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

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

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

#rockbox log for 2008-11-16

00:00:25*amiconn isn't 100% sure, but has no real motivation to find out
00:01:17 Quit robin0800 (Remote closed the connection)
00:02:30saratogaarea sampling is nice to have, though arguably it isn't needed
00:02:38saratogawould still be nice for targets with a lot of memory or flash disks
00:02:50saratogaperhaps it should be a compile time choice
00:03:18*kugel remember the wanted bmp resize shouldn't be only fast and small, but also give pretty results
00:03:47 Quit bluebrother ("leaving")
00:04:45LloreanPretty results are a bit less important, since if someone really wants the prettiest results they can resize in advance.
00:04:53LloreanThe goal should be "not bad looking results"
00:07:45 Quit {phoenix} ("Konversation terminated!")
00:08:14kugelsaratoga: I get 15k increased ram usage here
00:08:40 Quit merbanan (Remote closed the connection)
00:13:26jhMikeSamiconn: I looked at the beast voice issue and the PCM buffer is being filled and drained so for some reason voice simply isn't being heard.
00:14:25linuxstbamiconn: The patch on the tracker uses two cache lines for bilinear scaling, so yes, it's doable.
00:14:44saratogai really have no idea how big of a deal proper resizing makes, but I think having a compact and fairly well optimized high quality resampler is worthwhile, even if we don't enable it on most devices
00:15:17saratogaresizing algorithms are interchangable anyway, so theres little cost to having a few and picking the best one based on device resolution and memory constraints
00:17:15saratogakugel: yeah i've gotten a few numbers out of it, not sure why
00:18:03kugeli think 15k sounds reasonable if it's able to do 2500x2500
00:18:07linuxstbDoes anyone know how jpeg images are decoded? i.e. in what order pixels are output? I'm wondering what size temporary buffer is needed prior to resizing.
00:19:27amiconnlinuxstb: Further resizing jpeg without decoding the whole image first is very difficult
00:19:35saratogalinuxstb: they're stored as 8x8 blocks which are inverse transformed
00:19:44saratogai believe left to right, up to down
00:20:07amiconnPixels are output in macroblock order, but jpeg is YUV, and afaik Y, U and V are separated
00:20:25amiconnAlso, there are several possible chroma subsamplings
00:21:01saratogaif theh color samples are not interleaved this could be annoying
00:26:30saratogagoogle says images can have interleaved channels, but i'm not sure if they have to
00:26:50saratogathough it would seem quite stupid to transmit each color channel seperately
00:34:45*linuxstb does a test compile with the jpeg decoder (and the yuv2rgb conversion) in the core, and it adds about 14KB
00:55:11 Quit HellDragon (Client Quit)
00:56:43 Join HellDragon [0] (
00:56:45 Join lasser [0] (
01:05:43 Nick JdGordon|zzz is now known as JdGordon (n=jonno@rockbox/developer/JdGordon)
01:06:08 Quit tessarakt ("Client exiting")
01:06:15 Join domonoky1 [0] (
01:15:53 Quit saratoga ("CGI:IRC (EOF)")
01:17:15 Quit Rob2223 (Read error: 104 (Connection reset by peer))
01:17:44 Join Rob2222 [0] (
01:18:39 Quit domonoky1 (Read error: 104 (Connection reset by peer))
01:21:00 Quit ajonat ()
01:23:38 Quit domonoky (Read error: 110 (Connection timed out))
01:30:52 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
01:33:32 Quit lasser (Read error: 110 (Connection timed out))
01:35:55 Quit ender` (" In the west, Eastern religions are exotic and have strange powers. In the east, Western religions are exotic and have stran")
01:45:41 Quit XavierGr (Nick collision from services.)
01:45:54 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
01:45:55Unhelpfulif you're going to just do area averaging, and you're only averaging whole pixels... you could do just one line, but not of pixel values.
01:46:05***Saving seen data "./dancer.seen"
01:46:33Unhelpfulhave an accumulator line that you add values to, 32-bit values should be good for any reasonable scale, i'd think
01:47:00Unhelpfulthen divide the values in it after you've summed up however many lines you need to to get one output line
01:47:59Unhelpfulyou could even multiply the first and last lines that go into it by some scaling factor if they're not "all the way" in the region of interest
01:48:02 Join gabe565 [0] (n=440c6039@gateway/web/cgi-irc/
01:48:10 Quit gabe565 (Client Quit)
01:48:34kugeljhMikeS: ping
01:53:53gevaertsNico_P: if you split an mp3 file with a frame-level splitting tool (such as cutmp3, but there are lots), the resulting files play back without any noticeable noise or gap in rockbox
01:53:55kugelNico_P: we've played around with splitting mp3 files. the resulting files have a strange noise between each other in just about every media player except rockbox
01:54:48gevaertsSo the question is, why is rockbox different?
01:54:48Nico_Pinteresting... what do you think it could mean?
01:55:20Nico_Phonestly I wouldn't know... the parts of the playback engine I dealt with are higher level
01:55:22kugelwell, my guess is that rockbox skips 1-2 frames
01:55:41gevaertsYou'd hear that I think
01:55:49Nico_Pthe answer is most likely in the MP3 codec
01:56:13gevaertsIf the mp3 codec doesn't clear its state between tracks, you'd get this effect I think
01:56:40kugelgevaerts: I think you only hear that if those frames are noticeable different from the surrounding ones
01:57:08LloreanIf it regularly skipped a frame or two, wouldn't gapless never work?
01:57:18soapAre you talking about splitting A into A1 and A2 and then playing A1 followed directly by A2?
01:57:34soapor A1 followed by B1?
01:57:58Unhelpfulmpeg frames aren't overlapped, are they?
01:59:08kugelLlorean: do you hear that 1 frame is missing if they're very similar to the surrounding ones? I say you only hear that if that particular frame is very different
01:59:13soapyes, frame 1 is smeared into frame 2
01:59:34soapelse you would get level jumps at the boundry, no?
01:59:56Lloreankugel: Considering the kinds of bug reports we get regarding gapless, people have paid *very* careful attention to gapless under a variety of circumstances.
02:00:19gevaertsSo just cutting between frames would be expected to create audible artefacts I guess
02:01:29 Quit culture (Connection timed out)
02:01:37linuxstbgevaerts: I'm pretty sure Rockbox's mp3 codec doesn't clear the state between tracks - to enable gapless... apps/codecs/mpa.c should give you the answer though.
02:02:27Lloreangevaerts: The players you tried, any of them targeted at gapless playback?
02:02:44kugelwhat's the backdraw of not clearing the state? obviously pc apps don't do that (given that this is causing our issue)
02:02:55Lloreankugel: Most PC apps don't playback MP3s gaplessly
02:03:27kugelthe one I tried does gapless, but not in this case
02:03:33LloreanGapless MP3?
02:03:39LloreanOr just "gapless" in formats that natively support it?
02:03:49gevaertsLlorean: I don't think so. You hear a real noise however, not just a gap
02:04:04kugelmp3 too, I've a few live albums in mp3 which are gapless in both the pc app and rockbox
02:04:05Nico_Pkugel: have you tried with FB2K?
02:04:15gevaertsAnd you also hear it when just playing the second file
02:04:38Lloreangevaerts: And you don't hear it when just playing the second file from cold stop in Rockbox?
02:05:09kugelgevaerts: yea true
02:05:15soapalso remember the bit reservoir
02:05:32soapif frame A has room and frame B needs bits, it will store them in A
02:05:39Lloreangevaerts: My point about gapless, though, was that it may be something *we've* changed/fixed to improve gapless support in the past (maybe fixing some other bug) that they haven't had to address.
02:05:55soapso if you split between A and B, B won't have all the data it needs.
02:06:59gevaertsLlorean: I _think_ I hear it
02:07:22*gevaerts didn't hear it last time because of fade-on-pause being on
02:07:55LloreanSo it could be the lack of codec reset?
