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 2008-12-12

00:00:07*gevaerts hears some weird artefacts while playing mp3 on the clip
00:00:46gevaertsLike a short bit of audio repeating every now and then
00:01:23Zagorgevaerts: which build is that?
00:02:09gevaertsZagor: clean svn, with tagcache, quickscreen and pitchscreen disabled
00:03:02Zagorinteresting. my mp3 artefact is that it is skipping over small parts (perhaps 50-100ms) every second or so
00:03:08Zagorand of course that it crashes :)
00:03:53funmangevaerts: perhaps the pcmbuffer isnt updated correctly then
00:04:59ZagorI'm having trouble fitting a c200 build in 2MB...
00:05:26gevaertsIt sounds like some sort of echo, so I'm pretty sure it's not skipping. This is a 24 kbps 16kHz mp3
00:05:48ZagorI have only tested with 44kHz ones
00:07:00funmangevaerts: oh i have only tested 22 and 44kHz, perhaps we are not precise enough for 16kHz (that must be checked at least)
00:08:33LloreanPaulJam: In the end, only the Shuffle, Repeat, and Crossfade tags are redundant, as far as I can see. What to do with them has been discussed a little, but no final decision or anything.
00:11:03 Join krazykit [0] (
00:12:51 Join DerDome [0] (
00:14:13 Quit DerDome (Client Quit)
00:15:56PaulJamLlorean: wow, i somehow had the impression that it were much more tags. i guess those 3 (or 4 if you count %cf too) duplicates don't matter too much.
00:20:12Zagorhmm? with MEMORYSIZE=2 the DRAM section only becomes 512KB. no wonder it's hard to fit!
00:22:48 Join LambdaCalculus37 [0] (n=rmenes@rockbox/staff/LambdaCalculus37)
00:22:58gevaertsYou need to optimize more!
00:23:32LloreanPaulJam: If you look carefully, there are one or two other tags that seem to be duplicates (for example, volume seems to be one) but there's pretty much always some special behaviour involved that you can't replace with the generic tag.
00:24:03Zagorahh, it's my fault. does not depend on config.h
00:24:31Zagorwhen setting plugin and codec buffers to the same as on clip, it fits
00:25:46 Join akur [0] (
00:25:52funmanZagor: for what is worth the clip use 64kb of IRAM for rockbox, 320-64 for codecs and the rest for rockbox plugins and audio
00:27:15Zagorit has a huge iram
00:28:17Lloreanfunman: Regarding the Clip, and how mine isn't playing back.
00:28:26LloreanIt wasn't really "crashing"
00:28:47LloreanI could FF/RW to different times, change the play/pause state and see the icon change, playback just never actually started. I was sitting silently on the WPS screen.
00:30:00Lloreanfunman: As a clue, the "View buffering thread" says "alloc: 1832/ 29(something offscreen)"
00:30:07LloreanCould it still be trying to buffer 32mb for some reason?
00:30:12 Quit BigBambi ("Please insert girder")
00:32:03ZagorLlorean: how much buffer do you have?
00:32:06funmanI cant tell
00:32:07 Nick fxb is now known as fxb__ (
00:32:26 Join BigBambi [0] (n=alex@rockbox/staff/BigBambi)
00:32:43LloreanZagor: Is there another screen to check, it's offscreen and won't scroll on this one
00:32:48LloreanThere's a bunch of extra spacing.
00:33:25ZagorLlorean: "Rockbox info" show the buffer size
00:33:32LloreanZagor: Buffer: 593kb
00:33:40 Quit moos ("Rockbox rules the DAP world")
00:33:42LloreanSo that 29 at the start of it is definitely indicating *something* wrong
00:33:54ZagorLlorean: right, that's too small. I got the same symptoms as you with that buffer size.
00:34:11funmanLlorean: you got a point there ..
00:34:20 Quit perrikwp (" ajax IRC Client")
00:35:01Zagorand the same symtom on c200 compiled for 2MB!
00:35:46Zagorhaha. "Buffer: 45KB"
00:36:00Zagorit needs a diet
00:36:35Bagder45 feels a bit on the tiny side ;-)
00:36:42***Saving seen data "./dancer.seen"
00:36:46Zagorjust a bit
00:38:09funmanZagor: hum, what if you enable the check in playback.c for pcmbuf_init() return value ?
00:38:27LloreanZagor: Out of curiosity, does yours show the right value in the debug->view buffering thread screen?
00:41:47Zagorrefreshing database...
00:42:05funmanthats not a right value I think
00:42:09 Quit massiveH ("Leaving")
00:43:03 Join itcheg [0] (i=62db4767@gateway/web/ajax/
00:43:05Zagorfunman: yeah then I get panic. 578688 > 306720
00:44:14funmantry to enlarge the RAM by a few hundreds of kilobytes then
00:44:28funmanI understand that 306720 is the audiobuffer size
00:45:03Zagorthat's audiobufend - filebuf
00:46:23 Quit akur (Read error: 110 (Connection timed out))
00:48:18funmanfilebuf is defined by buffering.c ?
00:49:54Zagorno, it's a local variable in audio_reset_buffer. it's (malloc_buf + 15) & ~15)
00:50:46 Quit shotofadds ("Leaving")
00:51:55 Quit ender1 (" Why geeks like computers: look chat date touch grep make unzip strip view finger mount fsck more fsck yes spray umount slee")
00:52:00Zagoroh, malloc_buf is also local :-)
00:52:48 Join perrikwp [0] (i=18ac0c41@gateway/web/ajax/
00:54:28 Join akur [0] (
00:55:52 Quit funman (" ajax IRC Client")
00:58:30 Quit faemir ("Leaving")
00:59:56 Quit jhulst (Remote closed the connection)
01:01:06 Quit bluebrother ("leaving")
01:04:25 Quit herrwaldo ("Konversation terminated!")
01:04:59 Quit Nico_P (Remote closed the connection)
01:05:12Zagoralloc: and usefl: in the buffer thread display shows completely bonkers values when playback fails.
01:05:29 Join Thundercloud [0] (
01:05:34Zagorusefl: stopped at -1684852 right now...
01:07:42Zagorand it does not always trigger an sd panic
01:07:46 Quit tessarakt ("Client exiting")
01:08:11 Quit akur ("Leaving.")
01:10:33 Quit LambdaCalculus37 ("Ka-chunka")
01:10:35 Quit MarcGuay (" ajax IRC Client")
01:10:50Zagorflash targets should save settings more often
01:13:29 Join akur [0] (
01:17:44pixelmaor more flash wear?
01:18:11Zagoryeah, because that is a real problem with flash players...
01:18:55Zagorrather, they don't need to delay updating the config file.
01:19:08pixelmayou never know
01:19:19 Quit jhMikeS (Nick collision from services.)
01:19:22Zagorthough of coure this is mostly a problem on targets that tend to crash a lot :-)
01:19:25 Join jhMikeS [50] (n=jethead7@rockbox/developer/jhMikeS)
01:19:43 Quit itcheg (" ajax IRC Client")
01:19:45 Quit culture (Read error: 110 (Connection timed out))
01:20:08 Quit bmbl ("Woah!")
01:21:54kugelrasher, Llorean: I've looked at sorting tags a bit, and I don't think it's a good idea to implement this "show artist, but internally sort by sortartst" thing
01:22:24Zagorflash wear is only a problem when you write physical prom sectors yourself. all consumer flash media does wear leveling.
01:22:38PaulJamFrom the AlbumArt wiki: "%Cl|50|70|c100|b100| : displays the found bitmap at position x=50, y=70. Smaller bitmaps are centered horizontally at the bottom of this rectangle. Bigger bitmaps are cropped to 100x100." isn't this wrong? or does the resizing only support upscaling?
01:23:10 Join itcheg [0] (i=62db4767@gateway/web/ajax/
01:23:29kugelPaulJam: probably just out-of-date.
01:23:48kugelI suppose you try it out (and update the page if you're at it)
01:23:52 Quit Zagor ("Clint excited")
01:24:19 Quit akur ("Leaving.")
01:24:22 Join akur [0] (
01:24:30pixelmareminds me of a problem I have with the scaling and wanted to ask Unhelpful about...
01:24:51PaulJamwell, i'm going to try it out, but i won't be able to update the wiki.
01:24:54Unhelpfulyou have great timing, i just got back ;)
01:25:16kugelPaulJam: no problem. I'll do it after you told me the result
01:25:28 Part akur
01:26:31UnhelpfulPaulJam: the scaler is aspect-ratio-preserving, so if the the image is a different AR from the tag, the align tags may still come into play. there's not a very nice way to explain that that's as simple is explaining it without scaling. :/
01:27:12PaulJamUnhelpful: i was mainly referring to the "Bigger bitmaps are cropped to 100x100." part
01:27:31Unhelpfulno, that doesn't happen anymore, or shouldn't, anyway.
01:28:21Unhelpfulthe "smaller bitmaps are ..." is also not strictly true. the bitmap is scaled to the largest size that will fit the specified constraints
01:29:45pixelmaUnhelpful: have a problem most likely related to the scaling on my M5 (also in the sim). I'm playing around with a new WPS which uses 100x100 AA, but since I used one with 50x50 AA before I already downscaled the pictures. Now with a build of last Monday, the picture was displayed smaller in the center of course, now displaying these pictures seem to fail completely in the WPS they would have to be upscaled (or just displayed as before) but in the
01:29:45pixelma other WPS where they fit they are displayed
01:29:47Lloreankugel: Why would it be a bad idea?
01:30:08kugelI've looked a bit into the code
01:31:15pixelmaUnhelpful: the sim tells me that it finds the file and gives me the following error message: "read_part_line: error reading image, read returned 2 expected 24", rev 19389
01:31:18kugeland going by that it'll be a) overly complex to implement (not only because tagcache code is already complex enough), b) be most likely considerable wasteful ram and/or performance wise
01:31:21Unhelpfulpixelma: it sounds like something is wrong with the nearest-neighbor scaler, then. i'll play aound in the sim and see if i can duplicate & fix it.
01:32:00kugeltagcache loads the appropriate files (database*.idx) based on the format tagnavi passes to it
01:32:16pixelmaUnhelpful: thanks :)
01:32:20kugelit takes the data and does both, display and sorting, out of it
01:32:51Unhelpfulpixelma: it's somehow read past the end of the image. i've seen a few different ways that can happen... i'm not sure how it's doing it now, though. :/
01:32:54kugelnow if we want to sort and display different data, we need to load both files into ram.
01:33:02Lloreankugel: Why don't you populate the "SortArtist" idx file with both Artist AND SortArtist tags (put the Artist tag there if SortArtist doesn't exist) and ALWAYS use it for sorting, and then ALWAYS use the artist .idx file for displaying?
01:33:24LloreanYou need to load both files, but you need to load multiple files for more complex filters too, don't you?
01:34:01kugeland, of course, introduce special cases for the sorting tags
01:34:39Lloreankugel: The alternative is having things not work if people don't have both sorting tags and artist tags on *all* their files, though?
01:35:02kugelwell, I thought about that, but that means you need to load a file which contains twice as much data as you actually need
01:35:43Lloreankugel: If no file has sortartist tags, don't generate the sortartist .idx file, and use only the one?
01:35:47kugelLlorean: I implemented fall back for files which doesn't have the sorting tags filled, that's not my concern
01:35:51 Quit Thundercloud (Remote closed the connection)
01:36:05LloreanThe whole point of sorting tags is that they're not *supposed* to be displayed.
01:36:33LloreanIf you're going to display them, you may as well just tell people "put what you want it sorted by in the Artist tag, because we're gonna display it anyway"
01:36:42kugeldo you think generating a idx dependant on all your music files is a good idea?
01:37:02LloreanDisplaying the sort tags basically makes the patch pointless, because it then has to display "Beatles, The" if it's to sort it as B, and removes the point of the patch entirely.
01:37:04kugelit's still displayed in the wps. It's only about the database browser
01:37:14Lloreankugel: Which is what people are complaining about...
01:37:48LloreanThey want it to show "The Beetles" in the list, but have it listed in the B section.
01:37:50kugelthe ram usage will baiscally double just for displaying "the"'s and "a"'s
01:38:11LloreanAnd the doubled RAM usage won't affect people who don't use the database.
01:39:23Unhelpfulwhat if the sortartist index is only populated with files where it differs from the artist tag?
01:40:31kugelwell, then you still need to check the sortartist idx file regulary. And that's even more complicated to implement
01:41:10kugelI mean, as long as you still need to read the separate file, then you can just put everything you need into it, imho
01:41:19LloreanWell, the sort artist patch doesn't actually solve the problem it's meant to solve unless the display issue is fixed.
01:41:28LloreanSo it probably won't go in at all if some way isn't found, I suspect.
01:41:42kugelyea, I understand that
01:42:02LloreanAnd I don't think more RAM usage is a big deal for most people who use database.
01:42:09LloreanSince those people can save a lot of RAM just by not using it at all.
01:42:33pixelmawell, you can discuss if that's a display issue. Thinking of a record shop... I often see "Beatles, The" there
01:43:16kugelpixelma: I wouldn't have a problem if it's displayed like that, especially since I made the tags like that, so it's at least not unexpected
01:43:39 Quit MethoS- (Remote closed the connection)
01:43:57LloreanBut the long historical tradition of feature requests it's meant to address are people who want to see "The Beatles" in the list, in the B section.
01:45:17Unhelpful...what about offering separate displayas/sortby options? if you just set one, it sorts and displays from one index. if you set both, it has to load both files, and you get "The Beatles" in B
01:45:55LloreanUnhelpful: Most configuration of database behaviour is actually handled by the tagnavi.config file.
01:46:47LloreanI'm not sure complicating it with more options would simplify anything, since you'd still need to code for the most complex behaviour.
01:47:06LloreanSo you still get "if they wish to use both tags, it takes more RAM"
01:47:08pixelmayeah, I know. My point was only that you can see it two ways. Btw. *I* would be surprised to see something starting with "T" between other "B"s, but then I'm not one who would use it - at least when it's about articles, maybe with names like "Elvis Presley" or so)
01:47:13LloreanAnd if they just wish to use one tag, there's no need for an option.
01:47:42Unhelpfuland that's where the decision could be made. a generic displayas/sortby facility could also extend to the albumsort/albumartistsort tags just by putting those tags in the DB
01:47:52PaulJamUnhelpful: is it expected, that resizing doesn't seem to work when the filename of the bmp contains the dimensions (e.g. cover.75x75.bmp)? when i rename the bmp to cover.bmp it works like expected.
01:48:19LloreanUnhelpful: I think the decision could be 'made' by whether or not they included the tags in their files, rather than having them also have to adjust their tagnavi about it.
01:48:19UnhelpfulPaulJam: are those dimensions the ones of the current WPS?
01:48:31PaulJamUnhelpful: no
01:48:32LloreanUnhelpful: I think it should be 'transparent', if they've included sort tags, they want to use them.
01:48:46pixelma...then I'd expect it to be written Presley, Elvis"... ok, I'm not in that group at all
01:48:56 Part toffe82
01:49:29Unhelpfulpixelma: just so it's not done how the karma did "library sort", and put "Die Happy!" under H
01:49:33LloreanPaulJam: How would it know to look for cover.75x75.bmp if the WPS doesn't specify 75x75 cover art? It still has the exact same old search pattern.
01:50:23Unhelpfulexactly, the search is for <name>.<wpsAAdimensions>.bmp and then <name>.bmp
01:50:31 Quit robin0800 (Read error: 60 (Operation timed out))
01:50:46PaulJamLlorean: true, maybe this should be mentioned on the wiki (if it isn't already)
01:51:02Unhelpfulit can't reasonably search for each variation that might exist without loading the whole directory and pattern-matching the filenames
01:51:04LloreanPaulJam: Are we supposed to post somewhere "Yeah, the information already up is still correct" or what?
