through dirs while the playback is loading data in the background
21:25:04amiconnThen I can see only 2 dirs instead of 4 dirs and 4 files in the root
21:25:06[IDC]DragonI have some test variants, if you like
21:25:24[IDC]Dragonbut it doesn't make too much sense
21:25:47[IDC]Dragonnon caching, seperate caches, debug output
21:25:53amiconnI did not yet manage to trigger such an effect on the player at all
21:26:15amiconnMaybe ata is too fast to catch the right moment...
21:27:09[IDC]Dragonyou could insert a panic in the transfer function, on read errors (caused by out of range ata sectors)
21:28:04[IDC]DragonI get such a read error when returning into the root for the first time, but the problem only shows at the 2nd time
21:29:05amiconnThis should not happen at all
21:33:15zeekoejust a suggestion: did you thought of mounting the mmc card in /mnt?
21:33:16 Join Stryke` [0] (
21:33:23zeekoemight that be easyer?
21:33:39zeekoeno checking on double names, etc
21:34:46zeekoehm, there could rise problems when creating playlists on pc though
21:36:06amiconnThe concept works like this, only that the mount point is /<MMC1> (yes I know that there are forbidden characters in the path, as Linux and Windows see it)
21:36:46amiconnThe name is in fact arbitrary
21:37:58zeekoeah ok
21:38:12 Join EGM92 [0] (
21:38:46EGM92Hey guys I was wondering I can't get onto the rockbox site but is there any support for multimedia players yet?
21:41:51 Quit EGM92 (Client Quit)
21:41:57Quel|awayafaik no
21:42:05Quel|awayups.. gone
21:42:09 Nick Quel|away is now known as quelsaruk (~kvirc@
21:42:56amiconn[IDC]Dragon: I do not get a panic on read when the partial root dir appears.
21:46:52[IDC]Dragonperhaps an in-range corruption
21:47:35amiconnI'll check the root dir by hand, with a hex editor. I suspect the partial dir is only the first sector
21:51:56amiconnUhh, now that looks strange!
21:52:13amiconnI did a disk dump of the first megabyte of the internal flash.
21:52:46amiconnThe partition should start on sector 32 (0x4000), and the config sector being located at sector 30 (0x3C00)
21:53:08amiconnI have sectors looking like FAT16 sectors from sector 6 on....
21:53:18 Join lImbus [0] (
21:53:28lImbushuhuu \o/
21:55:22amiconn[IDC]Dragon: These stray fat sectors do not occur in the reserved area though. Real FAT starts at sector 38 (addr 0x4C00)
21:56:57[IDC]Dragonmaybe some stray cache flush brought them there
21:59:23amiconnOk, cleaned them. I should check from time to time whether they reappear
22:08:06amiconnThe 2 displayed dirs are in fact located in one root dir sector, however, it is neither the first nor the last.
22:08:36amiconnIt's sector 5 (counting from 0)
22:25:58 Join [IDC]Dragon2 [0] (
22:29:17***Saving seen data "./dancer.seen"
22:31:49amiconnHow many dragons are there? ;-)
22:34:05amiconnI checked a bit more. The directory entries that are displayed are part of the 2nd pseudo cluster. The entries from the 1st pseudo cluster vanish if there is activity on the other volume at that time
22:34:42 Quit thegeek_ (Read error: 104 (Connection reset by peer))
22:34:51[IDC]Dragon2sounds like a plain stupid bug
22:35:09[IDC]Dragon2I'll try my medidation soon
22:35:43[IDC]Dragon2real life reqieres some attention still
22:35:57 Join thegeek [0] (
22:37:33 Join muz [0] (
22:37:48muzhey guys
22:38:09muzhows the iriver port going?
22:41:15 Quit zeekoe ()
22:44:18 Quit [IDC]Dragon (Read error: 110 (Connection timed out))
22:44:29 Nick [IDC]Dragon2 is now known as [IDC]Dragon (
22:49:24[IDC]Dragonnow I got some time to meditate
22:49:33[IDC]Dragonbut am already tired
22:49:55[IDC]DragonI tried the separate cache variant again
22:50:11[IDC]Dragonwith the same reproduceable error
22:51:28[IDC]Dragonwhich puzzles me, how can they collide?