02:08:18gevaertsThat's what I think, but of course as you said this may be intentional
02:10:10kugelNico_P: No I haven't tried that
02:10:43Nico_Pit has explicit gapless support, so my guess is that you won't hear the artefact
02:10:55kugelok, let me try
02:14:15kugelNico_P: No, it doesn't show these artifacts
02:14:38gevaertsSo is this a bug or a feature?
02:14:40kugelmedia monkey does mp3 gapless though with showing these artifacts
02:16:18Lloreangevaerts: Sounds like a feature.
02:16:53 Nick fxb__ is now known as fxb (
02:17:10Lloreankugel: It looks like MediaMonkey might use the "remove silence" method of gapless.
02:17:22gevaertsLlorean: yes, until someone gets artefacts at the beginning of a correct file when it's played straight after some other files
02:18:08*gevaerts isn't convinced that this can't happen
02:18:19kugelLlorean: possibly, it achieves gapless playback
02:18:37Lloreankugel: "Achieves the appearance of gapless" and "has true gapless" are two very different things.
02:19:06gevaertsWasn't true mp3 gapless supposed to need special tags?
02:19:07Lloreangevaerts: I'm not convinced it's impossible either.
02:19:10LloreanIt does, yes.
02:19:13*gevaerts is ignorant
02:19:22LloreanIt needs either special tags, or MP3s that end on frame boundaries (IIRC)
02:19:45kugelhow's that different? no gap is no gap.
02:20:13Lloreankugel: "Silence detection" can mean removal of silences that's supposed to be there, for one thing.
02:20:21LloreanBasically, it can end up reducing or removing a gap that's intentional.
02:21:17kugel"true gapless" is questionable anyway when it comes to mp3
02:21:22LloreanNo, it's really not.
02:21:42LloreanWith the LAME headers it's an accurate statement.
02:22:04kugelwith lame headers maybe
02:22:17kugelbut not with mp3 in general
02:22:33LloreanI don't understand your point.
02:23:08kugel"there's one specific encoder which can do gapless" doesn't make it unquestionable. The format itself isn't designed for gapless, so it'll always be questionable
02:23:23 Join webguest73 [0] (n=50bb64d3@gateway/web/cgi-irc/
02:23:24LloreanThere is NO question as to whether MP3 can be done gaplessly
02:23:30LloreanThe only question is whether users WILL do it gaplessly.
02:25:06Lloreangevaerts: Should the codec maybe try to determine if the songs are sequentially from the same album?
02:26:25gevaertsLlorean: I think it should, as well as possibly checking for LAME headers. I'm not sure whether it's worth the extra complexity though
02:26:38Lloreangevaerts: It does check for LAME headers, because we make use of them.
02:26:47Lloreanwebguest73: If you have a question, please go ahead and ask.
02:28:59Lloreangevaerts: I'm kinda surprised MP3 depends on previous frames to avoid artifacts though. Doesn't that, in a way, break its ability to be streamed cleanly?
02:29:57gevaertsGood point. I'm a codec n00b though, so I don't actually know
02:30:32gevaertsLlorean: I meant, maybe it should reset state if no LAME headers are there. That should still allow frame-level gapless, _if_ the files are fully standalone. Of course, this entire discussion shows that that may not be a safe assumption
02:31:20Lloreangevaerts: I'd add a little more complexity (if codecs can be aware of shuffle status) with "always reset if shuffling, never reset if sequential unless current album != previous track's album"
02:32:42gevaertsLlorean: now how do you check the "sequential" bit reliably? IIRC those track number tags aren't always very well-formed
02:32:42LloreanThere's still cases where it could artifact but it'd reduce the odds a lot while still, I think, ensuring cases where it's supposed to be gapless always will be?
02:32:57LloreanSorry, by "sequential" I meant "if shuffle is off"
02:33:03gevaertsah, ok
02:33:12LloreanBasically, don't worry about the order of the tracks. If shuffle is off, and they come from the same album, assume they're "in order"
02:33:24LloreanIt's a little dumb, but it's less complex.
02:34:00*gevaerts thinks
02:34:36gevaertsI think it will work, as in _very_ unlikely to not play gapless when it
02:35:02gevaertss supposed to, and somewhat unlikely to attempt gapless when it shouldn't
02:35:38gevaertsNow the question is, does the codec have access to all needed info?
02:35:49LloreanNo clue. :)
02:36:21jhMikeSkugel: ring
02:36:23 Quit webguest73 ("CGI:IRC (Ping timeout)")
02:36:47kugeljhMikeS: I just wanted to get your attention on FS #6800
02:37:39kugelLambdaCalulus37 and me wanted to talk with you a bit about that patch in order to get it committed
02:38:13Zagornew make patch uploaded to FS #9534. please try and break it.
02:38:52gevaertsLlorean: as far as I know there's still the possibility that there is no real problem though. Maybe an mp3 frame can say something like "reset state", and maybe all "normal" mp3 files start with that
02:39:15gevaertsWho knows mp3 well?
02:39:18Lloreangevaerts: I was wondering the same sort of thing.
02:39:27 Join allele [0] (
02:39:56gevaertsI know mpeg video has that sort of thing
02:40:16alleleif I were to perform rb functions for a plugin from an h file how would I add rb as an argument?
02:40:38gevaertsallele: why would you?
02:41:15alleleso I can have seperate headers for each player and draw things on the screen according to the player
02:41:22jhMikeSkugel: Why is an out of range brightness being changed to MAX? Things get the default if not correct.
02:41:23LloreanjhMikeS: "There still seems to be a problem of cutting of voice samples from the beggining of clips when music isn't playing" <−− I'm pretty sure this is a global swcodec bug.
02:41:38alleleotherwise I guess I could just create a seperate plugin file for each player
02:41:54jhMikeSLlorean: Since when? If it is, it has been introduced.
02:42:08LloreanFor at least half a year.
02:42:20LloreanIt's been brought up dozens of times by people who thought it was a bug in the voice files.
02:42:44jhMikeSLlorean: It is every clip and not just the .talk clips which I though were the problem.
02:42:50gevaertsallele: Does that require calling functions from the header?
02:43:07LloreanjhMikeS: I've certainly experienced it with .voice files as well
02:43:13LloreanOn players other than the S.
02:43:16jhMikeSPerhaps I should check the other players first since I haven't really experienced it with voice menus.
02:43:29LloreanIf you generate voice files using rbutil you should see it.
02:43:40jhMikeSToday was the first time I used voice on the beast.
02:43:47Lloreani've heard reports that voice files generated by the other script don't experience the problem, though I can't confirm this (and think it'd be strange if true)
02:44:05alleleI believe so
02:44:12jhMikeSIt can't be bad frames since it doesn't do that if music is playing.
02:44:31 Quit Zagor ("Client exiting")
02:45:12allelebecause otherwise I would need at least a second set of functions for the sansa player, because the display on that would be rotated 90 degrees clockwise
02:45:22LloreanjhMikeS: Someone mentioned it might be an issue with the zero padding?
02:45:45gevaertsallele: usually this is just handled with some #ifdefs in the c file
02:45:46jhMikeSLlorean: I have had the beast cut some samples from the beginning of tracks too which I have never had on other players.
02:45:52Unhelpfulwhen somebody who knows the hardware is about, could they say if a multiply and add per input pixel, and one integer divide per output, (plus whatever loops end up costing) is reasonable, for image scaling? i'm working up a test at the moment...
02:46:05 Quit Nico_P (Remote closed the connection)
02:46:15jhMikeSLlorean: Every sample should get played, zero padding or not.
02:46:15LloreanjhMikeS: Well, maybe there's two things going on that are similar, but there's definitely been a voice issue like that.
02:46:51LloreanUnhelpful: Scaling should only be done on load, so even a relatively "slow" scale isn't a particular problem.
02:47:49UnhelpfulLlorean: i think i can pull off an area average with about that much computation, and one or two rows of uint32 the width of the image.