01:51:16kugelmaybe Slasheri pops up sometime, I think he can estimate best how much rewrite and stuff tagcache needs to support such a feature
01:52:04Unhelpfulpixelma: you wouldn't have a link to a 100x100 AA WPS for m5, would you?
01:52:36kugelI found the part where it sorts, and it sorts exactly the same variable which it displays, i.e. the result of the tagcache search
01:53:27pixelmaUnhelpful: well in that case the file would have no sorting tag at all. What I mean is just the use of the sorting tag for displaying in the list.
01:53:34kugeland the sorting function cannot take more arguments (iiuc), so there's no way to tell it to sort artist or sortartist based on whatever condition
01:53:52Unhelpful... it's horribly broken even w/ 50x50 file and cabbiev2 :/
01:54:33pixelmacabbiev2 uses 64x64 as AA size on those displays, just looked
01:54:39JdGordonthe sort comparer can use a global, or global_status to do it
01:54:46JdGordonlike the comparer for the file tree
01:55:01kugelah yea, those globals, right :)
01:55:12Unhelpfulpixelma: right, it looks like upscaling 50x50->64x64 corrupts the image :/
01:55:52Unhelpful32x32 -> 64x64 gets the same load error as you had, with it reading past EOF :D
01:56:05 Join XxJoshXx [0] (n=cf458923@gateway/web/cgi-irc/
01:56:43kugelbut it basically needs the search functions, the display functions, the sort functions, the parse functions and more to be (partly) rewritten
01:56:45pixelmaok, so I don't need to put my WPS up somewhere... good as it's still work in progress
01:57:20kugelsounds like fun
01:57:32 Join darknessxsurroun [0] (n=04f4b7bd@gateway/web/cgi-irc/
01:57:49 Join massiveH [0] (
01:57:56Lloreankugel: Probably why there's not a patch for it that's under any discussion for commit yet. :)
01:58:06LloreanIt's one of those 'big' projects.
01:58:10Unhelpfulpixelma: no, you don't. there seems to be some goof it scale_nearest for at least 2^N exact multiples, and possibly for others - i'll look for an M5 WPS that's not a 2^N size and see if i can break that.
01:58:12JdGordonkugel: or rework database internally so it needs less work for this...
01:59:12darknessxsurrouni need some help with replacing the firmware on the Philips GoGear Audio MP3 Player (SA30XX) is there a program availble besides the one from the main website(
01:59:48kugelI fail to see where it needs rework (apart from this sorting tag thing), besides that the code is a bit complex, it performs fast and relatively reliable (at least for me, I've heard of some weird bugs)
01:59:48Lloreandarknessxsurroun: This doesn't really sound like a Rockbox question.
01:59:49LloreanWhat are you trying to do, exactly?
02:00:14darknessxsurrounwell i want to replace the firmware with a rockbox made one
02:00:25darknessxsurrounsorry if i'm not explaining right
02:00:38*kugel didn't know SA30XX is a supported target :o
02:00:58darknessxsurrounits not supported?
02:01:02kugeloh, it isn't
02:01:07 Quit XxJoshXx ("CGI:IRC (Ping timeout)")
02:01:09Lloreandarknessxsurroun: The list of supported players is on the front page.
02:01:16kugeldarknessxsurroun: No, the front page clearly tells which targets are supported
02:01:34Unhelpfuldarknessxsurroun: if there's nothing on the website or forums about the player, chances are there is no rockbox for it. and if it's not on the front page, but you see it somewhere else, it's probably not very *well* supported.
02:01:56LloreanThere has been some work on the SA31XX GoGears, I believe, but it's all fairly early development work, and mostly just sitting around for some time.
02:02:16darknessxsurrounoh, cause the default firmware for the mp3 player is dead slow, and i was looking for a replacement
02:03:02darknessxsurrounby means. the browsing is slow on the mp3> searching or scrolling for songs is slow....
02:03:18Unhelpfulyou're going to need a different player if you want to use rockbox. the e200-series players are pretty cheap on, most of the time.
02:03:39darknessxsurrounhm... alright thanks.
02:06:07 Quit darknessxsurroun ("CGI:IRC (EOF)")
02:07:35 Quit lasser (Read error: 110 (Connection timed out))
02:07:45PaulJamUnhelpful: one last question: are there any restrictions on the size of the cover.bmp?
02:11:02UnhelpfulPaulJam: can't scale up from smaller than 2x2, or down to smaller than 2x2. scaling up to, or down from, a size with more than 16843009 pixels (4104x4104, if square) will silently corrupt the scaled output. i'm adding tests to catch the limits at both ends, and fail the load, but i don't expect them to be hit too often
02:12:04Unhelpfula 24-bit bitmap file in the max size is about 50MB, i don't think many people will have one
02:15:07 Quit kugel ("ChatZilla 0.9.84 [Firefox 3.0.4/2008102920]")
02:32:10 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
02:33:11 Quit massiveH ("Leaving")
02:36:43***Saving seen data "./dancer.seen"
02:46:09 Quit _lifeless (Remote closed the connection)
02:54:02 Quit HellDragon (Remote closed the connection)
02:55:38 Join HellDragon [0] (
02:55:58 Quit ajonat (Read error: 54 (Connection reset by peer))
02:56:41 Join ajonat [0] (n=ajonat@
02:57:34 Join perrikwp|lab [0] (i=98214d2a@gateway/web/ajax/
03:04:21 Join BHSPitMonkey [0] (n=stephen@unaffiliated/bhspitmonkey)
03:09:57 Quit jhulst (Read error: 110 (Connection timed out))
03:10:18 Quit avacore (Read error: 110 (Connection timed out))
03:10:48 Quit ajonat ()
03:15:14 Join avacore [0] (
03:15:23 Join CaptainKewl [0] (
03:18:13 Join fdinel [0] (
03:19:31 Quit martian67 ("gone")
03:20:20 Quit XavierGr (Read error: 54 (Connection reset by peer))
03:21:24 Join martian67 [0] (i=lol3izer@about/linux/regular/martian67)
03:24:58Unhelpfulgreyscale upscaling was broken, should i push the fix, or open an FS task first? also, should i push the other minor enhancements in my tree with it, or only the thing that actually fixes a bug?
03:27:06JdGordonone change at a time...
03:27:09JdGordonbut yeah push the fix
03:40:35 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
03:41:15 Join hillshum [0] (
03:44:35 Join advcomp2019_ [0] (n=advcomp2@unaffiliated/advcomp2019)
03:46:24 Quit advcomp2019 (Nick collision from services.)
03:46:25 Nick advcomp2019_ is now known as advcomp2019 (n=advcomp2@unaffiliated/advcomp2019)
04:12:15 Join blkhawk- [0] (
04:13:13 Quit Acky (Read error: 104 (Connection reset by peer))
04:17:47 Join Acksaw [0] (
04:24:07 Quit parafin (
04:24:07 Quit tvelocity (
04:24:07 Quit shodanX (
04:24:07 Quit preglow (
04:24:07 Quit killan (
04:24:07 Quit DaCapn (
04:24:07 Quit Zambezi (
04:24:07 Quit FOAD (
04:24:07 Quit tim__b (
04:25:18 Nick Bensawsome is now known as NinJew (n=Bensawso@unaffiliated/bensawsome)
04:25:46NJoinkillan [0] (
04:25:46NJointvelocity [0] (
04:25:46NJoinparafin [0] (
04:25:46NJoinFOAD [0] (
04:25:46NJoinpreglow [0] (
04:25:46NJoinDaCapn [0] (
04:25:46NJoinshodanX [0] (
04:25:46NJoinZambezi [0] (
04:25:46NJointim__b [0] (
04:29:44 Quit blkhawk (Read error: 110 (Connection timed out))
04:30:14 Nick blkhawk- is now known as blkhawk (
04:30:41 Join Kakashi666 [0] (
04:30:52Kakashi666hey all
04:32:04Kakashi666I'm having an issue with my Flyspray account, but Zagor is not online?
04:33:15Kakashi666I guess I'll try him by email, noone seems to be awake
04:33:19 Quit jfc (Read error: 54 (Connection reset by peer))
04:34:01 Join jfc [0] (
04:34:37 Part Kakashi666
04:36:13 Quit perrikwp|lab (" ajax IRC Client")
04:36:45***Saving seen data "./dancer.seen"
04:41:58 Quit jfc (Read error: 54 (Connection reset by peer))
04:45:44 Quit itcheg (" ajax IRC Client")
04:46:05 Nick NinJew is now known as Bensawsome (n=Bensawso@unaffiliated/bensawsome)
04:50:09 Quit hillshum ("Leaving")
05:00:32 Join itcheg [0] (i=62db4767@gateway/web/ajax/
05:08:24 Quit miepchen^schlaf (Read error: 110 (Connection timed out))
05:11:05 Join massiveH [0] (
05:21:51 Join FOAD_ [0] (
05:22:00Unhelpfulre: FS #7253, i am unable to "break" the dircache with the procedure steve describes, although shutting down while the disk is active just after startup seems to take much longer than if the disk has stopped - possibly any dircache updates are waited for on shutdown, now? it's also pretty old, filed 2007/06 and last updated 2008/03.
05:23:29LloreanMaybe leave a comment on it that you can't reproduce.
05:24:42Unhelpfulseems sensible. i'm wondering if the bug has been "fixed" by some other random work since then... a lot changes in a year.
05:26:51Unhelpfulmaybe bug some H3x0 users to try to reproduce, but it seems unlikely this would be target-specific
05:28:16LloreanWell, pondlife is still around from time to time. I imagine a comment on that is likely to get a response, and I don't think it's too harmful to let it sit a bit longer.
05:28:37LloreanNot that there's likely any harm closing it either, with a reason "Seems to be fixed, file again if this is incorrect" or something
05:34:16 Join Darksair [0] (n=user@
05:36:16 Quit FOAD (Read error: 110 (Connection timed out))
05:36:16 Nick FOAD_ is now known as FOAD (
05:37:12 Quit itcheg (" ajax IRC Client")
05:39:52 Quit Horscht ("Snak 5.3.3 Unregistered copy. Evaluation period is over. Program will now quit. Thanks for using Snak.")
05:52:49Unhelpfulhow hard would it be to get a binary for a particular target and revision built on a particular buildserver? i'm kind of curious as to whether might help determine what causes the server-to-server binsize fluctuations that seem to have come in with resize-on-load.
05:53:16Unhelpfulalso, best place to put a possibly useful tool? ;)
05:55:17 Join pixelma_ [0] (n=pixelma@rockbox/staff/pixelma)
05:56:10 Quit amiconn (Read error: 110 (Connection timed out))
05:56:22 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
05:56:35 Quit massiveH ("Leaving")
05:57:22 Quit pixelma (Read error: 110 (Connection timed out))
06:04:13 Quit Aurix_Lexico (Read error: 110 (Connection timed out))
06:04:52JdGordonUnhelpful: you need to speak to Bagder about that...
06:17:41 Quit HBK (Read error: 60 (Operation timed out))
06:19:32 Quit obo (Read error: 110 (Connection timed out))
06:20:56 Join Darksair` [0] (n=user@
06:29:29 Join toffe82 [0] (n=chatzill@
06:36:46***Saving seen data "./dancer.seen"
06:37:01 Join HBK [0] (
06:40:19 Quit Darksair (Connection timed out)
06:49:31 Quit daurnimator ("Cyas later...")
06:49:56toffe82I still have a problem with the install on an ipod 5.5 80 gb with a samsung hd
06:51:09toffe82it is 4k per sector, ipodpatche say it is a 2048 per sector, Lamdacalculus37 made me a version with the ata driver modified but it doesn't work
06:51:50toffe82I still have the same error, panic error sector 4096 or something like this
06:51:57toffe82anbody has an idea ?
06:59:28 Join Darksair [0] (n=user@
07:05:29 Join daurnimator [0] (
07:06:46 Join J-23 [0] (
07:08:03 Join Default_ [0] (
07:08:11 Nick Default_ is now known as slact (
07:08:14 Quit slact (Client Quit)
07:08:27 Join slact [0] (
07:08:44slacthi Rockbox. I've an ipodish question.
07:09:01slactmost of the time wheel response is kind of sluggish.
07:09:24slact...except when music is currently playing. then it seems a lot more responsive
07:09:29slact(most of the time)
07:10:06slactthis is for the latest svn build, but is the case for 3.0 as well several other builds that i've tried. i've got a 5.5G 80gb model
07:10:29 Join daurn [0] (n=daurnima@unaffiliated/daurnimator)
07:12:39Lloreanslact: This is normal behaviour.
07:12:53slact: (
07:13:03LloreanThe CPU is run pretty slow most of the time. When music is playing, it pops between the slow speed and the fast speed. When it's at the slow speed, the UI may be somewhat sluggish.
07:13:25LloreanThis improves battery life quite a bit.
07:13:52slacti see. that explains why it unsluggishes during disk activity, too
07:15:00slactis the general screen refresh rate slow when the CPU is slowed down, or is it just because of the processing for the keypresses?
07:15:29LloreanEverything is slower, and the music has higher priority than the user interface.
07:15:57 Quit Darksair` (Connection timed out)
07:16:46slactnow you're making me want to go code-digging.
07:17:19slact... 'cause now i naively think it might be a good idea to speed up the cpu when it detects keypresses
07:17:36LloreanThere's a patch to do that in the tracker already, I think
07:17:41LloreanI couldn't tell you which task it is.
07:17:49slacti'll go have a look
07:17:53 Quit reacocard ("AUGH")
07:18:05 Join reacocard [0] (
07:18:41slactis that it? 'cause that's nbot really what i'm talking about
07:18:59LloreanThat's not a patch.
07:19:19Unhelpfulthis is the one:
07:19:30Unhelpfulboost CPU for 1s after each button press
07:19:46 Quit Darksair (Connection timed out)
07:20:19slactthanks, unhelpful.
07:20:44Unhelpfulit looks like it also reduces the "normal" speed that is used when not boosted
07:21:13 Join Darksair [0] (n=user@
07:21:52 Join advcomp2019_ [0] (n=advcomp2@unaffiliated/advcomp2019)
07:22:02 Quit advcomp2019 (Nick collision from services.)
07:22:06 Nick advcomp2019_ is now known as advcomp2019 (n=advcomp2@unaffiliated/advcomp2019)
07:23:01Unhelpfulit looks like there's some work to do on it, regarding limiting the boost to appropriate contexts - apparently many of the games and demos use cpu-speed-sensitive timing methods, or no timing at all, and are quite jerky with the popping in and out of boost every time a button is pressed.
07:24:12slactguess i'll want to dig through the rockbox code after all
07:25:04Unhelpfuli can't say how hard this job might be - it could very well be trivial. be sure to put up a new patch if you get anywhere with it.
07:26:07slactwill do
07:29:21 Join MethoS- [0] (
07:53:30 Quit BHSPitMonkey (Remote closed the connection)
07:55:16 Join funman [0] (n=fun@rockbox/developer/funman)
08:05:54 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
08:11:28 Join bertrik [0] (
08:17:08 Join gromit`` [0] (
08:19:27amiconngcc is weird
08:20:23*amiconn checked why ata.o pulls __ashrsi3 on SH1, even though there is no (obvious) variable right shift in ata.c
08:21:59amiconnIt translates a check in set_features(): if (identify_info[features[i].id_word] & (1 << features[i].id_bit)) becomes if ((identify_info[features[i].id_word] >> features[i].id_bit) & 1)
08:22:41amiconnI don't understand why it uses an arithmetic shift though. identify_info is an array of unsigned short.