22:51:51amiconnMaybe this doesn't have to do with the fat caching, but with opening both root dirs at once.
22:52:06amiconnThe root dirs don't use the fat...
22:52:56amiconnI'm currently looking at dir.c
22:52:59[IDC]Dragonwith no caching, it works
22:53:36[IDC]DragonI committed dir.c some minutes ago
22:54:06[IDC]Dragonwith my pionter update, no behaviour change
22:56:32muzare you guys talkin about the iriver?
22:58:50lImbusmuz: no, it's a bug they are hunting.
23:10:58 Quit muz ("CGI:IRC (EOF)")
23:13:09amiconn[IDC]Dragon: I'm trying to understand the fat caching. I wonder why the cache consists of 2 parts - a small struct array and the sector cache itself.
23:19:31 Quit [IDC]Dragon (Read error: 60 (Operation timed out))
23:25:21 Join [IDC]Dragon [0] (
23:25:28amiconnhi again ;)
23:25:38amiconnI found a potential race condition...
23:25:49[IDC]Dragontonight t-online dislikes *me*, for a change
23:26:09[IDC]Dragonspeak up
23:27:11[IDC]DragonI just wanted to say that the read errors I get are in legal range!
23:27:38[IDC]DragonI moved the panic out of the write, and it doesn't trigger
23:27:44amiconncache_fat_sector() first checks if the entry is free, and if yes, it reads the sectors, then marks the entry as inuse. The sector reading may yield. In case another thread decides to use the same entry, it still finds it marked as free...
23:28:16[IDC]Dragonbut instead, ata_read_sectors returns an error
23:28:48[IDC]Dragonyou're the man to answer that
23:28:50amiconnAh, you mean your read error observations
23:29:11amiconnWhat error code do you get?
23:29:28[IDC]Dragon(race condition) yes, we should mark first
23:29:52[IDC]Dragonhave to output the error code, hang on
23:36:09[IDC]Dragonerror -12
23:38:32amiconnHmm, that is strange. -12 means the card did not send a data block (in time) as a reply to the read command.
23:38:48amiconnIn which place do you get this error?
23:39:09[IDC]Dragonwhen returning to root for the 1st time
23:39:18[IDC]Dragonor, where in the code?
23:39:23amiconnIn the code.
23:39:36[IDC]Dragonwithin transfer()
23:41:53amiconnArgh! In ata_read_sectors, you change the global variable current_card before locking the mutex.
23:42:07[IDC]Dragonuh oh
23:42:11[IDC]DragonI blew it
23:42:43amiconnI intended to remove that global variable completely when implementing multivolume.
23:43:10[IDC]Dragonyes, I just saved the extra #if
23:43:27amiconnI don't know yet for sure whether this is critical or not
23:44:00[IDC]Dragonit's worth removing those race conditions
23:45:28amiconnYes. What I don't know for sure yet is whether this creates a race condition at all.
23:45:52[IDC]Dragonthe read panic comes from reading 256 sectors, this must be the mpeg thread
23:46:05amiconnread panic?
23:46:22[IDC]Dragonmy locally added one
23:46:32amiconnAh ok.
23:47:29[IDC]Dragonremember I have to start playback + go back to root to cause it
23:47:57[IDC]Dragonmixing mpeg and browser access
23:48:30amiconnYes, me too, plus the (partly) vanishing rootdir occurs only if the mpeg thread is reloading at the moment I go back to the root.
23:48:42amiconnHowever, the music continues playing...
23:49:10[IDC]Dragonmaybe it has "won"
23:49:24[IDC]Dragonor, a glitch is less apparent
23:49:43[IDC]Dragonand some minute in the future
23:52:32amiconnYou're right - it plays a wrong part (from a totally different song!) after a while
23:57:56amiconnI still don't get what may cause the ata_read_sectors() error. Although not completely clean, the current_card handling should do no harm.

Previous day | Next day