02:48:19gevaertsUnhelpful: do I understand your idea correctly if I assume that that multiply is only needed for non-integer scaling?
02:49:06gevaertsI mean something like 500->120, while 500->125 wouldn't need it (as that's exactly 1/4)
02:49:28jhMikeSLlorean: for the beast it could be more than one thing. it could be failing to link the first buffer inserts.
02:49:39Unhelpfulgevaerts: indeed, the multiply would only be for "sub-pixel" contributions at the box edge, and could be tossed entirely if the scale happens to be an integer
02:50:40*jhMikeS just has to investigate and avoid random guessing that may or may not make a bit of sense.
02:51:56gevaertsUnhelpful: Speaking as someone who doesn't know image manipulation from audio codecs, I think your method is certainly worth trying
02:52:35Unhelpfulgevaerts: i'm working on a mock-up, that will just do raw, headerless grey data from stdin to stdout, just to see if it will produce passable output
02:53:12Unhelpfuldo you think it's worth special-casing the integer scale case?
02:53:43LloreanProbably not.
02:54:00LloreanThat'll make it bigger in a situation where the speed gain probably isn't very critical.
02:54:36gevaertsHard to say, but I think it's unlikely enough to not be worth the extra code size for a minimal speed up
02:54:53Unhelpfulalso, i don't think it *needs* to be very tightly coupled to the jpeg codec
02:55:24kugeljhMikeS: others drivers do change it to max
02:55:32Unhelpfulif it takes a line of input at a time, we could have options for 1/2, 1/4, 1/8-scale output from jpeg, and just fudge the input sizes by that amount, and pass lines to the scaler
02:56:08gevaertsI'd start by doing it just for bmp
02:57:10Unhelpfulwell, i'm starting, like i said, with grey data, to make sure the idea works and produces decent output
02:59:55Unhelpfuland probably not finishing tonight, saturday is far to tired a day for me
03:02:55kugeljhMikeS: Do you spot anything else that stops it from getting committed? Maybe you could just give it a test
03:03:14jhMikeSkugel: hmm...some don't do anything at all
03:06:02kugeljhMikeS: Ok, I can rever that change. I'jj did that because the condition was brightness >= array_len (i.e. higher that the max), so it made sense for to stay at the max level if something tries to go beyond that. But going to defaults seems logical as well.
03:08:50jhMikeSkugel: about the defaulting, I think I set it up to saturate at the backlight.c level. Others are to make sure bits aren't illegal.
03:09:19kugelok, sounds fine
03:09:41jhMikeSkugel: oh, it's (unsigned)brightness >= ARRAYLEN() which catches negative values too
03:10:06kugelaww right
03:10:47kugelmr. obvious used a '!=' though
03:11:06kugelscratch that
03:11:12kugelI'm way too tired
03:11:12jhMikeSkugel: I'll apply it and test it instead of trying to read the diff along. I have something else to do first and I'll be back.
03:11:41kugeljhMikeS: thanks.
03:13:27 Quit Gareth (Excess Flood)
03:17:16 Quit moos ("Rockbox rules the DAP world")
03:20:29 Quit allele ("Leaving.")
03:21:58kugelLlorean: I split that mp3 with another tool now (medieval cue splitter), and now there's no strange noise between the tracks in neiter rockbox nor my pc media player
03:26:38 Join reacocard [0] (n=reacocar@WL-112.CINE.HMC.Edu)
03:35:34 Quit DerDome (Nick collision from services.)
03:35:37 Join DerDome1 [0] (
03:35:49 Nick DerDome1 is now known as DerDome (
03:39:47 Join Darksair [0] (n=user@
03:43:05 Quit qwm ("leaving")
03:46:08***Saving seen data "./dancer.seen"
04:21:27 Join Hillshum [0] (
04:22:03 Join Minthe [0] (
04:22:21MintheExcuse me, but did you notice ?
04:24:39MintheExcuse me, but did you notice ?
04:24:50JdGordonno need to repeat yourself
04:24:57 Join blkhawk- [0] (
04:25:30MintheSorry but I got IRC error 477
04:27:04jhMikeSkugel: It's working mostly. If the brightness is 1, then the backlight doesn't come back on. I also wonder about fading brightness changes.
04:29:09 Join miepchen^schlaf_ [0] (
04:30:44MintheAnyway I'm looking forward to any comments for bugs 9536...
04:31:04 Join Davide-NYC [0] (
04:31:37 Quit Minthe ("Leaving...")
04:32:28Davide-NYCI spotted an elusive Sevakis commit! So I logged on and see a bunch of familiar names.
04:32:38Davide-NYCCheers folks!
04:34:13jhMikeSElusive? I'm sure I'll get back on track.
04:36:17 Join LambdaCalculus37 [0] (
04:36:38*LambdaCalculus37 spots a Davide-NYC lurking about :)
04:37:01Davide-NYCI am in and out of focus due to ongoing convo with the "boss"./
04:37:03LambdaCalculus37Davide-NYC: Yep, a Sevakis commit has occured. :)
04:38:27LambdaCalculus37Gotta go for now.
04:38:44*LambdaCalculus37 hopes the beast charging commit comes soon! :)
04:38:45 Quit LambdaCalculus37 (Client Quit)
04:38:54 Quit kugel (Nick collision from services.)
04:38:57 Join kugel_ [0] (
04:39:05 Nick kugel_ is now known as kugel (
04:41:01 Quit blkhawk (Read error: 110 (Connection timed out))
04:41:06jhMikeSheh...well, something will happen as soon as I get some sense of the whys and whats of the OF battery driver
04:41:30 Nick blkhawk- is now known as blkhawk (
04:43:59 Quit miepchen^schlaf (Connection timed out)
04:50:26 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
04:54:58 Quit kugel ("ChatZilla 0.9.83 [Firefox 3.0.3/2008101315]")
04:59:50 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
05:33:17 Nick Darksair is now known as Darksair{away} (n=user@
05:40:19 Quit Davide-NYC ("ChatZilla 0.9.83 [Firefox 3.0.4/2008102920]")
05:46:10***Saving seen data "./dancer.seen"
05:46:42 Nick fxb is now known as fxb__ (
05:52:04 Quit reacocard (Read error: 104 (Connection reset by peer))
05:53:46 Quit Hillshum ("ChatZilla 0.9.83 [Firefox 3.0.3/2008092417]")
06:02:04 Part toffe82
06:04:08 Nick Darksair{away} is now known as Darksair (n=user@
06:07:50 Quit homielowe ()
06:10:30obo_soap: the double entries aren't meant to appear in the log - there is code in apps/scrobble.c which should prevent it
06:24:57 Quit Bensawsome ("The awsome is gone :(")
06:32:15 Join AlHaz [0] (
06:34:32AlHazEvening. Just popping to ask if perhaps someone here has (or knows where i can get) a recent iriver h120 build w/ the rtc hack stuff enabled?
06:34:46AlHazjust had to check if i can save myself the trouble of installing a crossdev environment, etc
06:49:28soapobo_, do you understand my poor attempt at describing when the double entries are happening?
06:50:08 Quit XavierGr ()
06:56:00Unhelpfulhere's my area scaler. reads raw headerless greyscale data on stdin, writes same on stdout. takes input width, input height, output width, output height on command line. works fine, aliasing can be bad.
07:46:11***Saving seen data "./dancer.seen"
07:52:36 Join skipper [0] (
08:04:24 Quit aarcane (Read error: 104 (Connection reset by peer))
08:10:16 Join aarcane [0] (
08:46:19 Quit AlHaz ("Client exiting")
08:53:37 Join Rob2223 [0] (
08:57:38 Quit BHSPitLappy (Remote closed the connection)
09:09:38 Quit skipper (Read error: 110 (Connection timed out))
09:10:32 Quit amiconn (Nick collision from services.)
09:10:39 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
09:10:52 Quit pixelma2 (Nick collision from services.)