08:23:50 Quit Seed ("cu, Andre")
08:24:12Unhelpfulthat's bizarre. you want to get that binsize gain on recorder down to ondio's size, eh? ;)
08:24:15*amiconn tries sth
08:24:25amiconnyep :)
08:24:33*funman gives a try at -Os
08:24:36 Quit BigBambi (Read error: 113 (No route to host))
08:24:57Unhelpfuli saved my quick python hack into a file with prettified output. haven't decided where to put it yet, but it seemed possibly useful.
08:24:58 Quit daurn (Read error: 113 (No route to host))
08:25:40 Join daurn [0] (n=daurnima@unaffiliated/daurnimator)
08:27:21 Quit gromit` (Read error: 110 (Connection timed out))
08:28:04 Quit CaptainKewl (Read error: 110 (Connection timed out))
08:30:46amiconnUnhelpful: Using an unsigned '1' (1u) makes gcc switch to __lsrshi3 (logical right shift by n)
08:31:10funman~30kB saved on binsize, same on RAM usage
08:31:41Unhelpfulamiconn: but, bitwise operations are sign-agnostic? :/
08:31:50toffe82amiconn: it works with the first change
08:32:52amiconntoffe82: Try the other method anyway, if it works it's better (drive firmware will then handle partial sectors instead of rockbox)
08:34:19amiconnIf it does work, we need to think about detecting this capability. Afaik the ata standard says nothing about it - we either need to probe, or check for specific disk models...
08:34:44amiconnUnhelpful: Shifts are not.
08:34:45 Join Rob2222 [0] (
08:36:02amiconnLooks like gcc wants to play safe here. It wouldn't need to since for left shifts, logical == arithmetical, so it could just use a logical right shift even if the '1
08:36:09amiconn' for testing is signed.
08:36:49***Saving seen data "./dancer.seen"
08:37:00Unhelpfulthat's very odd. but now we know :D
08:37:35*amiconn will try this build un his recorder just to make sure, then commit this tiny change
08:37:46amiconnIt's just 1 ==> 1u
08:39:33toffe82amiconn: it is working
08:40:49Unhelpfulstill best to test. now one of us just needs to figure out how to get the rest of it back, preferably without forking much of the bmp reader code.
08:41:09amiconnShould be a little faster for small sectors. If you have the time, run test_disk with both core variants
08:42:04amiconn(@ toffe82)
08:44:22Unhelpfuli got it down by 80B or so, to get it to where it is now, by making read_part_line on mono read whole lines, and bmp_read_fd call it unconditionally once per line. i'm guessing a good chunk of the change that's left is due to packing various bits of data into a struct, and calling the function that unpacks it?
08:46:20toffe82amiconn: where is test_disk ?
08:46:40GodEatertoffe82: it's another thing you'll have to enable in the build manually
08:47:01 Join LinusN [0] (n=linus@gateway/web/cgi-irc/
08:47:13toffe82I will do it another day , time to sleep
08:47:23toffe82thank you
08:47:42*GodEater must get round to looking at how that's done with the new makefile
08:48:09 Part toffe82
08:48:18funmanedit apps/plugins/SOURCES (like old Makefile I think)
08:48:44GodEaterso it is ;)
08:49:45 Nick pixelma_ is now known as pixelma (n=pixelma@rockbox/staff/pixelma)
08:50:15 Join ender` [0] (
08:50:31funmanhum with a -Os build some buttons seem to not respond ..
08:52:06 Join Bagderr- [0] (n=daniel@gateway/web/cgi-irc/
08:52:50 Quit Rob2223 (Read error: 110 (Connection timed out))
08:54:57amiconnHmm, the Player seems to use __ashrsi3 somewhere else as well
08:56:23 Quit bertrik ("Leaving")
08:57:20funmanI wonder why playback.c define target specific CODEC_IRAM_(ORIGIN,SIZE) and doesn't use them ..
08:58:18funmanI see no include following the definition, so I think they should just go away, but I prefer to ask
08:58:23 Join kugel [0] (n=chatzill@unaffiliated/kugel)
09:00:11Unhelpfulamiconn: odd, that's the only place outside of plugins i remember seeing it.. :/
09:01:35funmanjust seem a leftover, amiconn could you check that i'm not misforgetting something here?
09:03:37 Join pondlife [50] (n=Steve@rockbox/developer/pondlife)
09:05:02Unhelpfulrechecked, i still only see it in firmware/drivers/ata.o :/
09:05:12pondlifeFor the logs: Could amiconn's neat "1u << x" trick be used in elsewhere - dsp.c, eq.c and replaygain.c do a few "1 << x"s.
09:05:23 Nick Bagderr- is now known as B4gder (n=daniel@gateway/web/cgi-irc/
09:05:54pondlifeOr would that hit performance?
09:05:57funmanpondlife: I understand that the right shift when using "& (1 << x)"
09:06:25funmani.e. it transforms to "( .. >> x) & 1"
09:07:10pondlifeAll aimconn did was specify that 1 was unsigned
09:07:53pondlifeSaved quite some SH1 binsize
09:07:59Unhelpfulwhich converts an arithmetic right shift to a logical on, in gcc's transformed version of the expression.
09:08:03 Join petur [50] (n=petur@rockbox/developer/petur)
09:08:26funmanbecause SH1 has no right shift so it uses a gcc function instead, i think other targets do have right shift?
09:08:27pondlifeOf course it might (a) not help other architectures or (b) hurt performance in the general case...
09:08:28amiconnpondlife: On coldfire and arm it doesn't matter at all
09:08:41pondlifeOK, that was the answer I was after
09:09:01amiconnSH1 has no variable shift instructions, only fixed ones, so variable shifts are subroutines on it
09:09:44amiconnBut even with those subroutines, logical shifts are a little better, because the subroutines for variable logical shifts are faster
09:10:35Unhelpfuli wasn't looking for common symbols of the same size - i'm assuming there's a logical shift right routine that's probably already used in many places?
09:10:49amiconnThis is because logical shifts by 1, 2, 8 and 16 are available, while arithmetic right shift is only availabe by 1 (and you can do by-16 using 2 instructions, and by-24 using 3 insns)
09:11:24amiconnYes there is - __lshrsi3
09:11:45 Join Zagor [0] (n=bjorn@rockbox/developer/Zagor)
09:12:01pixelmaUnhelpful: thanks for the fix, will have to try it out. And then find out if scrolling in conditional viewports is still broken and since when (must be something from last week too)
09:12:33pixelmawell or already fixed
09:13:08Unhelpfulpixelma: no problem, but i'm not really sure how i missed it before - i'm certain some of the test images i used in sliding_puzzle were smaller than display size
09:13:19amiconnUnhelpful: You probably didn't touch the Player, as it's charcell - no bmp scaler necessary :)
09:13:46 Join Darksair` [0] (n=user@
09:14:09Unhelpfulamiconn: it doesn't build the loader, or have album art, so i'm pretty sure anything i touched is ifdef'd out
09:14:11amiconnI'll check the .map and dig down the place(s) where it's used. It's a small optimisation, but still a good thing to do imho
09:14:45Unhelpfulyay, macro-ized version of at least GET_FB_WIDTH works
09:15:57 Quit Darksair (Connection timed out)
09:17:03amiconnfunman: Btw, while I think this optimisation (using a right shift for preparing the test) is quite clever, gcc goof directly after that on SH1:
09:18:56amiconnIt uses and #1, r0; tst r0, r0 instead of reducing that to tst #1, r0
09:19:06pixelmaUnhelpful: ah works... and scrolling works again too :)
09:19:17Unhelpfulpixelma: i doubt i fixed *that*
09:20:09amiconn('tst' is just an 'and' that sets the flag instead of storing the result. Normal alu instructions don't touch the flag on SH)
09:22:19pixelmaUnhelpful: and doesn't look too bad. Well it was in the same WPS maybe something got "confused" when scaling didn't work as I noticed it falls back to displaying viewports for the non-AA case even when AA was present but couldn't be displayed) who knows
09:24:22Unhelpfuli only tested with "extreme" upscaling, which makes it very obvious that it's working, because, HUGE GREY BLOCKS. :D
09:25:19*pixelma runs away
09:27:22funmanZagor: with SVN I see playback starts and hangs immediately on the Clip; by looking at the buffering thread debug screen. I'll try without tagcache since it let you play at least (sometimes)
09:27:35Unhelpfuli also found out that i needed to special-case scaling up by exactly one pixel, and i believe we got a free binsize reduction on greyscale targets, as well.
09:27:39funmanI noticed the audiobuffer is a bit below 1MB with tagcache enabled
09:28:00Zagorfunman: are you still using the direct-from-flash patch?
09:28:14Zagorah, "with SVN".
09:28:36 Quit JdGordon (Remote closed the connection)
09:28:44funmanno, I'm not using it because I want to reproduce problems, not avoid them ;o)
09:29:04kugelfunman: that's exactly the issue I got on my fuze as well (using that startscreen+.playlist_control+nvram.bin trick), which has a planty more ram
09:29:12Unhelpfulum, does the new build system even work on git that's not git-svn? i seem to recall that it calls git-svn info to get the svn revision...
09:29:43funmankugel: and did disabling tagcache help ? what's the size of audiobuffer ? (look in
09:30:11funmanUnhelpful: how can you get the svn revision on 'git that's not git-svn' ?
09:31:12Unhelpfulfunman: no clue how, since git doesn't do keyword subs, either...
09:32:11funmanthen it will work but report revision0 (like on a modified git tree)
09:32:22kugelfunman: with svn (only my lcd fixes applied) audiobuf is 0x54DFCC (5 562316)
09:32:41funmanperhaps the trick should be to use git log and grep magic as well to get the rev
09:32:48funmankugel: hm can you paste your map?
09:33:13Unhelpfulhrm, the cloned git log *should* have the svn revs in the commits, right?
09:33:37*kugel thought funman doesn't read so hugh files
09:34:15funmanUnhelpful: yes it has, but it's just a bit harder to extract than using git svn info
09:34:43funmanZagor: disabling tagcache highers filebuflen from ~28k to ~256k and indeed there is working playback
09:35:37funmannow I check the buffering thread stats and wait for a crash ! :)
09:36:00ZagorUnhelpful: git-svn-id: svn://
09:36:39Lloreanfunman: For me, I never had tagcache enabled in the first place, and didn't have working playback.
09:36:41Zagorthat's what git log says on the server at least
09:36:45Unhelpfulwould git log + grep be faster or slow in general than using git-svn, i wonder?
09:37:08funmanUnhelpful: also you (do you plan to fix it ?) should check if rev from git svn info and the rev from git log to see if the revision has been modified (and append a M to the revision)
09:37:20funmanLlorean: did you check audiobuffer and filebuffer sizes?
09:37:27funmanUnhelpful: not important ;)
09:38:12Unhelpfulfunman: that'll be tricky, if we don't import some form of .gitignore info into the actual svn repo... :/
09:38:21Unhelpfulalso, i volunteered? ;)
09:38:32Lloreanfunman: What values do you refer to, on what screen? Do you mean the PCM and compressed buffers?
09:39:21funmanUnhelpful: i can do that if you want
09:39:58funmanLlorean: buffering thread debug screen to check filebuf, and to check audiobuffer (it seems that playback doesn't start if it's below 1MB)
09:40:46Lloreanfunman: There's not a value on that screen called "filebuff"
09:41:07funmanLlorean: alloc and usefl
09:41:30kugelfunman: well, what could be the reason the issue is also apparent on fuze?
09:41:44funmani had to edit debug_menu.c to make the values fit on screen
09:41:46Lloreanfunman: I can't see those values. They're off the right side of the screen. Remember, I mentioned they started with 29 earlier?
09:41:48funmankugel: no clue
09:42:00Zagorfilebuflen is 284KB for me and playback works
09:42:01Unhelpfulfunman: i'll take a look, but i need get my current work commitable and switch to another branch first.
09:42:06Zagorwell, starts
09:42:14funmanLlorean: the important is how many digits they have :/ it could be 29k or 290k
09:42:54Zagorfunman: well you see that from the left-side value. if that has six digits, so does the right-side
09:42:57funmanLlorean: i removed the 08 from the format specifier, and used 4/5 chars for the description
09:43:55funmanZagor: do you have change directory enabled?
09:44:09Zagorwhen playback works, filebufused() and useful_data both hover around 240-280 KB. when it fails, it goes much much lower. 67KB now.
09:44:45Zagorfunman: no
09:44:47funmanlike, instantly ? or does fail after having dropped until ~67k?
09:45:10Zagor"instantly". I can't tell which happens first.
09:45:34Lloreanfunman: Well, if Zagor's right, it's four digits. so 29XX
09:45:36funmanI see arough 10kB per second decrease
09:45:56funmanLlorean: definitely too low, I hadn't playback working with 29kb
09:46:06Llorean2.9KB rather?
09:46:27Unhelpfuli think it might take an unless grep or sed has a first-match-only option that i'm not remember?
09:46:41funmanUnhelpful: cut?
09:46:51Lloreanfunman: I'm confused. The value is not in Bytes?
09:47:06Zagora strange thing is that usefl: varies with codec. the wav file shows values in a large range (50-280K) while mp3 is always >240KB
09:47:14funmanLlorean: yes; i meant I had 29kb in my build and no playback, so your 2.9kb is definitely too low
09:47:22 Quit jhulst (Remote closed the connection)
09:47:32amiconnZagor: Different watermarks, I'd guess
09:47:35Unhelpfulfunman: no, no, i think head is the thing... and when it gets its one line and closes stdin, it should stop grep and git-log
09:48:08Zagoramiconn: could be. but why does mp3 use such a high watermark?
09:48:18*amiconn thinks that watermark handling in swcodec playback needs a thorough makeover
09:48:29funmangrep -o only keeps the matching regex
09:48:46amiconnIt's a thing I am complaining about almost since the beginning of swcodec playback...
09:48:52Lloreanfunman: Why would mine be so low with a clean SVN build and default settings?
09:48:55Zagornow mp3 changed, and decresed 16KB/sec as expected
09:49:10Zagorlots of funny symptoms...
09:49:10LloreanOr better yet, what file is that screen in, so I can fix formatting and get the values displayed, to be sure?
09:49:10funmanLlorean: I have no clue, i only begin to read that code ^^
09:49:27LloreanSince it's not even filling completely
09:49:39LloreanNeither the PCM buffer nor the compressed buffers fill.
09:49:53Lloreanfunman: Also, for audiobuff, to I just subtract audiobuff_end from audiobuff in the map and give you that?
09:50:15funmanhum playback stopped itself (but status remains on pause) and I see: pcm buffer empty, alloc/usefl buffer nearly full, track count 0, handle count 1
09:50:30kugelLlorean: yes
09:50:35funmanLlorean: yep, just to check if >1MB gives playback while <1 doesn't
09:51:19funmankugel: i don't read so huge files, just grep the minimal info i need from it ;)
09:51:27 Join Darksair [0] (n=user@
09:51:49kugelfunman: what info you need? I told you the audiobuf size before
09:51:53Unhelpfulgot it: git log | sed -nre '/git-svn-id/ s|.*@([0-9]+) .*|\1|p' | head -n 1
09:52:06Zagorwhat is data_rem? it shows a bit over 6MB here!
09:52:09funmankugel: yeah; and I read 500kB, no 5.5MB .. so i wanted to check what was wrong
09:52:10Lloreanfunman: 990,352 in dec
09:52:18funmanZagor: data remaining to read in the file
09:52:26kugelfunman: no, it's 5,5MB
09:53:01funmanLlorean: what if you try to remove some code to get it over 1MB (reducing the plugin buffer for example, and ignoring plugins not building) ?