09:11:01 Join pixelma2_ [0] (n=marianne@rockbox/staff/pixelma)
09:11:03 Nick pixelma2_ is now known as pixelma2 (n=marianne@rockbox/staff/pixelma)
09:11:50 Quit Rob2222 (Read error: 110 (Connection timed out))
09:16:59 Quit Darksair (Read error: 60 (Operation timed out))
09:23:57 Join Darksair [0] (n=user@
09:31:50 Quit jhulst (Remote closed the connection)
09:44:41 Join sin613 [0] (
09:46:12***Saving seen data "./dancer.seen"
09:48:06 Quit sin613 (Client Quit)
09:52:46 Quit cg (Read error: 145 (Connection timed out))
09:52:52 Join reacocard [0] (n=reacocar@WL-112.CINE.HMC.Edu)
10:09:50 Join cg [0] (
10:11:59 Join n1s [0] (n=nils@rockbox/developer/n1s)
10:15:54 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
10:16:04*JdGordon would like to get the wps mode patch commited
10:17:35 Quit Darksair (Read error: 131 (Connection reset by peer))
10:18:25 Join Darksair [0] (n=user@
10:21:44 Join stoffel_ [0] (
10:28:42 Quit Darksair ("Everything that has a beginning has an end.")
10:40:10LloreanJdGordon: Did people ever really accept that patch?
10:40:17*Llorean doesn't recall if it was talked about on the list.
10:40:23LloreanI mean that's pretty major, remapping controls.
10:40:34JdGordonit was talked about in a few places and mostly got lost in other arguments
10:40:45LloreanWas it talked about on the list?
10:41:01LloreanThis kinda feels like another "people didn't like it so I'm trying to slip it by" thing.
10:41:15LloreanI don't really remember it having much support at all among the devs when I've seen it discussed.
10:42:13*amiconn severely dislikes this patch
10:42:27amiconnIt changes controls, possible even without the user knowing why
10:42:33JdGordonno, it didnt have a thread to itself on the list, but it was discussed a fair amount, and from what i remember it either got "dislike and wouldnt use" or "wait and see what the authors come up with"
10:42:57LloreanJdGordon: So, no dev is in favour of it but you, so you wait until nearly everyone's asleep and commit it?
10:43:20amiconnIirc the outcome was that unless a better way than stealing a button is found, it shouldn't go in
10:43:26JdGordonwell most are waking up now... and im not the only dev who likes it
10:43:53amiconnName any other one?
10:44:05LloreanJdGordon: I seem to remember a *lot* of devs don't like it though.
10:44:14JdGordoniirc linuxstb wasnt against it
10:44:43amiconnI'm not against the idea of having wps modes, but I'm firmly against changing controls, and against configurable controls through the backdoor
10:45:11LloreanI still think just adding a "switch WPS mode" option in the context menu would've been fine.
10:45:16LloreanAlmost as fast, and no stolen controls.
10:45:42JdGordonand I still say that its too slow and makes it less sueful
10:45:43LloreanJdGordon: It really feels like you haven't even tried to address the concerns everyone had with it.
10:45:46JdGordonuseful even
10:45:57LloreanJdGordon: Less useful, maybe. But it also means it doesn't remap controls invisibly.
10:46:04 Join GodEater_ [0] (n=ge@rockbox/staff/GodEater)
10:46:06LloreanIt's still *useful* just VERY SLIGHTLY slower.
10:46:09LloreanOh no.
10:46:18JdGordonnot very slightly at all
10:46:28LloreanIt's ~1.5 seconds slower for a single mode switch.
10:46:37LloreanAssuming you put the option as the top context menu entry.
10:46:45JdGordonyou can flip through every mode in the time it takes to change to the next mode if it has to go through the context menu
10:47:09LloreanJdGordon: 1.5 seconds is "slightly"
10:47:27JdGordontimes by how ever many modes there are... is not slightly
10:47:41LloreanIt's not times however many there are unless you leave, look, switch again, leave, look...
10:47:56LloreanAnd even then, it's a few seconds. Most WPSes won't have more than 2 or 3 modes.
10:48:09LloreanAnd I still think it's a worthwhile price to pay for not silently remapping controls.
10:48:48linuxstbJdGordon: I wasn't against it, but you shouldn't have committed it - it was far too controversial...
10:48:54LloreanBut here you are, changing controls arbitrarily, avoiding addressing concerns, and committing a rather major change without discussing it on-list when over the last few months we've explicitly been trying to get people to talk about patches more on-list first
10:49:04martian67Llorean, there could always be a visual wps indicator
10:49:16Lloreanmartian67: But there ISN'T one right now.
10:49:41Lloreanmartian67: That's the point, the patch as he's committed it contains problems that greatly upset some of the devs and he's chosen not to even attempt to address.
10:51:17JdGordonthats up to themers.... and its not silent, every theme that uses it will make it pretty obvious
10:51:23LloreanJdGordon: They don't have to.
10:51:39JdGordonusers arnt stupid... they'll figure it out quickly
10:51:46LloreanIt's silent, as Rockbox doesn't require the user to know the theme has changed their keymap on them until they attempt to use a key they thought they knew how to use, and it fails.
10:52:01amiconnIt *is* silent for users which unknowlingly load a theme that uses it.
10:52:12*linuxstb doesn't like debating patches _after_ they've been committed...
10:52:13LloreanJdGordon: I'm sure blind users will be able to readily tell their WPS is changing modes, rather than their Rockbox suddenly not talking to them.
10:52:50*Llorean points out that blind users may share an MP3 player, so it's not unbelievable that they may have a WPS set.
10:53:05LloreanOr simply someone using one blindly.
10:53:26amiconnJdGordon: There's another important point: You take away the browse button to switch modes, which means that the browser becomes much slower to enter
10:53:34 Quit J-23 (Remote closed the connection)
10:53:36JdGordonLlorean: and you really think a blind user will use this sort of wps?
10:53:48LloreanJdGordon: Did you read the rest of what I said?
10:53:55 Join J-23 [0] (
10:53:57JdGordonamiconn: thats what the emnu button is for
10:54:06JdGordonLlorean: yes, though after hitting enter
10:54:14LloreanJdGordon: The menu button does not instantly go to the browser, nor does it follow playlist.
10:54:31LloreanYou're complaining about slowed access to the WPS modes, but you're slowing access to the browser in exchange.
10:54:48*Llorean thinks the patch should be reverted until it's at least properly discussed.
10:55:16LloreanNot everyone has to agree, but committing something when people aren't expected to be around because it's controversial really should be discouraged.
10:56:30 Join bertrik [0] (
10:56:38 Quit amiconn (Nick collision from services.)
10:56:45 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
10:57:01amiconnJdGordon: Before your unwanted commit, I could go to the browser with a single keypress. You really expect me to go to the menu, browse to "Files" and enter that? That's at least 4 keypresses, and on wheel targets, it even involves using the wheel, i.e. not having tactile feedback
10:57:18JdGordonbut your not going to use the tag anyway so its irrelevant
10:57:26 Join merbanan [0] (n=banan@
10:57:32amiconnIt's not irrelevant. Re-read what I said earlier
10:58:35n1sHow about using a currently unused button on targets which have one (like rec buttons on some players) and do it with the context menu or quickscreen for others?
10:58:59Lloreann1s: See, this is an idea that could've come up if it had actually been widely discussed...
10:59:02*linuxstb agrees the commiit should be reverted until it's discussed, but will probably argue in favour of the feature
10:59:17JdGordonunused buttons are probably out becuase its nice to keep targets consistant
10:59:26*JdGordon thinks linuxstb is about 2min too slow...
10:59:38LloreanJdGordon: Unused buttons are fine if the context menu option is still there on all targets.
10:59:40JdGordoncontext menu is imo too slow
10:59:48LloreanSo some targets have button+context, others just use the context.
10:59:57LloreanPretty consistent.
11:00:15LloreanJdGordon: But you're telling us to use your new feature, we have to sacrifice existing features.
11:00:24LloreanI can't use your feature, and features I'm used to using, at the same time.
11:00:29linuxstbJdGordon: So you only reverted because it gave warnings?
11:00:35LloreanIMO that's worse than "slow", since you're actively disabling a feature.
11:00:49 Join pixelma [0] (n=Marianne@rockbox/staff/pixelma)
11:01:22JdGordonyour not disabling... you can still go through the files menu...
11:01:46LloreanJdGordon: You've disabled the ability to quickly reach the file browser.
11:01:48JdGordonwould adding a "go to file" option in the context menu be ok?
11:01:58LloreanYou've still disabled the existing feature.
11:02:12JdGordonevery target still can get to it through the menu
11:02:13LloreanWhy is it "I want to destroy your feature for mine" rather than "I want to preserve the existing feature set and add mine on top of it"?
11:02:28JdGordonits only breaking it IF YOU ACTUALLY USE THE TAG
11:02:31LloreanEvery target could get to your additional WPS screens through the menu
11:02:45LloreanJdGordon: As I said earlier, you prevent me from using an existing feature with the new feature.
11:02:55LloreanWhy not make them work alongside? There are several ways to do it as have been suggested.
11:02:56JdGordonthen its up to you which is more important
11:03:10LloreanIf I have to pick which is more important, I'm going to vote against your feature period.
11:03:16amiconnJdGordon: Yes, and I would like to be able to use that tag (as I already said) *without* taking away another favourite feature of mine (the ability to browse quickly)
11:03:24LloreanIt's not impossible for you to design your feature to work alongside existing ones.
11:03:49JdGordonno, but imo one of the main strongpoints is the quick mode changing...
11:04:05LloreanBut apparently not in the opinion of several others.
11:04:45LloreanRockbox should strive not to have mutually-exclusive features.
11:05:35LloreanBut especially not silently exclusive ones.
11:06:36LloreanJdGordon: You gonna wait and discuss it on-list?
11:06:38amiconnDon't forget that there are targets where reaching the menu from the wps takes significantly longer than going to the browser. That does not even include going from the menu to the browser.
11:07:30amiconnAnd the target I have in mind is one of those which would profit most from wps modes (small screen)
11:08:27amiconnJdGordon: On Ondio, wps->browser is short Mode. wps->menu needs 2 times long Mode (first one goes to the context menu)
11:09:25*JdGordon thought the ondio couldnt goto the browser and had to go through the menu?
11:11:05JdGordonLlorean: as for bringing it up on the list... really is there any point? the people who arnt here now are mostly the ones I dont expect to hear from anyway, and then its going to be the same stalemate
11:11:41LloreanJdGordon: So you honestly think only amiconn, linuxstb and I would comment on it?
11:11:44pixelmano, and the way it is (as amiconn explained) is usually quite quick
11:12:05n1sJdGordon: so you see absolutely no other acceptable way to do this but stealing the browser button?
11:12:27JdGordonLlorean: no, bluebrother will say its remapping buttons, alex wallis will say "what does this patch do", ddalton will complain about it breaking voice
11:12:34Lloreann1s: It's another case of "a feature we find important isn't important to him, so he can't understand any argument against sacrificing it"
11:12:56JdGordonditto Llorean
11:12:56LloreanJdGordon: So, your view is "Everyone will complain against it but me, so it should be committed?" that doesn't sound like sound logic.
11:13:13LloreanJdGordon: The difference is, you're talking about REMOVING AN EXISTING FEATURE.
11:13:19JdGordonbecause the quiet users and the those that dont care either way group are much bugger
11:13:24JdGordonno im not
11:13:36JdGordonand the one which could be disabled doesnt even work correctly anyway
11:13:40LloreanI'm sure you'd moan quite a bit if I decided customizable quickscreens weren't important to me and decided to remove them for something *I* thought was better.
11:13:45JdGordonor at least consistantly
11:14:00LloreanWhat do you mean "It doesn't work consistently"?
11:14:05JdGordonLlorean: if you took the time to write the patch, no i wouldnt...
11:14:28LloreanJdGordon: I find that very hard to believe.
11:14:45JdGordonso prove me wrong
11:15:44LloreanIn that case, I'll just wait until you commit this WPS tag, and replace its feature with something I value more.
11:15:49*JdGordon doesnt mind people saying "your way works, this way works better" and then brings up a patch... /me minds people saying "do it this way or not at all"
11:16:08LloreanJdGordon: You're the one saying "the patch has to be this way, or not at all" to everyone who's suggested alternate ways of controlling it...
11:16:22LloreanSo maybe you should be careful about the statements you make.
11:16:36LloreanYou haven't once attempted to compromise.
11:16:40JdGordonbut ive done the patch... do your suggested change and ill be more receptive
11:17:15LloreanSo we should break something rather than waiting until the patch is more acceptable overall?
11:17:52JdGordonif you value this over the browser then its not breaking.. if you dont the patch should be pretty quick...
11:18:38LloreanYes, it's darn quick.
11:18:44LloreanIt's a diff between your commit and the previous version.
11:19:40LloreanI assume the statement about me going for something *I* thought was better applies here too, or is it just for the quickscreen?
11:19:49LloreanAll I'm saying is cool down a bit, and discuss alternatives on the list.
11:19:52 Join lasser [0] (
11:19:58JdGordonhere, everywhere
11:20:01*JdGordon is cool
11:20:03LloreanThere are other ways it can be controlled that mean it's not disabling another feature while in-use.
11:20:17LloreanBut while those of us who don't like the feature are offering compromise, you're saying "It has to go in now, MY WAY"
11:20:28amiconnIt *is* breaking a feature. A feature valued by several people, even though that obviously doesn't include you. It's also breaking consistency, and that was agreed upon to be unwanted.
11:20:33LloreanYou're not even willing to talk about it on the list, and see if maybe somebody *does* want to do the work, or attempt to wait.
11:21:10 Join dany_21a_ [0] (
11:21:45 Quit pixelma2 (Read error: 110 (Connection timed out))
11:22:17amiconnI'm not against the wps modes feature (would btw be especially useful on the Player), but against the side effects of your way to invoke it.
11:23:09LloreanAnd I think the feature is somewhat nice, as long as it never interferes with existing use patterns.
11:23:40LloreanI really like n1s' idea of using the unused record button (or another unused button on say, the S) where possible, and context/quickscreen where not.
11:24:28 Quit n1s ()
11:24:43JdGordonunused buttons makes it harder to get working (and support) but is fine. menu almost makes it too slow to bother with
11:25:12LloreanJdGordon: If the feature is *so* good it's worth sacrificing existing functionality for, surely being able to use it slightly slower is still good...
11:25:20pixelmacontext would be a must then (as the Ondio and Player have no spare buttons and no quick screen) - the unusability is your opinion
11:25:25JdGordonthe modes are not (currenlty) named, so having them in a list in the menu means its hard to find the info you want to see
11:25:43LloreanYou think users can't remember what mode they're on?
11:26:01LloreanInstead of showing a list, you could juts have a "next mode" option in the menu that cycles them, maybe?
11:26:51JdGordonthey know which mode they are on now, they dont know which mode they need to change to to find what they want to see
11:27:02LloreanJdGordon: They do after they've used it once.
11:27:19LloreanWasn't that your excuse for how to deal with the silent keymap change?
11:27:20JdGordondo they?
11:27:42LloreanYou're saying a user can't remember "1 is standard WPS, 2 is large album art, 3 is up next, 4 is lyrics"?'
11:27:57JdGordon"was AA on mode 2 or 3?" "hmm it was on mode 2 except when there is next track info, then its mode 4"
11:28:02 Join bluebrother [0] (n=dom@rockbox/staff/bluebrother)
11:28:08JdGordonor was that the last theme i used?
11:28:25JdGordonit really hurts its usability changing it to a menu
11:28:33LloreanOnly if themes are stupidly designed.