09:53:06kugelfunman: 5 562316 (there's a space after the first 5, maybe the number got split at your screen)
09:53:32funmanno it wasn't split but i overlooked the first 5, sorry for that
09:53:35Unhelpfulit runs a tiny bit faster than git svn info here, too
09:53:37 Quit kachna (Read error: 113 (No route to host))
09:54:06funmanUnhelpful: how does that work on osx (i mean without gnu sed) ?
09:54:28Lloreanfunman: Ah, it is 29k, not 2.9k on mine, btw. Just double checked
09:54:35Unhelpfulthat would be bsd sed? i don't know which options work with it...
09:54:57funmanLlorean: and did you have disabled tagcache intentionally ?
09:55:16funmanUnhelpful: mentioned in posix spec, i'll check that
09:55:37funmanoops the spec is on the dead laptop, but there is freebsd manpages
09:55:48 Quit Darksair (Client Quit)
09:56:04 Quit Darksair` (Client Quit)
09:56:18Lloreanfunman: I never enabled it to begin with
09:56:20Unhelpfulfunman: found the mac os x sed man page online, it needs -E for the -r, otherwise we should be good.
09:56:27funmanUnhelpful: uyup
09:56:58funmanLlorean: it's enabled by default since r19382
09:57:34Unhelpfuli can't find this, though, that the revlog says does the work
09:57:34Zagorpcm buffer seems a bit overkilled. it uses ~50 of its' 500 KB.
09:57:45Unhelpfulnevermind, i was in my build dir :/
09:57:51funmanUnhelpful: in tools/
09:58:19Zagornow I get a full mp3 playback, without skips
09:58:31Zagorhow frustrating :-)
09:58:41funmanZagor: how ?!
09:58:54Zagorluck. I haven't changed anything.
09:58:59funmanah ;)
09:59:05Lloreanfunman: I thought you meant "enabled" as in "turned on", not "included"
09:59:19LloreanAs I said before, an unmodified SVN build
09:59:28 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
09:59:30funmanLlorean: sorry for the confusion
09:59:42LloreanI'm compiling one without a plugin buffer now.
10:00:10LloreanOr rather, a quite small one
10:00:17Unhelpfulfunman: if sed -r &>/dev/null ;, i guess? i don't have an osx around to test.
10:00:21funmannot building tagcache is faster ;)
10:01:00funmanUnhelpful: that's probably the right test (with a correct expression because gnu sed -r returns 4 here)
10:01:11Unhelpfulecho | sed of course.
10:01:38Lloreanfunman: I'll do it that way then.
10:01:52funmanjust comment out HAVE_TAGCACHE in config-clip.h
10:02:22LloreanWhat is "audiobuff" then if it's not the PCM buffer and it's not the compressed audio buffer?
10:02:26Unhelpfuldoes &> work in whatever shell osx uses?
10:02:44Unhelpfulfurthermore, is there a /dev/null? :/
10:02:50funmanLlorean: audiobuffer is used for voice, compressed audio, pcm, and probably more things
10:03:10funmanit's just the RAM available which isn't used by rockbox, plugins, or codecs
10:03:25amiconnAh, that reminds me
10:03:29funmanUnhelpful: yes.. it's a bsd ;)
10:03:42amiconnOn the lowmem swcodec targets, voice should probably be handled the same way as on archos
10:04:19 Join mofux [0] (
10:04:21Lloreanamiconn: Swap with playback rather than in parallel?
10:04:29amiconnI.e. instead of permanently reserving a chunk of audio buffer for voice, the voice file should use the main audio buffer, and be swapped out during playback
10:04:59amiconnThis makes voice unavailable during playback, but might be the only way to make it work
10:04:59funmangood idea
10:05:36Lloreanfunman: Alright, playback works (briefly) once I get enough memory free.
10:05:52funmanLlorean: welcome in the club!
10:06:09LloreanWell, I'd already used it with the flash_buffering patch.
10:06:33funmani'd divided the pcm buffer size by 2, let's see what happens
10:06:57Lloreanamiconn: Well, right now we could (presumably) just drop the largest few plugins and still have it do voice the old way, I think
10:07:34funmana quick crash, and I see the same values of filebuf_used and useful_data than when I had a 29kB filebuffer (alloc = 1832, usefl = 1224)
10:07:39 Join Thundercloud [0] (
10:07:54funmanplugin buffer is 384kB I think
10:08:02amiconnMight work if you also switch to flash buffering, but then some nice-to-have plugins might be unable to run
10:08:07ZagorI get totally bogus data sometimes. looks like it gets overwritten.
10:08:08Unhelpfulfunman: works on linux, you want me to put it on FS?
10:08:21amiconnThe jpeg viewer is rather smaller, but needs some buffer for the images
10:08:25funmanUnhelpful: no, better let macosx people who do not want to use GNU fix it ;)
10:08:26amiconn*rather small
10:08:52amiconnSo more free audio buffer is useful in many places
10:08:54 Join lasser [0] (
10:09:16LloreanI think I'd rather not have to stop playback for voice, than not having to stop it to view jpegs.
10:09:22Zagorwhere is the pcm buffer size defined?
10:09:38funmanpcmbuf.c : pcmbuf_get_next_required_pcmbuf_size()
10:09:39LloreanThough actually, if it *always* just stopped playback for voice, that might be better all around.
10:10:04LloreanSave some memory on BigMem targets, resolve the complaint that "sometimes i can't hear the voice over the music" and fix the issue "I can't hear voice while music is paused" all in one go.
10:10:12Unhelpfuli've got the test as described for the bsd-sed args, i just would want an actual OSX user to be sure it works... so FS might be the place
10:11:02funmanUnhelpful: hum i think i have a shorter test
10:11:29amiconnLlorean: "fix" as in "make it the way it's intended to be"?
10:11:36Unhelpful... we definitely want *something* like this, if you run git svn info on a non-svn git repo, it "fixes" it for you
10:11:48funmanUnhelpful: oh
10:11:51Unhelpfulmine's still not done fixing for a few minutes, now. :/
10:12:10funmangit log HEAD^..HEAD|tail -1|cut -d@ -f2|cut -d\ -f1
10:12:12Lloreanamiconn: You don't "fix" something that's working right. :)
10:12:16funmancross platform
10:13:14amiconnLlorean: Pausing != stopping. On archos, voice also doesn't work while paused (because the audio buffer is still occupied, it cannot laod voice)
10:13:39Unhelpfulfunman: erm, that only looks at the most-recent git rev, and if the most-recent git rev is not an svn rev, you get junk.
10:13:54funmanUnhelpful: then origin^..origin ?
10:13:55Lloreanamiconn: that seems odd. Why can't it flush, if it's going to have to rebuffer anyway?
10:14:08amiconn(and it actually pauses the MAS data transfer, allowing to continue from the exact same frame where it was paused)
10:14:22funmanthat assumes user use git pull −−rebase after svn rebase though
10:14:24amiconnIt doesn't have to rebuffer after pause...
10:14:35LloreanIt has to rebuffer after voice though, doesn't it?
10:14:51LloreanI'm confused.
10:14:57amiconnIt doesn't do voice when paused. Only when stopped.
10:15:10funmanor using grep git-svn-id|head -1 : that solves the cross platform issue, perhaps at the cost of tiny cpu cycles (i don't think it's an issue)
10:15:13Unhelpfulfunman: that fails for my clone, at least.
10:15:16LloreanBut when voice should be played, why does it "stop, voice, resume"?
10:15:27Llorean*doesn't it
10:15:36amiconnThat'd be wasteful
10:15:42funmanUnhelpful: git show origin doesn't show you a git-svn-id in the log?
10:15:57amiconnIt would mean stop->load voice->talk->rebuffer->resume
10:16:01Unhelpfulfatal: ambiguous argument 'origin': unknown revision or path not in the working tree.
10:16:13Lloreanamiconn: Which is necessary if you want to use voice anyway, you just have to do it by hand instead.
10:16:29amiconnYes, hence it's better not to do it automatically
10:16:39Unhelpfuli cloned this one from svn myself... maybe zagor's is not broken this way? :/
10:16:45Zagorwhy are we using (NATIVE_FREQUENCY*4) in pcmbuf.c:434 ? last I checked we're not supporting 4-channel audio...
10:17:06funmanhum .. what do you have in .git/refs/remotes ?
10:17:07amiconnZagor: 2 channels * 2 bytes/sample ?
10:17:16Zagoralso the "seconds += 2" ensures minimum 3 seconds, not 2
10:17:24LinusNZagor: could the 4 mean the 4 bytes per sample?
10:17:35funmanUnhelpful: and after git pull −−rebase ?
10:17:45amiconnIn fact I prefer the hwcodec behaviour of not doing voice during playback
10:17:58Zagoramiconn: that makes sense. but then why is it NATIVE_FREQUENCY*2 if MEM <= 1 ?
10:18:21amiconnI have no idea
10:18:22Unhelpfulfatal: 'origin': unable to chdir or not a git archive
10:18:23Unhelpfulfatal: The remote end hung up unexpectedly
10:18:39Unhelpfuli pull from svn
10:18:50funmanusing the instructions on the wiki ?
10:19:13funmangit update-ref refs/remotes/git-svn origin/master
10:19:23Zagoris sansa m200 working? it's the only target with MEM<2.
10:19:38Zagor(well, ifp too but I know that isn't working)
10:19:59Unhelpfulthat fails for me, too :D
10:20:02 Quit axionix_ (Read error: 110 (Connection timed out))
10:20:19Unhelpfulat some point, i need to pull from the "official" git mirror, and see if things work any better.
10:20:23*Zagor regrets not bringing the usb cable to work today
10:20:29funmanUnhelpful: i'm not sure but i think origin ref should be defined
10:21:38funmanUnhelpful: could you paste your .git/config please?
10:22:44funmanZagor: see r12843
10:23:31Unhelpfulfunman: lemme make a fresh clone really quick, of the last few revs, and see if something's not clearly broken...
10:24:06Zagorfunman: yeah I saw that. doesn't really explain the strange code though.
10:24:21funmannope but at least points to its author ;)
10:26:48funmanhum i just saw a drop into used buffer
10:27:06funmani mean very high drop at the moment of the crash, like you told me
10:27:24funmanit displays BUF_USED in buffering.c
10:27:35Zagorchanging pcm buffer to the specified 2 seconds instead of 3 would free up 172KB ram
10:27:48funmanwhich uses scary macros
10:27:48Zagorand I question why we need 2 seconds pcm buffer on flash targets
10:27:57funmanfor effects?
10:28:16Zagorcrossfade is added on top of those 3 seconds
10:28:35funmanlike fade on pause?
10:28:52Zagorthat doesn't affect the pcm buffer
10:28:54LloreanWith the flash_buffer patch at least, the PCM buffer almost never even drops below about 4/5 full from the look of it.
10:29:15Unhelpfulstill going :/
10:29:16 Quit Thundercloud (Read error: 54 (Connection reset by peer))
10:29:20LloreanI don't know how it does with the normal buffering code, since I get a data abort after a few seconds.
10:29:28ZagorLlorean: exactly. I'm guessing the 2 seconds are to allow disks to spin up.
10:29:33funmanLlorean: looks normal
10:29:54Unhelpfuland my config has no origin ref in it
10:29:55funmani'm more interested by the file buffer
10:30:05LloreanZagor: Even that *could* be handled in the compressed buffer instead of the PCM buffer.
10:30:09funmanUnhelpful: does it have any remote ?
10:30:16LloreanIt would make more sense since 2 seconds of compressed audio is a lot smaller, most of the time.
10:30:17ZagorLlorean: crossfade?
10:30:23funmanexcepting the svn-remote
10:30:30Unhelpfulnope! :D
10:30:32amiconnThe pcm buffer needs to be large enough to avoid too frequent boosting/unboosting
10:30:40LloreanZagor: Possibly? Was just saying, disk spinup wouldn't need to depend on the PCM in my mind. :)
10:31:02amiconnOn targets where we don't do that yet, it could very well be < 1 second
10:31:09ZagorLlorean: ah, you're right. pcm shouldn't have anything to do with spinup.
10:31:16Unhelpfulit looks like you get none of that when you run git svn clone
10:31:34Zagoramiconn: is boost transition expensive?
10:31:58amiconnYes, although it varies with SoC
10:32:27amiconnIt's mainly the wait-for-pll-resync that costs time
10:32:31Zagorhow expensive? bigger pcm also means smaller compressed buffer = more spinups/flash accesses
10:32:39Unhelpfula fresh git svn clone of the last few revs is exactly the same
10:33:14amiconnOn coldfire, each transition also adds a bit of random timer uncertainty
10:33:41Unhelpfula git clone of the the git svn clone has a nice origin ref :/
10:33:42amiconn(due to the timers being hooked to the core clock)
10:33:45Unhelpfulremote rather
10:34:04funmanUnhelpful: git remote add origin git:// ?
10:34:12Zagoramiconn: yes, but what is a bit? are we talking microseconds or 100 milliseconds?
10:34:33amiconnZagor: Up to 10 msec (typical 2 msec) on coldfire, and around 500 ┬Ásec on PP iirc
10:34:39Zagorok. thanks.
10:34:55funmaninstant on as3525 :o)
10:34:56 Quit Nibbler (Read error: 110 (Connection timed out))
10:35:02Zagorfunman: really?
10:35:12Unhelpfulfunman: that does 'er. :D
10:35:12funmanthe PLL setting doesn't change
10:35:23Zagorah, only the divider?
10:35:56LloreanSo we could do with a smaller PCM buffer there?
10:36:13amiconnWe could do that too on PP (only changing a post divider), but not on coldfire
10:36:15Zagorwe might want to do some stuff #if MEM < 8 or MEM==2
10:36:39Zagoramiconn: why don't we already on pp?
10:36:44funmanI was wondering if a global define LOWMEM would be useful
10:36:50***Saving seen data "./dancer.seen"
10:36:55amiconnIt would limit us to certain 'normal' frequencies, e.g. 1/3 of the max.
10:37:51Zagorfunman: nah, let's try not adding too many defines.
10:37:56amiconnActually we could run the pll at a higher frequency, and divide even for the max. clock
10:37:59Zagorwe already have a LOT
10:38:39Zagoramiconn: but coldfire must get pll change?
10:38:43amiconnOn PP5022 the pll needs to be run at >=96 MHz anyway, so we wouldn't lose anything, but on PP5020 it doesn't, and if we run it higher, it might need more power
10:39:31amiconnZagor: yes
10:40:56amiconnOn PP we use 80 and 30MHz now. With simple post divider, normal clock could be either 40MHz, or 26.667MHz
10:41:17Unhelpfulbut it still doesn't make any of your suggested methods work for me... :/
10:41:30funmanUnhelpful: git show origin doesn't show anything?
10:41:34Llorean40mhz would be nice for improved UI performance, though if the UI boost patch ever works, maybe 26.67 would work too.
10:41:34amiconnStarting from 160MHz (as we need to do anyway on 5022), we could also use e.g. 32MHz
10:42:02Unhelpfuli think the only thing for it may be to clone from the official repo, and push −−mirror my work onto that
10:42:04amiconnIirc the post divider is 1 <= divider <= 16
10:42:45ZagorLlorean: surely the clip doesn't need a UI boost
10:42:51funmanUnhelpful: if you name the svn remote "origin" does git svn still work ?
10:43:06 Join einhirn [0] (
10:43:11amiconnOf course we can always use 24MHz, but then it makes sense to disable the pll in order to save power. Powering back up then implies resync again.
10:43:13LloreanZagor: Yeah, but the 40 and 26.67 are for PP targets, which tend to sit on the edge of needing it, depending on screen size.