11:28:36*bluebrother wonders if he missed the actual discussion about this new wps tag
11:28:49pixelmayour way also only cycles through the modes...
11:28:57LloreanThere's no sense in having it move if there's next track info, if it's in a menu, since order is irrelevant from a menu
11:29:16JdGordonLlorean: I never suggested themers where smart... jist that users arent stupid
11:29:23LloreanSo any theme that moves it based on whats available is, frankly, poorly designed. We can't help poorly designed themes.
11:29:35LloreanFrankly, "going by sight" is just as bad in that case
11:29:49LloreanOne page could have alternating sublines, so they visit the page and it's not showing what they wanted so they move on and are unable to find it entirely
11:30:02LloreanIt's possibly, always, to design a theme where they won't see what they want sometimes
11:30:09LloreanHell, the information could be hidden so it only shows on hold
11:30:12LloreanUsers *have* to remember some things
11:30:17LloreanIt's a given if a theme is designed such
11:31:04JdGordonsure, but making them remeber is not as nice as just being able to press a button untill what they want is on the screen
11:31:11*bluebrother goes checking the logs
11:31:34LloreanJdGordon: In response "Taking a feature away is not as nice as not silently changing the controls."
11:32:02 Join fml [0] (
11:32:15 Quit fml (Client Quit)
11:32:27 Join fml [0] (n=4fd3ecb1@gateway/web/cgi-irc/
11:32:34JdGordonits not silent.. its documented.. if you use this tag the button behaviour changes
11:32:49LloreanJdGordon: They don't have to remember. It's documented. It's in the wps file.
11:33:03 Quit amiconn (" bbl")
11:33:03JdGordonwe could even add a splash when the theme is loaded "browse button changs wps mode!"
11:33:15JdGordonthe WPS file is not documentation...
11:33:23JdGordonyou have a pretty twisted mind if you think it is
11:33:26LloreanJdGordon: It's at least *with* the player when they use it.
11:33:29bluebrotherIMO all questionable patches should be brought up on the ML before committing
11:33:35LloreanAnd "we could even add strings to the WPS tag so that each page IS labelled"
11:33:50Lloreanbluebrother: At the VERY least "any patch that removes or changes existing behaviour"
11:34:29*JdGordon disagrees... patch discussion should be on the tracker
11:34:35fmlI must say that I'm on JdGordon's side. IMO, the whole point of the modes is to be able to switch really quickly. If you can't then it's nearly worthless. It's like saying: you have a car but in order to use it you have to walk 1km. So the patch lets you decide what's more important to you: browser or the modes. What's wrong with it?
11:34:38JdGordonI will concede that a warning post to the ml is good though
11:34:48bluebrotherbut the ML is kinda neglected for such stuff. Discussion things on IRC cuts out all those people that miss the discussion
11:34:49Lloreanfml: The whole point of the modes is *to be able to see them*
11:34:55_lifelesshello guys, lcd on my player shows rockbox color logo just fine, but colors in rockbox menu is inverted or so, where i should look to fix it?
11:34:59Lloreanfml: Switching quickly is secondary. it's nice, but it is not ESSENTIAL to be able to use them
11:35:22bluebrother_lifeless: there's a setting to invert the lcd on at least all non-color players
11:35:34LloreanJdGordon: Most patch discussion can happen on the tracker
11:35:42LloreanJdGordon: If you're going to undo someone else's work, you REALLY should discuss it on -dev.
11:35:46pixelma_lifeless: what player is this?
11:35:51linuxstbJdGordon: BTW, the reason for your yellow on all ARM targets is that the "current_mode" variable needs to be a "signed char", not just "char".
11:35:52LloreanJdGordon: Changing existing behaviour is undoing someone else's work.
11:36:07_lifelessphilips gogear hdd1620
11:36:29bluebrotherJdGordon: well, discussion itself could also happen in the tracker, but irc is quite bad for this. Also, not all people are watching all tasks on the tracker, so there should at least some "call for tracker discussion" on the ML imo
11:36:31JdGordonlinuxstb: oh? it wasnt the is_remote?1:0 line?
11:36:51*JdGordon didnt really look that hard
11:36:53bluebrother_lifeless: this player isn't supported. There's only work in progress so you have to expect glitches
11:36:58JdGordonreverted with that message to save face :p
11:36:59LloreanJdGordon: Are you going to wait and discuss this some, or are you really going to re-commit once the yellow is fixed?
11:37:01pixelma_lifeless: are you aware that Rockbox on the GoGear is still in early stages of development?
11:37:40fmlLlorean: I disagree. The point is to be able to switch them *quickly*. To see all the info, you have the "show track info" screen.
11:37:58_lifelessyea, i expect it, i'm trying to port rockbox
11:38:00linuxstbJdGordon: That's not helpful −− people should know the real reason for every commit, not a fake reason...
11:38:24bluebrotherbah, who needs real reasons? Or even real discussions?
11:38:40JdGordonor new features...
11:38:55Lloreanfml: Which can't show the album art in different sizes. It can't show a more granular volume indicator. etc, etc.
11:39:23Lloreanfml: Quickness adds to the feature, it isn't the whole of the feature.
11:39:33*bluebrother likes to strengthen that the menu is NOT a button to enter the browser
11:39:44_lifelessi just wondering, why bitmaps is ok and ui not
11:40:00LloreanJdGordon: We've offered suggestions for how to compromise and still have the new feature... You're the one insisting on "all or nothing" which is going to force people who won't accept "all" into the "nothing" position.
11:40:09JdGordon_lifeless: is the invert option enabled? that would explain it
11:40:29bluebrotherfor myself I really don't want quickly switching between useless wps screens but be able to quickly enter the browser
11:40:53JdGordonthen you wont use the tag and the discussion is irrelevant...
11:40:58Lloreanbluebrother: I can see some use in it, *especially* once we have multifont
11:41:02LloreanA large clock screen would be handy.
11:41:03bluebrotherthis main menu thing is ... well, it's used as excuse for making things slower
11:41:10fmlLlorean: correct (about what you can't see in that screen). But I still think that the main point of the modes is not only the information but also how it's accessed.
11:41:11LloreanI don't see "quick" being necessary. It'd be nice, but non-essential
11:41:26*JdGordon does see quick as being essential
11:41:30Lloreanfml: If you had to pick one to drop, the information or the quickness, which would it be?
11:41:32bluebrotherLlorean: well, it might be useful. I won't use it though. And most important, I absolutely don't want to use quick access to the browser
11:41:50_lifelessJdGordon, nope, HAVE_LCD_INVERT is not set
11:41:50Lloreanbluebrother: Well you only lose quick access if your WPS has the tag.
11:41:55LloreanSo if you won't use the tag, it really won't affect you.
11:41:57*bluebrother wants *quick* for the *browser*
11:42:21bluebrotherLlorean: maybe. But this makes the UI inconsistent.
11:42:30Lloreanbluebrother: Yes, and that's one of the important points.
11:42:32JdGordon_lifeless: are colours really inverted or you have a bad combination of back/foreground colours?
11:42:40LloreanBut I think more important is the fact that it disables one feature while it's in use.
11:42:54bluebrothereither there is a dedicated button combo that works *always* or there isn't a button for it
11:42:57LloreanSo rather than just lack of consistency, you *lose* a feature by using it.
11:43:02LloreanRather than being able to use it alongside a feature.
11:43:26fmlLlorean: information. But I use rather simple WPS anyway so probably wouldn't need the modes. So I'm discussing with my imagination.
11:43:29pixelma_lifeless: would have been nice if you could have explained this youself - that's more of a technical development question. I have no idea (guessing some bits aren't set correctly for that display yet) but I'm sure that with a bit of patience you'll find people around who know when asked accurate question
11:43:33 Quit J-23 (Remote closed the connection)
11:43:44 Join J-23 [0] (
11:43:51Lloreanfml: You'd drop the information? What would you do with just quickness?