10:43:32Zagoryeah the pp targets are usually a bit more taxing
10:43:42LloreanZagor: The Clip has probably got the best "amount of screen data" to "CPU speed" ratio we've got...
10:43:49amiconnLlorean: Actually I'd say it's just the Video. Color is fine afaics
10:44:15Lloreanamiconn: Yeah, but it's probably close enough that it'd need it if we dropped unboosted down a little further, right?
10:44:45LloreanNot that the patch is ready, anyway.
10:45:00Zagoranyway, since the ams targets use div boosting we have no reason to not shrink the pcm buffer
10:45:50Zagornot that it solves the immediate problems. but memory is rather scarce on these little critters.
10:46:04LloreanZagor: Well, it might bring the audiobuffer back up to "large enough for playback"
10:46:21ZagorLlorean: yeah, without having to cut major features
10:46:35funmanmaybe even "resistant to crashes"
10:46:38LloreanZagor: Where'd it defined, I'll test a build?
10:46:54Zagorgoing from 3 to 1 second pcm buffer will save 344 KB
10:47:06ZagorLlorean: apps/pcmbuf.c line 432
10:47:06funmanUnhelpful: how did you clone that tree, git svn clone ? I'll try to see what i can tweak
10:47:22Zagorsimply remove "seconds += 2;" to get a 1-second buffer
10:47:57funmanhow a larger pcm buffer reduces the number of boosting needed, because boosting is needed by pcm playback?
10:48:27Zagorfunman: boosting is needed by decoder. and decoder can only decode as much as can fit in the pcm buffer
10:48:41Zagorso small buffer = many small decodes.
10:48:47Zagor= many boosts
10:49:46funman-2.3MB of filebuffer used, that can't be right !
10:49:57Unhelpfulfunman: yes, that's how the "broken" one got cloned... i'm still thinking maybe i should just mirror my branches onto a "proper" clone of the official git mirror.
10:50:30Zagorfunman: yeah I get that sometimes too. I suspect overwriting.
10:50:32funmanright, and we should remove the instructions to svn clone then
10:50:52funmanZagor: of buf_.idx , yes me too
10:50:57Unhelpfuldo we even have those? i just followed git-svn's instructions... :/
10:51:33Unhelpfuli think there are more than a few things broken, anyway. pushing svn revs via git to a git-svn repo can very easily make git-svn too confused to be able to rebase :/
10:52:03 Part LinusN
10:52:26funmani would prefer a native git server too .. :o
10:52:36LloreanZagor: Doesn't make playback work. My compressed buffers are much bigger now, but it still fails to play. I'll disable tagcache and see if it at least seems to work once the total audiobuff is big enough
10:55:16 Join gregzx [0] (
10:55:24Unhelpfulwell, it's good enough as a conduit to move revs out of svn and into git, and apparently if we all clone our git repos from a git-svn clone, ours will be fine
10:56:43LloreanZagor: With the smaller PCM buffer playback seems to work well enough. It's unreliable, but I think it's just the usual unreliableness.
10:57:27 Quit jhulst (Remote closed the connection)
10:57:59ZagorLlorean: did you ever try standard pcm buffer without tagcache?
10:58:22LloreanZagor: Yeah, got data aborts after ~4 seconds of playback
10:58:30LloreanI'm not getting data aborts like this.
10:58:35Zagoryet :)
10:58:43LloreanI had a little difficulty getting playback to start
10:58:55LloreanI clicked on the file, got to the WPS, nothing happened, I paused then unpaused and it started.
10:59:08funmanthe reliable unreliableness
10:59:13LloreanIt's been going for quite a bit so far, with me watching the audio buffer screen. It *might* make sense to change the low watermark for the PCM buffer now.
10:59:30ZagorI wonder if we have ever successfully ran swcodecs on a <8MB target
11:00:05LloreanZagor: Someone declared an enormous plugin buffer on an iPod Nano once.
11:00:16funmanwith the in-progress buffering_flash.c i'd think yes
11:00:17 Quit gregzx ("ChatZilla 0.9.84 [Firefox 3.0.4/2008102920]")
11:00:20LloreanSo they, effectively, had a less than 8mb target I think
11:00:33Zagorthere seems to be a lack of checks in the buffering/playback code
11:00:34Lloreanfunman: So far my music is still playing, and none of the audio glitches I get wit the buffer_flash patch.
11:00:55funmannormal, because buffering.c use watermarks
11:01:22LloreanI do need to pause and unpause to start playback, which is strange.
11:01:37LloreanThat's reproduceable like this, at least
11:02:14Zagorfunman: how is the status of the sd driver? do we know why we get fifo full?
11:02:21 Join AndyIL [0] (i=AndyI@
11:02:28Zagoror when
11:02:44funmanZagor: why, no, you might want to save the function params at each call and display them in the panic
11:02:50 Join robin0800 [0] (
11:02:51funmanwhen, when buffering bugs
11:02:56LloreanThe compressed audio buffer seems to refill *very* often
11:03:00funmanthe state : additional error handling wouldn't harm
11:03:05LloreanIt seems to refill at about 512k
11:03:13LloreanWhich is a bit early when it's only 630k large
11:03:27ZagorLlorean: yes, I noted that too. the watermarks are totally off.
11:03:34Zagorexcept for wav, for some reason
11:03:37funman#define BUFFERING_DEFAULT_WATERMARK (1024*512)
11:03:45Zagorfunman: ah! :-)
11:04:09funmani think all the keys are in buffering.c since replacing this file gives completely different results
11:04:18Zagornew targets are good. it gives us reason to revisit and improve old code.
11:05:09*Llorean tries with a smaller watermark
11:05:14LloreanFlash doesn't need to wait for spinup
11:05:49Zagoryeah, seems the buffering code could use a number of #if HAVE_FLASH_STORAGE
11:07:17LloreanI imagine the watermark being higher than the total buffer size caused at least some problems
11:07:20kugelZagor: Or use the flash buffering patch, maybe with a higher buffer?
11:08:08Zagorkugel: that's definitely an option. we need to determine which approach is the most power efficient.
11:08:25funmanZagor: hum now i record : the corruption i had seen was in the linked lists struct memory_handle
11:08:26ZagorLlorean: indeed
11:08:39LloreanI tried changing the DEFAULT_WATERMARK to 1024*128 and it's still rebuffering at 512k
11:08:43LloreanIs there something else I need to look at?
11:08:49funmanand the useful_data flag comes from the active struct
11:09:04kugelZagor: I think a combination of both. As in use the available ram on as buffer for flash buffering too (although it potentially only needs very little)
11:09:40Zagorkugel: well we can't make it too special, since the audiobuffer is a core element that is used by a lot of other code too.
11:10:55kugelZagor: That's what I meant: Full the very little flash buffer with the available ram to form the "classic" audiobuffer
11:11:58 Join Darksair [0] (n=user@
11:12:00funmanand the memory_handle s are stored in buffering.c's "buffer" (= playback.c's "filebuf")
11:12:23*Llorean assumes the problem is that he left high watermark alone
11:12:34Zagordo we have an SD guru? do SD cards normally have power management? i.e. is it more efficient to do few long reads, or is it just the same to do many frequent small reads?
11:12:57funmanhum that reminds me the pl180 has a 'powersave' bit ..
11:13:16funmanSD controller inside the as3525 SoC
11:13:24kugelZagor: when I tried the flash buffer patch on my e200, it definitely had an impact on battery life
11:13:32Zagorkugel: negative?
11:13:50kugeland that was before the "sd_enable()" commit
11:14:00funmanthat's what i was going to ask : what about now?
11:14:11 Quit AndyI (Read error: 110 (Connection timed out))
11:14:32Unhelpfulfunman: your short test w/ origin in place of HEAD seems to be the way to go
11:14:47Unhelpfulgit log origin^..origin|tail -1|cut -d@ -f2|cut -d\ -f1
11:14:51funmanas long as there is an origin ;)
11:15:12Zagorwhat exactly do we gain by reading directly from flash? it makes a lot of other things more difficult, such as MoB.
11:15:13funmanand note that if you don't use git pull −−rebase, origin will not be updated
11:15:46*Llorean is apparently failing to convince it to use a different watermark.
11:15:58Zagormy feeling is that it's a last resort for very lowmem targets that simply cannot work otherwise
11:16:18LloreanThat being said, as long as I sit in the buffer screen rather than the WPS, I currently have perfect playback, it seems.
11:16:26LloreanIf I go to the WPS it eventually freezes up.
11:16:43Unhelpfulfunman: what about using git rebase?
11:16:50ZagorLlorean: try changing tracks in the buffer screen (press up & down)
11:16:57funmanUnhelpful: yes that's probably better, i'm too used to git pull ;)
11:17:04LloreanZagor: Works fine
11:17:20Unhelpfulgit rebase doesn't seem to work w/o arguments
11:17:36Unhelpfulso it's either git rebase origin or git pull −−rebase :/
11:17:45funmanLlorean: if repeat again multiple times you'll get other (iconsistent) crashes i believe
11:17:57Lloreanfunman: Repeat what?
11:18:07funmanany steps related to playback in fact
11:18:18LloreanNo crashes
11:18:27LloreanI just skipped about 12 tracks.
11:18:36funmanbut you only did that for 1 or 2 hours
11:18:56LloreanIf I skip many tracks, very fast, playback pauses, but if I go to the WPS and pause then unpause, it resumes.
11:19:01 Quit kugel (Remote closed the connection)
11:19:06funmanfrom my longer experience i think that the crashes happen randomly and not because of particular conditions
11:19:42LloreanWell, it was crashing _very_ easily before making the PCM buffer smaller. I currently have 1 second PCM buffer, and ~730K of compressed buffer.
11:20:01LloreanUnfortunately, I failed to set the watermark lower.
11:20:11LloreanSo it's still rebuffering quite often.
11:21:00LloreanHm, "backlight on hold" doesn't seem to work on the clip.
11:21:19ZagorLlorean: watermark is also set in each codec. ci->configure(CODEC_SET_FILEBUF_WATERMARK, 1024*512);
11:22:05B4gderhaha, how... eh, weird
11:22:10LloreanThat seems strange to me.
11:22:17Lloreanit seems like the watermark should rather be a calculation.
11:22:20ZagorI won't argue with that...
11:22:23LloreanAverage bitrate * X seconds, or something
11:22:33LloreanBasically "long enough to spinup, plus a bit"
11:22:49B4gderthe hwcodec has it based on the hd spinup time iirc
11:22:57LloreanThough I guess it'd get quite complicated with mixed codecs.
11:23:04ZagorB4gder: at least it did, last I looked at it :-)
11:23:16B4gderI don't think that has moved much during recent times
11:23:40ZagorLlorean: well there are still much better ways to do it than this way
11:24:08LloreanAnyway, for whatever clue it's worth, if I'm in the WPS I seem much more likely to get a freeze or halt in playback.
11:24:44 Join gregzx [0] (
11:25:59ZagorLlorean: could simply be that wps touches a lot more memory areas than buffer screen, and so if things get overwritten wps will go wrong faster than buffer screen
11:26:44LloreanCould be.
11:26:57Zagormy theory remains buffering overruns, due to faulty assumptions about buffer size.
11:27:52LloreanI'm actually most curious why playback doesn't start until I pause and unpause.
11:28:08LloreanSince I can also get it "stuck" like that by skipping many tracks very quickly.
11:29:22 Quit linuxstb (Read error: 110 (Connection timed out))
11:33:16 Join aliask [0] (n=chatzill@rockbox/developer/aliask)
11:36:39 Join kugel [0] (n=chatzill@unaffiliated/kugel)
11:49:22 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
11:52:10 Quit kugel (Remote closed the connection)
11:54:11 Join culture [0] (
11:55:51 Join n1s [0] (n=nils@rockbox/developer/n1s)
11:56:32 Join moos [0] (
12:08:54 Quit Darksair ("To Arch or Gentoo? That is the question...")
12:13:53funmani don't have playback anymore, perhaps it's due to too small pcm buffer?
12:14:05 Quit robin0800 (Read error: 104 (Connection reset by peer))
12:16:33funmanaccording to the debug menu, the pcm buffer fills and then everything just stops
12:18:33n1shmm, this looks weird
12:18:59*n1s wants an iaudio [xm]5 tester
12:21:01funmanand using an higher pcm buffer makes playback start again
12:21:59Lloreanfunman: I had playback with a ~176kb PCM buffer.
12:22:15LloreanEr, 192 perhaps
12:22:34 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
12:22:35funmanif i use 1 second, i got nothing
12:22:42Lloreanfunman: Did you pause and unpause?
12:23:03LloreanHow big are your compressed buffers?
12:23:09funmanvery big
12:23:11LloreanEr, is your compressed buffer.
12:23:22LloreanMine was about 900k
12:23:33funmanhumm i think it was less
12:23:43funmanwhere did you get this extra space i haven't?
12:23:49LloreanI disabled tagcache, and lowered my max playlist to 2000, and max files in folders to 100
12:24:16Lloreanoh, sorry, 720
12:24:18LloreanNot 900
12:24:23*Llorean wonders where he got 900 from
12:24:24funmanah now pause/unpause starts playback
12:24:58funman172kB pcm , 615kB filebuf
12:25:45LloreanSame PCM, slightly more compressed here.
12:27:37n1shmm, not so weird after all, just a confusing keymap, still if someone could confirm FS #7572 on [mx]5 would be nice
12:27:45pixelman1s: what for?
12:28:49n1spixelma: No one has confirmed it, it's old and the code seems like everything should work the same as the irivers etc
12:29:08pixelman1s: can't help you with this as the M5 doesn't have radio...
12:32:28funmani'll stop reading that code else it'll give me headaches
12:33:24 Quit einhirn (Read error: 104 (Connection reset by peer))
12:36:52***Saving seen data "./dancer.seen"
12:39:20n1srasher: i'm pretty sure FS #8819 has not been fixed but the only way to test is to commit a broken amnual update :)
12:40:14Lloreanfunman: So far with the small PCM buffer and large compressed buffer, the only crash I've gotten has been a Panic for the SD FIFO
12:40:19 Join _lifeless [0] (n=lifeless@
12:42:00funmanLlorean: i also get deadlocks, data abort (in buffering.c code messing with memory_handle linked lists, or memcpy, memset ..), stack overflows (less common)
12:46:03 Join Aurix_Lexico [0] (n=comrade@
12:56:05*Zagor wants to see static symbols in the .map file
12:56:50ZagorI get a data abort in buffering.c, but it's not in one of the public symbols :-(
12:57:12funmanZagor: isn't it a struct access ?
12:57:23funman(or close to)
12:57:29Zagorpointer access, at least
12:57:44Zagorbut there's quite many of those in that file...
12:58:18funmanIDA makes it easy to read rockbox.elf , dunno about objdump
12:58:34Zagorgood idea
12:59:04ZagorI forget, do we already have some support for assert()? methinks it would be useful in cases like this.
13:00:35funman#define assert(x) do { if(x) panicf(##x); } while(0) /* here it is */
13:00:55*n1s wonders if jhMikeS has a different definition of 'freeze' than everyone else :P
13:00:55funmanif(!x) even
13:02:44Zagorwe have include/assert.h, but I can't find any implementation of __assert
13:03:06*jhMikeS missed the "freeze" announcement? Or is keeping things sane a bugfix?
13:03:35ZagorjhMikeS: what are you referring to?
13:03:36LloreanIt's probably one of those gray areas anyway.
13:04:01jhMikeSn1s: wondering about my definition of "freeze" :)
13:04:11funmanZagor: i found also some other functions declared in headers but not defined
13:05:34Zagorthe data abort is in fill_buffer()
13:06:08funmanZagor: m = m->next ?