11:44:33fmlLlorean: I mean I wouldn't want to have modes without quickness.
11:44:48Lloreanfml: So you can't see it as having *any* use without quickness?
11:45:00bluebrotherfml: but are you willing to give up quick access to the browser for that?
11:45:12_lifelessJdGordon, yea, looks like bad combination yep
11:45:16_lifelessanyway thx guys
11:45:28fmlbluebrother: no. Quick access to the browser is a must.
11:45:32LloreanJdGordon: You really haven't answered what you intend to do now. Will you wait some and discuss it on the ML?
11:46:06JdGordonprobably hit the ml
11:46:14***Saving seen data "./dancer.seen"
11:46:21fmlbluebrother: but there might be other opinions. And they should be free to decide.
11:46:28bluebrotherfml: which is exactly why I'm against this. If we can get a custom button combo for this screen switching, ok.
11:46:57JdGordonso only targets with masses of buttons should be able to use the feature?
11:47:07Lloreanfml: We're trying to negotiate a way it can still be used, and in some cases just as quick in others at least able to be used even if it does take an *entire* 1.5 seconds to switch screens, without dropping quick access to the browser.
11:47:07fmlbluebrother: see my second reply. We don't necessarily have to wait.
11:47:24LloreanJdGordon: Where does "it's less fast" mean "it cannot be used"
11:47:40JdGordonhaving the menu is only useful if you know which screen you want
11:47:49LloreanYes, which you can REMEMBER
11:48:13 Join Darksair [0] (n=user@
11:48:52LloreanWe've got a long history of not engineering things with the dumbest user we've got targeted.
11:49:10fmlHe-he, I have a proposal. What if we could give the modes names (in the WPS). Then we could offer a context menu to switch to desired mode (with meaningful names)
11:49:40Lloreanfml: I brought that up earlier.
11:49:55LloreanIt was either found unacceptable (as anything that deviates from his current patch apparently is) or ignored.
11:50:13fmlLlorean: oh, I didn't see, sorry
11:50:18JdGordon1) that will bloat it out a bit... 2) thats not pretty (mode: aa with next track, AA without next tyrack, next track (if exsists..) it limits the amount of flexibility you have
11:50:32LloreanJdGordon: how does that limit flexibility?
11:50:42bluebrotherlimits the amount of flexibility?
11:51:02LloreanGoing from "no names" to "names" surely *increases* flexibility, even if names can't be paragraphs of text.
11:51:12fmlJdGordon: how the modes are named is left to WPS authors
11:51:46JdGordonit also kills internationalisation
11:52:16LloreanJdGordon: How many WPSes have text already that's not localized?
11:52:20JdGordonalso requires more ram to implrement
11:52:41LloreanI'll take "a little RAM lost" over "an entire feature lost" any day.
11:52:47fmlJdGordon: true, that wouldn't be int..... But is that a problem? It's not hard coded, so everyone can adjust his own copy of the WPS
11:52:51JdGordonalso you then need to put a limit on the mode count
11:53:31fmlJdGordon: it's like the named FM stations we have
11:53:32 Join MegafEee [0] (n=kvirc@unaffiliated/megaf)
11:53:49JdGordonso hit the disk every time you open the mode menu option?
11:53:50 Join culture [0] (
11:53:56 Join petur [50] (n=petur@rockbox/developer/petur)
11:54:02fmlNow I'm starting to like the idea
11:54:09LloreanJdGordon: The WPS is buffered in RAM already.
11:54:10bluebrotherthe fm stations aren't read each time you open the presets menu
11:54:13LloreanJdGordon: So why would it hit the disk?
11:54:30JdGordonhow are they stored then?
11:55:03JdGordonhmm... is the wps kept in ram?
11:55:10fmlBut it's only because I wouldn't use the modes anyway. So, for me, it's just a free cake. For others who really want it, it'll still be not very usable in this form.
11:55:24 Join ender` [0] (
11:55:37LloreanJdGordon: You don't think we load all the bitmaps and text every time someone returns to it, do you?
11:56:17JdGordon.... the text from the .wps... is it kept in ram? I thought it was dumped and only the tokens were ketp
11:56:27LloreanJdGordon: You can have actual strings in a WPS.
11:56:34LloreanYou can have "Artist: %ia"
11:56:44LloreanSurely that has to be kept *somewhere*
11:56:56LloreanUnless that feature vanished while I wasn't looking.
11:57:07JdGordonyes, but are they referenced into the .wps text? or copied somewhere?
11:57:17LloreanCopied to the WPS buffer, I'm sure.
11:57:25LloreanBut couldn't your list of screens also be copied to it somewhere?
11:57:28fmlJdGordon: the tokes bear data with them
11:57:37 Quit MegafEee (Client Quit)
11:57:53 Join MegafEee [0] (n=Linux@unaffiliated/megaf)
11:58:18 Join PaulJam [0] (
11:58:21 Quit J-23 (Remote closed the connection)
11:58:41fmlSo you have a token "text", and the data "tralala" along with it
11:58:43 Join J-23 [0] (
11:59:09JdGordonnot without wasting more ram... strings are copied into a new buffer.,..
11:59:31 Part kachna|lappy ("Konversation terminated!")
12:00:47fmlJdGordon: for the names, we can use the same buffer that's used for storing string constants in the WPS
12:02:36*JdGordon is starting to see advantages of having the menu, but still doesnt like the idea of losing the button
12:02:54LloreanJdGordon: using "Rec" gets you the button on a decent percentage of targets.
12:03:21JdGordonyeah, I'm not convinced rec sholdnt be used for something closer to its name
12:03:30JdGordonlike changing to the rec screen perhaps :p
12:03:33LloreanHolding record *is* used for recording, already.
12:03:43LloreanBut short-press is unused AFAIK.
12:03:56bluebrotherit's "toggle backlight" ;-)
12:04:42bluebrotherbut isn't using a button that's currently unused in the wps too obvious as a solution? :P
12:04:49JdGordonok, bugger it... lets make a camel... menu, setting to choose to use the browse button, context menu option for "return to browser" and anything else I forgot?
12:05:00Lloreanbluebrother: It'd still be pretty close to "Toggle backligh" even with modes, since changing modes doesn't change any playback factor you can cycle around to mode 1 again
12:05:24bluebrotherLlorean: well, that "toggle backlight" isn't an intended functionality from my understanding anyway ;-)
12:05:36JdGordonits more of a bug than a feature
12:05:40bluebrotherso no loss here
12:05:42LloreanJdGordon: I'm sorry, what? That sounds like the exact opposite of what people are leaning toward... customizing a button, and slowing down getting to the browser even further?
12:06:07 Join MethoS [0] (
12:06:12Lloreanbluebrother: It's not, but it's one people think is an intentional feature these days. But it's no loss for cycling WPS modes really.
12:06:15JdGordonyou dont know the reference? camel is a horse made by a commitee... that option makes everyone happy
12:06:42LloreanJdGordon: Never heard that, no.
12:06:43bluebrotherLlorean: peple do thing that? Oh my. I still agree with you
12:07:08Lloreanbluebrother: Well even if they think that, cycling WPS modes is still "better" than pressing a functional button to bring up the backlight.
12:07:49LloreanJdGordon: I honestly just thought you'd mistyped something with 'camel' though I couldn't figure out what.
12:09:55JdGordon... next topic... the scroll padding setting.... :D
12:10:29*Llorean still thinks "setting a custom string" is the best overall way to go.
12:11:18LloreanNo choice we have is liked by all, but at least the only objection to that one (afaik) is the standard "we've got too many options already"
12:11:36bluebrotherwell, maybe we should discuss if we want "hidden" settings or not. We already have some, and padding might be a good candidate for another
12:11:48bluebrother*if* we decide to use hidden settings at all.
12:11:51JdGordonwe have a single 1 hidden setting
12:12:00LloreanI'm okay with hidden "theme" settings (settings that should always be dealt with in a theme.cfg file) only.