13:06:23funmananyway that means corruption of the memory_handle
13:06:46funmanor perhaps first_handle
13:07:26ZagorI'll disassebmle and see
13:08:08funmandisassmatroskae, disbuttmatroskae, (ok i stop)
13:08:24jhMikeSbeep (OT warning)
13:09:24 Quit funman ("leaving")
13:14:43 Nick fxb__ is now known as fxb (
13:22:08 Join pvbcharon [0] (n=charon@
13:22:18 Quit pvbcharon (Client Quit)
13:32:44ZagorI wonder if periodical list walking/checking would yield anything
13:36:07*Zagor feels a little guilty for playing with a new target instead of fixing bugs for the release...
13:39:35LloreanYou're just getting a big head start on 3.2
13:39:39LloreanOr maybe 3.3
13:42:07 Quit Nico_P (Remote closed the connection)
13:44:03 Quit Aurix_Lexico ("Leaving.")
13:45:21 Join kugel [0] (
13:45:41 Join mc2739 [0] (
13:47:33 Part B4gder
13:48:06mc2739can I get a dev to fix FS #9625, I hit add task too soon
13:48:35mc2739it should be: tasktype=patch, category=music playback, and player=another
13:49:35 Quit aliask ("ChatZilla 0.9.84 [Firefox 3.0.4/2008111318]")
13:49:54kugelmc2739: I assume you got sound?
13:50:47Lloreanmc2739: Fixed..
13:50:59mc2739Llorean: thank you
13:51:30mc2739kugel: yes, flac, wav, all play - mp3 has problems
13:51:43kugelI tried ogg and it didn't play
13:52:06mc2739ogg worked for me also
13:52:14kugelI tried to ge to the wps using a playlist_control and nvram.bin in combination with start screen: wps
13:52:41mc2739that is what I had to do to get it to play
13:52:58Zagorsounds like we should do what the comment above that #if says...
13:54:04Zagor"TODO : If this turns out to apply to all ams3525 targets, consider simplifying the precompiler condition to #if defined(AS3525)."
13:54:12*kugel highly assumes it also applies for fuze (and probably c200v2 too)
13:54:25kugelso, agree to Zagor
13:54:54 Join stripwax__ [0] (
13:54:57 Quit stripwax__ (Client Quit)
14:01:53 Quit mc2739 ()
14:04:06 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
14:04:12 Join DerDome [0] (n=DerDome@
14:04:19 Join massiveH [0] (
14:10:02 Join LambdaCalculus37 [0] (i=44a04303@rockbox/staff/LambdaCalculus37)
14:12:39 Join CaptainKewl [0] (
14:15:17soaplinuxstb: congrats again on your polite and well spoken interaction with TBeck. I hope his offerings prove useful to you.
14:15:29 Quit culture (Read error: 60 (Operation timed out))
14:16:50 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
14:17:58 Join culture [0] (
14:18:19 Nick fxb is now known as fxb__ (
14:26:11 Join Schmogel [0] (
14:28:40 Join Jaykay [0] (
14:36:45 Quit CaptainKewl (Remote closed the connection)
14:36:57***Saving seen data "./dancer.seen"
14:38:35*Jaykay breaks the silence
14:38:53*Jaykay hopes that the developers are not annoyed by those "i suggest closing of...."-messages...
14:39:13Zagornot if they are reasonable
14:39:50Jaykayi suggest closing of because when the player says 1020 min remaining you just think wtf......
14:41:29 Join itcheg [0] (i=41d59de2@gateway/web/ajax/
14:43:33ZagorJaykay: I disagree. I might not want it myself, but that is no reason to reject it. It's a valid feature and it doesn't add much code.
14:44:09 Quit itcheg (Client Quit)
14:44:18 Join itcheg [0] (i=41d59de2@gateway/web/ajax/
14:45:39JaykayZagor: du you really think somebody wants the battery lifetime given in minutes? you wont know how much hours 1000+ minutes are....
14:46:11Zagorobviously the person who bothered to write the patch wants it
14:47:39Zagorthe patch is adding a wps tag to be used by whoever chooses to. it's not changing anything, only adding.
14:48:27Jaykayok so i am allowed to write a patch, say "i do want this" and hen the patch remains in the tracker for years? ;)
14:48:52LloreanIt'll stay in the tracker until we make up our minds as to whether we want it or not.
14:49:03ZagorJaykay: of course you are allowed to do that.
14:49:23LloreanYou need to be just as certain about rejecting it as you are about committing it.
14:49:35Jaykayknow i know why there are 340+ patches^^
14:50:20LloreanIt's better than none, probably.
14:51:36Jaykayanother patch:
14:52:01Jaykayit does something automatically.... without asking for anything.
14:53:17Jaykayand those total track/disk tags can be useful (nevertheless i dont use them)
14:54:51 Nick fxb__ is now known as fxb (
14:55:36LloreanJaykay: The patch opener is a dev, so it should probably wait for his response at least before closing.
14:56:31PaulJami think you can even acieve this without the patch, by just displaying this information in a vievport that just fits 2 numbers, so if the tracknumber is 08/12 then only 08 would be visible in the WPS.
14:57:03Jaykayyeah but as far as i see this patch is deleting the tags...
14:57:17Jaykaynot only showing part of it
14:57:26LloreanIt's ignoring part of it.
14:57:46LloreanRockbox has no means of writing tags back to the files.
14:58:45Jaykayok im sorry..
14:59:20Zagoron this one I agree with jaykay. rasher showed already aug27 that the patch goes againt the spec.
14:59:40LloreanZagor: I stated I don't like the idea of the patch in the very first comment.
15:00:10LloreanI'd just think we should wait for lear to have a last response before closing, is all. Or at least, give him a day or two to respond, the last comment's only from yesterday.
15:00:28ZagorI'm ok with that
15:09:10LambdaCalculus37I think that the patch doesn't really have any use at all, but yes, it's lear's patch, so let's have him speak about it.
15:13:14 Quit PaulJam (".")
15:31:08 Quit pondlife ("Leaving.")
15:37:39 Quit Jaykay (Read error: 110 (Connection timed out))
15:38:34 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
15:39:48 Join MethoS-- [0] (
15:40:46 Join Horschti [0] (
15:41:32 Quit Horscht (Nick collision from services.)
15:41:42 Quit MethoS- (Read error: 60 (Operation timed out))
15:43:22 Quit Rob2222 (Read error: 110 (Connection timed out))
15:45:34 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
15:45:35 Join jgarvey [0] (
15:53:00 Quit perrikwp (" ajax IRC Client")
15:53:08 Join perrikwp [0] (i=18ac0c41@gateway/web/ajax/
16:08:42 Join massiveH_ [0] (
16:13:56 Join Seed [0] (
16:16:10 Quit DerDome ("Leaving.")
16:25:48 Quit massiveH (Read error: 110 (Connection timed out))
16:37:00***Saving seen data "./dancer.seen"
16:53:07Lloreandomonoky: Are you interested in accessibility in RBUtil or is it primarily Bluebrother who's worked on that?
16:53:51 Join toffe82 [0] (
16:54:33domonokyLlorean: i am interested in it, but i miss the time for it. I think we 8me and bluebrother) both worked on accessibility issues.. :-)
16:54:58Lloreandomonoky: You may be interested in
16:55:04Zagorto build the database pc tool, we should probably just start with the uisim, put in the five-line database.c main() and then chip away what isn't needed. doing it the other way is ... frustrating
16:55:10*domonoky also already replied to brandon
16:55:22LloreanOkay, just making sure.
16:55:43LloreanNever do know who's subscribed to that list and who isn't.
16:56:15domonokytrue, so thanks for noticing me. :-)
16:56:49domonokydid you find time to try out your m200v4 ?
16:57:26Lloreandomonoky: Not yet. I keep getting distracted with my clip.
16:59:56 Join kachna [0] (
17:04:32LambdaCalculus37amiconn: Ping
17:07:42 Quit Zagor ("Client exiting")
17:10:20 Nick massiveH_ is now known as massiveH (
17:13:17 Join CaptainKwel [0] (i=2669ecc2@gateway/web/ajax/
17:14:35 Quit petur ("connection reset by beer!")
17:15:13 Nick Horschti is now known as Horscht (n=Horscht@xbmc/user/horscht)
17:20:54 Join DrMoos [0] (
17:21:04 Quit DrMoos (Read error: 104 (Connection reset by peer))
17:29:17n1shmm, i wonder how we can have only 23 "known" bugs in the 3.1 release notes when there are 169 open bugreports...
17:29:48kugelmarketing :)
17:30:12 Quit kharo (Read error: 110 (Connection timed out))
17:30:13 Join kharo1 [0] (
17:30:31n1son a related note, i wonder why we list still open bugs in the release notes at all?
17:31:58kugelanti-marketing :)
17:32:07*kugel runs
17:32:18LloreanIf we tell them it sucks, they'll be even happier when it's awesome.
17:32:18*LambdaCalculus37 chases kugel with a squeaky hammer
17:32:27n1sa question about FS #9427 can tracks in a file played with a cue sheet have individual meta data?
17:38:58kugeln1s: yes, I don't know if rockbox supports it though. The cue file can declare some fields individually for each track
17:41:21linuxstbn1s: Yes, Rockbox stores per-track metadata for a file with a cuesheet.
17:41:25linuxstb(assuming that info is in the cuesheet...)
17:46:46n1sso it _could_ be feasible to scrobble individual tracks from a cue sheet-file thingy?
17:47:34linuxstbWhy only "could be" ? I would say it _is_ feasible.
17:47:59 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
17:48:25linuxstbI guess it could get complicated when determining % of tracks played...
17:48:32kugelthe problem seems to be that scrobble thingy doesn't actually read the same thing as the wps (I seem to remember the wps shows the individual track info from the cue)
17:48:46n1sDoesn't the scrobbler require a bunch of fields? I guess i just don't know enough about cue sheet metadata stuff?
17:49:24n1sah, well thanks for clarifying though :)
17:49:32linuxstbThe cue contains artist name, track name and album name. I guess anything else would be from the whole-file metadata in the tags.
17:49:39 Join perrikwp|lab [0] (i=982129cf@gateway/web/ajax/
17:50:02kugeltrack number is probably too (at least indirectly) in the cue file
17:50:38 Quit massiveH ("Leaving")
17:51:27kugellinuxstb: I wonder if you could commit , I tested it in the bootloader
17:51:59*kugel wonders if funman is going to show up again before he starts his travel tomorrow
18:00:35 Join evilnick [0] (i=0c140464@gateway/web/ajax/
18:00:58 Join Rob2222 [0] (
18:03:16 Join Strife89 [0] (n=michael@
18:09:20 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
18:11:33 Quit MethoS-- (Read error: 104 (Connection reset by peer))
18:11:41kugeldoes anyone already have an idea what to do about the resizers in the pluginlib? smooth_resize and simple_resize
18:12:07kugelI think they're slightly obsolete with resize in the core
18:16:05evilnickIs there anyone with a Sansa E-series around?
18:17:03 Quit Llorean (Read error: 104 (Connection reset by peer))
18:17:23evilnickAfter the addition of the software fade-out, I've had a few instances when a button-press as the screen starts to fade-out will NOT reset the time-out (i.e. the screen still fades out)
18:17:28 Join Llorean [0] (
18:18:04evilnickIt's damn hard to reproduce though
18:18:19 Quit TMM (Remote closed the connection)
18:19:33kugelevilnick: I made the patch. I have not experienced what you describe though
18:19:42 Join TMM [0] (
18:19:52kugelyou maybe didn't press hard enough?
18:20:36evilnickIt's whilst I was in Solitaire, and pressed Rec to deal another three cards - the Rec button performs the correct function in Solitaire but the screen continues to fade-out
18:21:01 Join MethoS- [0] (
18:21:23evilnickThis happened only 3 or 4 times in about half an hour, so it's usually working as expected. I just wanted to see if anyone else had experienced it.
18:21:28 Join karashata [0] (n=karashat@
18:22:58 Join miepchen^schlaf [0] (
18:23:09kugelevilnick: I have indeed a hard time trying to reproduce it
18:24:13kugelevilnick: have you only experienced it in solitaire so far?
18:24:21evilnickkugel: I'm guessing that it's such a small window of opportunity that it might not be worth looking into.
18:24:40evilnickkugel: yes, so far. Would the fact it's a plugin make a difference?
18:25:05kugelcould be
18:25:45evilnickIt seemed to happen if the fade-out had *just* started if that's any help
18:27:20kugelevilnick: sorry, I try hard, but I'm not able to reproduce
18:28:09evilnickYeah, just tried on the wps screen using play/pause and can't reproduce it either, so perhaps it's a problem with the Solitaire plugin.
18:28:34kugelI'm trying solitaire
18:29:14evilnickkugel: try during a commute so that you can spend quite some time on it!
18:29:14 Quit Schmogel ("Miranda IM! Smaller, Faster, Easier.")
18:29:53kugelevilnick: Ok, I got it now, with 1s backlight timeout
18:30:32 Join Schmogel [0] (
18:30:44evilnickcool - it's very rare, but I'm glad I got to describe it to the correct dev!
18:31:01 Part evilnick
18:31:55 Nick fxb is now known as fxb__ (
18:32:41kugelI fail to reproduce it in other situations though
18:37:03***Saving seen data "./dancer.seen"
18:37:30slactquick question: rockbox will accept dirty USB power, right?
18:37:38 Quit Schmogel (Read error: 60 (Operation timed out))
18:37:48 Join jeffdameth [0] (
18:38:30slact(that is, I wouldn't need to negotiate a "usb 2-capable, powered hub" on the data lines to get the battery to charge)
18:38:36slact(...on a 5.5G ipod)
18:40:05rasherslact: I actually don't think so.. gevaerts is the man you want to ask to get a real answer
18:40:35slactwell, it wouldn't kill it, right?
18:40:48slacti mean, 5 volts is 5 volts...
18:41:26rasherI've no idea, honestly.. I just remember hearing reports about not charging right
18:41:37slacteh, guess i'll just give it a try.
18:42:57slactseems to be charging okay...
18:43:11slactdebug screen says so, voltage graph says so
18:43:35slactthat makes my life easier, but i'm not sure that's great in general...
18:53:14 Join bertrik [0] (
18:55:19 Quit Strife89 ("Bye all.")
18:56:30 Quit perrikwp|lab (" ajax IRC Client")
18:56:34 Quit nplus (Read error: 110 (Connection timed out))
19:00:29 Join nplus [0] (n=nplus@141.25.Globcom.Net)
19:03:32 Join massiveH [0] (
19:08:13 Join herrwaldo [0] (
19:11:14 Quit linuxstb (Read error: 113 (No route to host))
19:21:24 Quit avacore (Read error: 110 (Connection timed out))
19:27:50*kugel finally got a final version of his pictureflow patch out!
19:27:56kugelNico_P: You might have a look at it
19:31:13LambdaCalculus37kugel: What does your patch do?
19:32:37 Join avacore [0] (
19:32:39kugelLambdaCalculus37: it adds configurability to the album title (at top, at bottom or not at all), introduces resizing (to lcd_height/2, configurable), and fixes 2 outstanding bugs
19:33:01 Join eric [0] (
19:33:05kugelparts of it have been committed already during the previous freeze ;)
19:33:19 Nick eric is now known as Guest68088 (
19:33:32LambdaCalculus37kugel: Would one of those outstanding bugs happen to be PF not properly refreshing its database when told to do so via its option menu?
19:34:02Guest68088I recently swapped out the HD in my Gigabeat F10 and I have tried all the recovery procedures to no avail....I was wondering if someone can help...
19:34:08LambdaCalculus37Refreshing the database in that manner often cause PF to display certain album covers in place of blank covers.
19:34:17kugelLambdaCalculus37: No, not that one. I suppose you mean the one where covers of deleted files still show up?
19:34:28kugelor changed covers aren't refreshed
19:35:12kugelthat's basically solved by really deleting the old cache. Yea, that should be doable (once I know how to batch delete files in plugins, I probably should look at disktidy)
19:37:49Guest68088Broke rule #9
19:38:22 Quit slact (Read error: 60 (Operation timed out))
19:38:42LambdaCalculus37kugel: I've been doing that already (deleting the old cache and letting PF rebuild), but it sometimes tends to be a little tedious, not to mention a bit kludgey.
19:38:43kugelrb-remove(/absolute/path/to/file) ?
19:39:01kugelLambdaCalculus37: yea, pf should do that for you, I agree
19:39:46 Quit MethoS- (Remote closed the connection)
19:40:14LambdaCalculus37kugel: I think disktidy would be a good starting point to find out how to delete files from within plugins. Once you figure it out, adding an option to PF (say, "Rebuild art from scratch") would make the job much easier.
19:40:22Guest68088Swapped out my Gigabeat F10 hard drive with a new HD from the list provided on the website. Everything is connected fine, used QT parted to create a partion and formated it with FAT32. Attempting to transfer the GBSytem folder to the drive, but the drive is not recognizable in Ubuntu, or XP. I can see it in Vista through Disk Management, but states not initialized. Need help getting HD recognized so I can reload System and
19:40:22Guest68088 RockBox
19:40:41kugelLambdaCalculus37: Why adding another item? Just let the existing rebuild cache do that
19:40:57LambdaCalculus37kugel: True.
19:44:05advcomp2019Guest68088, just wait for someone that knows it
19:47:00 Nick Guest68088 is now known as J0hnnyBr (
19:47:03toffe82Guest68088: if the drive is no recognized, you have to do the recovery again
19:48:33 Join fml [0] (n=4fd3c4a5@gateway/web/cgi-irc/
19:49:23fmlHello. Wouldn't it be good to have some last entries from the MajorChanges page on the start rockbox page?
19:49:56J0hnnyBrtoffe82: I've done the recovery a few times already...I don't understand why the drive shows up until I add a partition and format...weird
19:50:15LambdaCalculus37J0hnnyBr: Please stick with a nick.
19:50:32toffe82after the format, if you don't disconnet, you don't see it ?
19:51:49J0hnnyBrtoffee82: I took the drive and plugged it in to my Ubuntu shows up but cannot mount. I then ran QT Parted and created a single FAT32 partition and formated with FAT32. When I go back into the file manager, the disk no longer shows up
19:53:56J0hnnyBrtoffee82: When I follow the recovery procedure on my Vista box, I get a message that the Disk is not initialized and when I try to initialize it, it states there is an I/O error
19:54:46 Quit itcheg (" ajax IRC Client")
19:56:44 Join stoffel_ [0] (
19:57:28kugelLambdaCalculus37: fixed!
19:59:48toffe82J0hnnyBr: did you try back your old drive ?
20:01:18LambdaCalculus37kugel: Cool, what's the FS#?
20:02:46kugelLambdaCalculus37: Do you know which FS# that was?
20:03:16kugelI mean the bug report on that
20:03:55kugelhm, I think I have a better plan
20:06:52LambdaCalculus37kugel: No, I don't remember.
20:07:38 Join funman [0] (n=fun@rockbox/developer/funman)
20:08:02 Join bluebrother [0] (n=dom@rockbox/developer/bluebrother)
20:12:39 Quit fml ("CGI:IRC (EOF)")
20:12:58 Quit stoffel_ ("leaving")
20:18:27 Join itcheg [0] (i=62db4767@gateway/web/ajax/
20:21:31rasherkugel: how about getting the artist / albumartist information before searching for albumart? (to fix FS #9606 and remove the need for FS #9622)
20:22:03kugelrasher: I'll have a look
20:22:38kugelif it fixes fs#9606, then yea, that's proper
20:24:02rasherYeah, I mean actually getting the artist and albumartist value and putting that in id3.artist/id3.albumartist.. that would fix FS #9606, and FS #9622 wouldn't be necessary I guess
20:26:40 Quit Bensawsome (Remote closed the connection)
20:28:12 Join aarcane [0] (
20:28:50kugelLambdaCalculus37: the fix is at FS #8335
20:29:16 Join petur [50] (n=petur@rockbox/developer/petur)
20:30:16 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
20:30:16rasherkugel: I'm going to commit FS #9622 (since it's currently buggy) and add a note that it should ideally be set to the actual artist/albumartist value instead
20:31:08kugelrasher: Ok. On a side note: It seems I somehow fixed that issue with the latest patch FS #8335 (No idea why though, but the sim doesn't crash anymore
20:31:29rasherkugel: you're probably just lucky concerning memory usage
20:31:50kugelfixed as in fix 9622, not as in fix#9606 probably
20:32:11LambdaCalculus37J0hnnyBr: What's the drive model you're trying to use?
20:32:30kugelrasher: hm, you mean that my sim gets the same portion of ram again and happens to retrieve correct data by that?
20:32:58rasherkugel: or another portion where the memory happens to be null (or something else that doesn't cause trouble)
20:33:11*domonoky wonders why video.c in plugins/SOURCES is dependend on CONFIG_LCD == LCD_SSD1815
20:33:26kugelrasher: yea, that's possible
20:33:34*kugel haven't thought of that
20:34:44*kugel fails to get segfaults.
20:35:02kugelcan anyone else try to see if FS #8335 "fixes" the segfaults?
20:37:06***Saving seen data "./dancer.seen"
20:38:29pixelmadomonoky: that's the Archos video player
20:39:19domonokypixelma: yes, and the m200v4 has a LCD_SSD1815, so i have to change the check.
20:41:02*domonoky finds chopper very hard to play on m200v4, the display is so small, you nearly always crash before you can even press select :-)
20:41:23LambdaCalculus37domonoky: The tiny displays make it much harder to play.
20:41:44J0hnnyBrLambdaCalc: The drive is MK1231GAL
20:42:01J0hnnyBrLamdabCalc: Toshiba...
20:42:12pixelmadomonoky: already using the half chopper size as Archos?
20:42:36 Join Reizo [0] (n=4232bff8@gateway/web/cgi-irc/
20:42:47LambdaCalculus37J0hnnyBr: Have you read before?
20:42:53pixelmabtw... amiconn ;)
20:43:03domonokypixelma: i only added keys, not more.. everything else can come later. Now i need a solution to the video.c and i can commit plugins for m200v4.
20:43:22*LambdaCalculus37 should look for an m200v4 and get in on the fun as well... ;)
20:43:39J0hnnyBrLambdaCalc: Yes...and the other docs on the site as well
20:43:52pixelmadomonoky: maybe check for swcodec/hwcodec too? Though I don't know if that's the clean solution (TM)
20:44:02ReizoHello. I wanna join TWiki. And I read, that I needed to introduce myself here. Or did I mis interpret?
20:44:20LambdaCalculus37Reizo: Well, you do have to come here to ask for write permission.
20:44:34LambdaCalculus37J0hnnyBr: Bad news: (read a little further down).
20:44:38J0hnnyBrLambdaCalc: OK...I see why you are telling me that....
20:45:05LambdaCalculus37Reizo: What's your wiki name? I can give you permission.
20:45:27J0hnnyBrLambdaCalc: I missed the *. Yep....kicking myself. Saw the drive was cheap and jumped at it...
20:45:35ReizoThat's the thing. I'm kinda lost when it comes to joining. :o
20:45:43gevaertsLambdaCalculus37: maybe flashing the rockbox bootloader helps?
20:46:02J0hnnyBrLambdaCalc: Thanks for your help! ID10T error ;)
20:46:38funmandomonoky: why not checking the defined model ?
20:46:49linuxstbdomonoky: Adding && (CONFIG_CODEC != SWCODEC) would seem OK to me.
20:46:52LambdaCalculus37gevaerts: I'm not sure if it'll help. The guy who was attempting the upgrade tried the recovery procedure as documented, and found that the drive wouldn't work whatsoever.
20:46:58funmanthere is only 1 with LCD1885 which is not m200v4 no?
20:47:10gevaertsLambdaCalculus37: yes, but that's using the OF
20:47:25*domonoky will use the CONFIG_CODEC check, video.c probably also depends on this.
20:47:37linuxstbfunman: No, all the Archos devices have it
20:47:42 Join mc2739 [0] (
20:47:42linuxstb(apart from the Player)
20:47:57 Part J0hnnyBr ("Leaving")
20:48:25LambdaCalculus37gevaerts: I'm not sure if it'll work in Rockbox, either. If some brave soul is willing to step forward and try it, all the more power to them.
20:48:42*gevaerts would try, but he doesn't have a MK1231GAL :)
20:49:35*LambdaCalculus37 doesn't have one, either :)
20:50:04LambdaCalculus37Anyone want to buy me a new iPod Classic so I can cannibalize it and take the drive? ;)
20:50:16LambdaCalculus37Then we can try this out.
20:52:41rasherCan anyone give me a good reason why the background logic for the sim isn't reversed? New users are pretty much *always* mystified as to how to control the thing.. experienced user could run it with −−nobackground if it annoys them
20:53:35funmanrasher: nobody did commit a change to this logic
20:53:42kugelrasher: pf only searches for the first song per album to search album art. Artist might differ though for the tracks, but pf isn't able to show AA on a per-track basis anyway, so I'd just set albumartist
20:55:29ReizoQuestion: You guys "specialize" in hacking/modding ipods?
20:56:43 Quit Reizo ("CGI:IRC (EOF)")
20:57:01 Join Reizo [0] (n=4232bff8@gateway/web/cgi-irc/
20:57:37krazykitReizo, not really. rockbox runs on several devices, and ipods are simply one of them
20:57:48 Quit Reizo (Client Quit)
20:57:53 Quit itcheg (" ajax IRC Client")
20:58:14rasherkugel: You should check if there's an albumartist set though
20:59:02linuxstbrasher: Reversing the background logic would get my vote.
20:59:09kugelrasher: albumartist is always set (per fall back from artist)
20:59:17 Join {phoenix} [0] (
21:00:33rasherkugel: Ah, alright.
21:02:54funmanpixelma: indeed chopper is much more easy in archos mode !
21:03:19funmani think the check should be made on height <= 64, not width
21:04:10 Join Bensawsof [0] (
21:06:11 Nick Bensawsof is now known as Bensawsome (
21:08:03kugelrasher: I got it
21:08:18*domonoky added a more detailed status table on, other sansa-ams devs should feel free to fill in more details :-)
21:08:58rasherkugel: the artist?
21:08:58 Join einhirn [0] (
21:09:11kugelrasher: all, I just do get_metadata actually :)
21:09:35rasherSo id3 is fully populated?
21:09:58kugelrasher: yes
21:10:09rasherThat should be more future-proof also
21:10:35kugelthat doesn't harm though, and makes the fix relatively easy. and yes, it's surely future-proof (with respect to a possible database integration)
21:10:37rasherkugel: You sould probably open a separate task for that, since it's a bugfix that can go in now
21:10:42 Join freddy__ [0] (
21:10:48 Quit freddy_ (Read error: 110 (Connection timed out))
21:10:59LambdaCalculus37domonoky: Nice! I've got a couple of things to mention on the table.
21:11:00kugelrasher: uhm, my pf patch fixes 2 other bugs as well, they might be go in too
21:11:09*LambdaCalculus37 goes to add what he needs to add to the table
21:11:20kugel(not the features, of course)
21:11:39LambdaCalculus37After funman gets done first. :)
21:11:57rasherkugel: Then you should at the least separate the bugfixes from the features and open a task for those
21:12:15 Join itcheg [0] (i=62db4767@gateway/web/ajax/
21:12:16rasherA task for each would be preferable I guess, but if they're small fixes, a fix-em-all patch is fine
21:12:27kugelrasher: yup, that's exactly like in the last freeze, where the first load of my pictureflow bug fixes was committed :)
21:13:43 Join mc2739_ [0] (
21:18:16kugelrasher: "artist.bmp" in /,rockbox/albumart is valid, isn't it?
21:18:21funmanthe fuzev2 firmware seems to have the same format than clipv2
21:19:32rasherkugel: Nope, that won't get picked up - check out apps/recorder/albumart.c line 154
21:19:36funmanand the same structure than other firmwares ("AS3525" string)
21:23:01kugelbut artist-album.bmp it seems
21:23:18kugellets try that
21:23:34rasherHow else would you get the right album art if you have more albums by one artist?
21:24:05kugelgood point
21:24:53funmando you have a link to a list of the flags in the program status register on ARM ?
21:25:39kugel2nd third of the right side page
21:26:26kugeluhm, that one should work better:
21:26:29 Join Aurix_Lexico [0] (n=comrade@
21:27:42funmankugel: thanks!
21:28:06 Quit jhulst (Read error: 113 (No route to host))
21:28:10kugelyou're welcome :)
21:28:52*kugel didn't expect a floating point exception in pictureflow
21:29:29rasherFS #9626 - Reverse simulator background logic - up for discussion
21:29:51pixelmaI would be for reversing too
21:30:24 Quit mc2739 (Read error: 110 (Connection timed out))
21:32:03*bluebrother was thinking about adding WASD-style button mappings to the sim for those guys that don't have a num pad
21:32:21rasherThe arrows work
21:32:39rasherMaybe the number row should just map to the numbers as well?
21:32:48funman < more detailed
21:33:18 Nick mc2739_ is now known as mc2739 (
21:33:41gevaertsrasher: I think reversing the logic is a good idea
21:34:37 Quit kharo1 (Read error: 110 (Connection timed out))
21:35:05funmanthe fuzev2 also use the armv5 BLX instruction, so it suggests and upgrade to the as3525
21:35:33funmanbecause i can see what seems access to as3525 registers (same addresses)
21:36:36 Join kharo [0] (
21:39:45 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
21:40:24*bertrik wonders if we could make the radio screen a little prettier
21:43:03 Quit bmbl ("Woah!")
21:43:17 Join akur [0] (
21:45:41kugelrasher: Ok, it works. It finds my artist-album in /.rockbox/albumart
21:46:57LambdaCalculus37bertrik: It would be a nice idea.
21:47:34rasherkugel: smashing
21:47:49gevaertsI think FS #5744 should be rejected. Configuring buttons is a specific NODO, and reversing the behaviour sounds like a bad idea to me
21:48:42bertrikfor example, we could quite easily give a visual indication of the "location on the dial" using a horizontal slider
21:49:25rashergevaerts: I agree
21:50:08kugelgevaerts: I think this one is a special case. And it doesn't really change the functionality of the button imho, rather "what happens when usb is inserted"
21:51:47bluebrothergevaerts: agreed
21:52:32gevaertskugel: I'm not sure how you define "button funtionality" then
21:52:58rasherkugel: it configures how Rockbox interprets the buttonpress, I'd say that counts
21:53:05*gevaerts adds a c before anyone notices
21:53:21kugelwell, it changes the behaviour if you don't press that button at all
21:53:29bertrikyou put the fun in functionality
21:54:47kugelI think button configurability implies that that the you get different results when you press the button _only_. With this one, you don't press the button and still change behavior
21:55:20kugeland yea, it's annoying imho that you'll need to press select (which leads you to the context menu) in order to NOT reboot
21:56:10rasherkugel: It's the expected behavior in most cases, for most users though: Attaching cable enables USB connection.
21:56:48rasherI'd fight a reversal of this logic tooth and nail, since it'd "break" USB in the eyes of a lot of users
21:57:41rasherkugel: your patch is unclean
21:57:58kugelrasher: I tend to claim that most rockbox users aren't able to use rockbox usb (ipods and sansas have the biggest user base as per Bagder's statistics)
21:58:35rasherkugel: I'm not talking about rockbox usb, but about rebooting to the OF that has USB
21:58:37gevaertskugel: are you going to do the support for all the people who complain that their ipod doesn't connect?
21:59:05kugelgevaerts: I didn't vote for reversing it, but I'd surely like to have the option
21:59:21kugelif the default is kept, no user will complain
21:59:29 Quit LambdaCalculus37 (" go home time is now")
21:59:48kugelanyway, if it goes under configurable buttons, then it should probably rejected
21:59:58kugelrasher: I answered
22:00:35rasheroh, yuck
22:00:49 Quit funman ("leaving")
22:02:25 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
22:05:43 Quit jhulst (Remote closed the connection)
22:06:06kugelrasher: Ok, I should've mentioned in the first post (I forgot about it), but don't you agree to what I answered?
22:06:31rasherYup, the yuck was at the custom scrollwheel design
22:07:32 Join Zagor [242] (n=bjst@rockbox/developer/Zagor)
22:09:50rasherI'm unsure if the scrollwheel thing counts as a bugfix
22:10:00rasherProbably not, really
22:10:35kugelimho it does...
22:10:51rasherWhere's the bug? Everything works as expected/intended
22:10:59kugelbesides that it's totally safe
22:12:13rasherIt does seem pretty obvious
22:12:21 Quit Llorean (Read error: 131 (Connection reset by peer))
22:13:05kugelrasher: visit the task to get a bugfix only version
22:13:44*rasher debates a bit back and forth
22:14:58 Quit itcheg (" ajax IRC Client")
22:15:04kugelI don't mind either way. I'm sure that that FS #8335 will get committed some time after the release, and that can contain the scrollwheel stuff
22:15:05gevaertsbluebrother: is FS #5737 still relevant?
22:17:10*bluebrother spots an "okok" in the tracker ...
22:17:33 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
22:17:39bluebrothergevaerts: yes. I still want to get it updated and committed.
22:17:48gevaertsok. I won't touch it then
22:17:54bluebrotherthough I'm a bit unsure if the save format is the best idea
22:18:09kugelbluebrother: Anything against a "okok"? ;)
22:18:14*gevaerts now looks at rasher with FS #5519
22:18:18bluebrotherif you have a good idea about it let me know ;-)
22:18:36bluebrotherkugel: shall I call you Casainho? :P
22:18:53rasherkugel: What about the MAX_PATH+1 change? Most uses of MAX_PATH don't do this.. Though some do.
22:18:54gevaertsI'm just going through the patch tracker to see if anything can be closed
22:19:36rashergevaerts: People claim it takes ages.. I haven't tested it since I first wrote that patch though.
22:20:08bluebrotherI'm planning to try to look into rbutil tasks during my holidays. And give the h100 remote a new shot. Maxim chips arrived :)
22:21:06 Join akur1 [0] (
22:21:18 Quit akur (Read error: 110 (Connection timed out))
22:22:00Zagorrasher: I agree with you about simulator background. it should be enabled by default
22:24:35kugelrasher: it should be +1, for the \0
22:25:27kugelelse you dont have really max path available, but max path -1
22:26:04rasherkugel: And you're sure that's not already included in the value of MAX_PATH?
22:26:16bluebrotherwell, does MAX_PATH include the terminating NULL? IIRC POSIX was unclear about that.
22:27:11*rasher is going to leave that in - won't do any harm
22:27:16kugelrasher: max path is max path, it's a convention to not include the \0 in that symbol, from what I've heard
22:27:52Zagorre #5744, how does car adapter mode work with usb-powered targets?
22:27:57rasherMAX_PATH doesn't consider multi-byte characters anyway, so it's already awfully broken
22:30:08gevaertsZagor: no idea. As far as I know there is no car mode related code near the place where the button is checked in usb.c
22:31:43 Quit perrikwp (" ajax IRC Client")
22:33:59kugelbluebrother: don't! :)
22:37:09***Saving seen data "./dancer.seen"
22:40:52 Join Chronon [0] (
22:43:23ChrononI notice that there are a number of dead entries on the TWikiUsersGroup page. Are there any objections to me removing them?
22:45:05bluebrotherok ... -> a filename is up to MAX_PATH including the terminating NULL. The Filename entry instead says "1 to {NAME_MAX} bytes used to name a file" and doesn't mention the NULL at all
22:45:29ZagorChronon: go ahead. I did a rather big cleanup a few weeks back and never got around to clean that page
22:45:45Chrononediting. ..
22:47:14rasherbluebrother: I'd say that MAX_PATH most likely *should* include the terminating NULL. If nothing else, then to avoid trouble.
22:47:24 Quit massiveH ("Leaving")
22:48:00bluebrotherrasher: yep. My point was just that the standard itself is unclear about it :)
22:48:02*amiconn returns, and sees LambdaCalculus pinged him...
22:48:16 Quit bmbl ("Woah!")
22:48:40gevaertsWhat should we do with FS #8266?
22:49:00rasherbluebrother: It also seems more likely that the standard things the NULL should be included, since it specifically says that once, and just omits mention the other time
22:49:14kugelwell, if you do sizeof(MAX_PATH) to know how much chars you can use for your filename, then you should get it (what you see is what you get, anyone?). And not that, effectively reduced by 1 imo
22:49:56bluebrotherrasher: well, I bet there are other places where it explicitly tells it does not include NULL.
22:49:57 Quit {phoenix} (Remote closed the connection)
22:50:00kugelAlso, I've seen +1 more in the rockbox source, especially in the settings
22:50:10bluebrotherMAX_PATH is a macro, so sizeof(MAX_PATH) doesn't make sense
22:50:12gevaertskugel: sizeof(MAX_PATH) is going to be 4 or 8 ;)
22:50:15Zagorkugel: string variables always have room for a terminating null. that is not unexpected behaviour.
22:51:02kugelgevaerts, bluebrother: ok, "imagine you could do sizeof() on MAX_PATH"
22:51:23bluebrotherkugel: you can do it ... it just doesn't yield what you want it to do ;)
22:51:28 Join tessarakt [0] (
22:51:56 Quit XavierGr (Read error: 110 (Connection timed out))
22:52:08 Quit mc2739 ()
22:52:31kugelthat makes me wonder how it's defined
22:52:41rasher#define MAX_PATH 260
22:52:46bluebrother*la la la*
22:53:09rasheramiconn: Your take on the MAX_PATH thing?
22:53:12kugelbluebrother: singing shakira is strictly off-topic :)
22:53:45Zagorrasher: can I get a quick recap of the issue? are you wondering if variables should be declared with or without +1?
22:53:46kugelrasher: oh, thanks for that info
22:54:16kugelZagor: yep, that's what we all wonder and discuss
22:54:18rasherZagor: Pretty much. Also, how we handle multi-byte characters
22:54:46CaptainKwelHi - so aside from the codec buffer, is there any substantial block of memory a codec can exploit to use for some internal memory stuff?
22:55:25bluebrotherhmm, according to limits.h definitions it does include NULL. Oh my.
22:55:41 Quit petur ("*plop*")
22:55:44bluebrotherkugel: you sure it's Shakira?
22:55:57ZagorI definitely think MAX_PATH includes \0
22:55:58kugelnot anymore ;)
22:56:56kugelgood that PATH_MAX != MAX_PATH, else I would've been proven wrong :D
22:57:25*kugel hides
22:57:27*bluebrother now listens to a track called "La La La" brought by player ...
22:57:54Zagorsnprintf takes the size including 0, for example
22:58:43kugelI see, I see, I should renew my C references as soon as possible
23:00:15rasherI wonder how things would look if we did MAX_PATH = 780 (iirc, characters can be up to 3 bytes?)
23:00:29kugelZagor: and how about MAX_FILENAME? It's been used with +1 in settings.h
23:01:15 Join itcheg [0] (i=62db4767@gateway/web/ajax/
23:01:24Zagorsame thing, in my opinion. declaring buffers string[MAX_SIZE+1] is ugly
23:02:07 Join EspeonEefi [0] (i=eefi@STRATTON-SIX-FOURTY-SIX.MIT.EDU)
23:02:44rasherFor this reason alone, I think the MAX_* defines should include the \0
23:04:02 Quit itcheg (Client Quit)
23:08:42amiconnrasher: Things would most probably crash very often
23:09:01 Quit Zagor ("Leaving")
23:09:20amiconnAnd UTF-8 chars can be more than 3 bytes (but only for chars outside the unicode bmp)
23:09:38rasheramiconn: Why's that?
23:10:05amiconnMany buffers for that kind of stuff are allocated on teh stack
23:10:38ender`utf-8 characters are allowed to be up to 4 bytes (though originally 6 were allowed)
23:11:46rasherI think restricting to the BMP is acceptable. But I can see how allocating the better part of a KiB on the stack (often several times, I'll bet) would be problematic
23:12:52amiconn.bdf doesn't support extended unicode anyway
23:13:00 Quit akur1 (Read error: 60 (Operation timed out))
23:15:17rasherI think we can safely declare anything past the BMP "the boonies", and refuse to support it
23:16:42rasherStill, it seems a bit wrong that we potentially have a limit on MAX_PATH of less than 90 characters.
23:17:48amiconnIirc the fat driver switches to using shortnames if the longname doesn't fit
23:18:10rasherAh right, the shortname! A lucky escape for FAT32 there
23:18:33 Quit freddy__ (Remote closed the connection)
23:21:13amiconnWe might be better off using UTF-16 for pathnames internally.
23:22:00*rasher would think UCS-2 instead?
23:22:40rasherAvoid all the surrogate pair stuff, since we don't support that anyway
23:23:43 Join MethoS- [0] (
23:27:01 Join BenCrinion [0] (n=5ec0f2a2@gateway/web/cgi-irc/
23:27:42BenCrinionhi, did that show?
23:28:21gevaertsBenCrinion: yes
23:28:28 Quit karashata ("G'bye everyone!")
23:29:57 Join akur [0] (
23:30:04 Quit BenCrinion (Client Quit)
23:30:27 Part akur
23:39:36 Join BenCrinion [0] (
23:39:58 Join sajes [0] (n=sajes@
23:40:06BenCrinionhi, is there anybody out there
23:40:13bluebrotherno, only bots in here
23:40:29bluebrotherand those are in, not out
23:40:45bluebrotherhi, I'm Eliza. Please tell me your problem.
23:40:53BenCrinionso, is there any chance of getting rockbox help in here?
23:40:58BagderBenCrinion: fire away your prob!
23:40:59BenCrinioni installed the latest version
23:41:09BenCriniononto an olympus mrobe
23:41:16BenCrinionand it works, but i get no sound
23:41:24bluebrotherInteresting. How do you feel about olympus mrobe?
23:41:36*gevaerts hits bluebrother over the head
23:41:42gevaertsBenCrinion: let me try on mine
23:41:44BenCrinionit looks like its playing, i can fast forward, upload
23:42:09BenCrinioni liked the mrobe until i broke it by deleting some files
23:42:16bluebrotherHave you feeled upload before?
23:42:18BenCrinionand now i dont like it any more, cos its broke
23:42:28BenCrinionfeeled upload?
23:42:46gevaertsBenCrinion: ignore bluebrother. He seems to be in a not very helpful mood today :)
23:42:49pixelma/ignore bluebrother ;)
23:42:54bluebrotherBenCrinion: ignore me. I'm playing Eliza
23:43:05BenCrinionits fine
23:43:58bluebrotherso, what kind of files did you delete? Have you chkdsk'ed the player?
23:44:23BenCrinioni reformated the player, then rehooked it up to the mtrip
23:44:31BenCrinioni havent chkdsked it
23:44:38BenCrinionim confident that its fine
23:44:49BenCrinionit was playing ok with mrobes firmware yesterday
23:44:59bluebrotherhmm. Well, if you reformatted it there shouldn't be issues with crosslinked files
23:45:06gevaertsIt's a bug
23:45:15bluebrotherdoes the m:robe firmware play fine today?
23:45:20gevaertsI don't get any sound either with the latest build
23:45:34gevaertsBenCrinion: did you use Rockbox Utility to install?
23:45:42BenCrinionno, it didnt work
23:45:48*bluebrother hasn't tried the m:robe for a while
23:45:50BenCrinionsomething about cannot find firmware
23:46:07*jhMikeS did broke it? (somehow)
23:46:09BenCrinionmaybe cos the original files were missing
23:46:11gevaertsOK, just checking what I should say next :)
23:46:13bluebrotherlet me guess ... the original firmware file wasn't present?
23:46:25BenCrinioncos i formatted it
23:46:39gevaertsBenCrinion: did you install the current build, or 3.0?
23:47:01BenCrinioni think anyway
23:47:06bluebrotherwell, that file is required as otherwise it can't make booting the OF (and thus disk mode) work
23:47:09BenCrinionthe link from the releases page
23:47:39gevaertsJust to check, can you go to the System menu and select Rockbox info?
23:47:43BenCrinionit boots fine
23:47:59 Quit kugel ("ChatZilla 0.9.84 [Firefox 3.0.4/2008111319]")
23:47:59*gevaerts has sound now...
23:48:15jhMikeSgevaerts: what was the problem?
23:48:15gevaertsThat's 3.0, yes
23:48:17 Join kugel [0] (n=chatzill@unaffiliated/kugel)
23:48:25gevaertsI rebooted
23:48:41 Quit CaptainKwel (" ajax IRC Client")
23:48:44BenCrinionrebooted the mrobe?
23:48:48BenCrinionwith the rest button?
23:48:53 Quit jgarvey ("Leaving")
23:49:02bluebrotherthat's not reboot, that's reset
23:49:27gevaertsActually the current build froze on the USB screen, so I held power for a while to reset it
23:50:32BenCriniongot sound
23:50:38BenCrinionwhat the F?
23:50:45gevaertsGood question...
23:50:45BenCrinionim sure i turned it off already
23:51:08gevaertsThere must be a bug that makes sound break now and then
23:51:28*jhMikeS sees something in wm8751.c that should be corrected anyhow
23:51:32BenCrinionif it makes any difference i just activated the database whatever that means
23:52:18BenCrinionthis is really good guys, great job
23:52:26gevaertscan you try something?
23:52:57gevaertsIf you connect it to USB, so it switches to the OF disk mode, and then disconnect and reboot to rockbox, do you have sound?
23:53:38gevaertsI seem to have sound on cold boot and on reboot from rockbox, but not on reboot from diskmode
23:53:51BenCrinionwhats OF disk mode?
23:54:06BenCrinioni take it thats where it mounts as a hd?
23:54:23gevaertsOF stands for Original Firmware. Rockbox doesn't have its own USB code yet, so it uses that
23:54:38*gevaerts submits a bug report
23:55:09BenCrinionive still got sound
23:55:35bertrikkugel, have you experimented yet with the i2c thing, maybe it could be some kind of I/O expander thingy that is involved in button reading
23:56:10kugelbertrik: I gave funmans patch a quick test, without success though
23:56:27bertrikwhat patch?
23:56:28kugelthe read returns -4, and button passed to it doesn't change
23:57:45gevaertsIt seems to be a bit more complicated than that. Anyway, rebooting helps so it's not too bad
23:57:47 Join helpo [0] (i=4b3a331d@gateway/web/ajax/
23:58:09BenCrinionright guys im off, thanks for your help
23:58:09helpoi have a sansa sandish e260R
23:58:23helpoi am not sure how to install it
23:58:26 Part BenCrinion
23:58:39helpoit has wierd instructions, can anyone help?

Previous day | Next day