12:12:08LloreanJdGordon: Only one?
12:12:13bluebrotheryou sure?
12:12:20LloreanI'm pretty sure we've got at least two
12:12:27JdGordonwhich is the other?
12:12:32LloreanFiletype colors are one. The other I don't know of, but sdoyan called me on it when I said it wasn't possible.
12:12:36bluebrotherwhich one is the first?
12:12:50JdGordonhold left/right in the browser which should probably go anyway
12:13:26LloreanJdGordon: What's hidden there?
12:13:26bluebrotherwhy? This makes going to the root folder a pain
12:13:33LloreanIsn't that just "scroll lines for viewing"?
12:13:45JdGordonhonestly I dont even remember what it does
12:13:54JdGordonI remeber fighting for it to being with but never change it :p
12:13:56bluebrotherthat's "screen scrolls out of view", and that is a setting
12:14:01Lloreanholding left/right in the browser forces scrolling lines to scroll in the direction you hold
12:14:33LloreanIt scrolls all lines, rather than just the highlighted one.
12:15:24bluebrotherah, right. But only if you triggered the scrolling −− if you're "just" in the browser holding left goes to the root folder
12:15:37JdGordonhold_lr_for_scroll_in_list is the hidden setting
12:15:48LloreanThere's a setting for it?
12:15:51JdGordondefaults to true
12:16:06JdGordonthat one actually changes the keymap in the browsers iirc
12:16:23bluebrotherin any case we now have at least two hidden settings ...
12:17:00LloreanWell, I'm for colours being allowed to stay "hidden" since I think it should always be handled by themes, and I'm against this other setting staying hidden since it's not even documented as far as I know
12:17:11LloreanBeyond in the list of config settings, I'm sure
12:17:36*JdGordon doesnt mind documented hidden settings
12:17:50LloreanI'd actually say "spacer string" qualifies as a theme setting, since a valid size is *very* dependent upon font and how you display the strings (small viewport, large viewport, etc)
12:17:50JdGordonand would mind them even less if we could localise plugins
12:18:02bluebrotherwell, we could have hidden settings and configure them using rbutil
12:18:20JdGordonthe argument against hidden settings is you cant easily change them without a computer
12:18:39bluebrotherone could also argue that wps and font should get hidden as they are handled by the theme
12:18:50bluebrotheroh, and the icon set is a hidden setting as well.
12:19:16JdGordonit needs a nice choose-file/folder widget though so themes can be built on the dap
12:19:17LloreanI think most "theme" settings shouldn't be arbitrarily changeable.
12:19:34bluebrotherwell, you can change those hidden settings using cfg files
12:19:35LloreanEspecially since with list viewports and such, they're just going to get more complex and easier to mess up being able to use the device without a reset.
12:20:06bluebrotherplus, if we want hidden settings we could consider an "extended settings plugin" that allows changing them −− like gconf on gnome
12:20:13 Quit MegafEee (Remote closed the connection)
12:20:14bluebrotheror windows regedt
12:20:17*JdGordon almost suggests removing theme settings from global_settings and the usual .cfg file completly
12:20:36JdGordonbluebrother: yes, but not until plugin localoisation
12:20:40bluebrotherand where put them?
12:20:48JdGordonnot sure
12:20:56JdGordona .theme maybe
12:21:19bluebrotherJdGordon: no. I don't see it as problem to have some regedt-like plugin that is not localized
12:21:51bluebrotherwhy another file type? That way I can't easily restore my configuration anymore
12:22:00JdGordoni dunno
12:22:16JdGordona simple config editor can be done very quickly if localisation isnt important
12:22:44JdGordonbut there is still the problem with it being a nasty non-contectual list of cfg item names
12:24:18bluebrotherthat could get worked around too −− just make setting names something like general.filetype_color
12:24:41bluebrotherand then derivate the location in the plugin depending on the name.
12:24:51JdGordonthats an idea which should be discussed regardless...
12:25:05bluebrothersuch a plugin could even do that dynamically based on a configuration file.
12:25:37bluebrotherwould have the advantage that it wouldn't need adjustment when a new setting comes in (apart from i18n when that's done, of course)
12:26:01_lifelessfixed lcd, there was wrong pixel format
12:27:37*JdGordon would be very much in favour of adding groupings to the setting cfg names
12:27:56JdGordonit would at the very least mean settings_list.c would get some order
12:33:56 Join {phoenix} [0] (
12:36:24 Join Thundercloud [0] (
12:39:06*fml sees the chances of the scroll padding string patch to be committed raising (maybe after stripping off the menu code)
12:41:24fmllinuxstb: shouldn't HZ be defined somewhere else? I find inclusion of kernel.h in wps_parser a bit weird.
12:42:30linuxstbfml: I find it weird as well - but using HZ in wps_parse.c is the weird part, rather than defining HZ in kernel.h
12:44:15 Quit _lifeless (Remote closed the connection)
12:44:35JdGordonwhy is using HZ in wps_parse.c wierd?
12:45:19 Join _lifeless [0] (n=lifeless@
12:45:34fmllinuxstb: you may be right. It's used as a default for the timeout of sublines. We could use seconds as the unit and perform tranlation in ticks in the interpreting code.
12:47:39linuxstbJdGordon: Maybe "weird" is the wrong word, but it's not expected that a wps parser requires a kernel #define. I'm not saying it should be changed though.
12:48:05fmlErr... It's for volume being changed. But the same function.
12:51:27fmlIsnt' the code wrong anyway? IIUC, the value is given in tenth of a second. But the default value is HZ which is 100.
12:54:31 Join Nico_P [50] (n=nicolas@rockbox/developer/NicoP)
12:55:25fmlHe-he. When we talk about WPS Nico shows up! Hello!
13:03:29 Nick fxb__ is now known as fxb (
13:03:59 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
13:04:50 Quit {phoenix} (Read error: 110 (Connection timed out))
13:06:25*fml thinks that the code for the button_volume token is wrong. The tag should take a value expressed in 0.1 sec. But the code interprets the value as 1/100 secs. So if you specify 2.5 you should get 0.25 secs, not 2.5.
13:06:31 Join MegafEee [0] (n=Linux@unaffiliated/megaf)
13:06:34*fml goes check
13:08:51fmlYes, my assumtion is correct
13:10:08 Join {phoenix} [0] (
13:10:21linuxstbfml: What does "2.555" give you?
13:12:47 Join n1s [0] (n=nils@rockbox/developer/n1s)
13:14:20 Quit BigBambi (Read error: 104 (Connection reset by peer))
13:15:08 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
13:18:11pixelmathat hold_lr_for_scroll_in_list settings was probably "just" made a setting for the Ondio because there holding "right" envokes the context menu (on a file/folder) but later this forcing the horizontal scroll was changed to a button combo on the Ondio (basically it's useless though) and I remember this disabling through the setting didn't work reliably or some other problems with it (at least as long as it wasn't changed to a combo the context me
13:19:06pixelmathe context menu appeared sometimes and other times it didn't...)
13:19:39 Join simongmzlj [0] (
13:20:56 Quit MegafEee ("KVIrc 4.0.0 Insomnia")
13:22:55fmllinuxstb: 2.5 is very short. I have a patch and will upload it in a sec.
13:23:15 Join MegafEee [0] (n=Linux@unaffiliated/megaf)
13:24:38 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
13:25:33simongmzljwhat does rockbox require for it to recongnize album art?
13:25:46 Quit Thundercloud (Read error: 54 (Connection reset by peer))
13:25:51 Join Thundercloud [0] (
13:25:59linuxstbsimongmzlj: See
13:26:21pixelmafml: why should I get .25 seconds if I specify %t2.5 ? Or am I misunderstanding what you mean?
13:26:35simongmzljlinuxstb: perfect, thank you. feel stupid for not finding it myself :D
13:29:12linuxstbfml: Shouldn't TIMEOUT_UNIT be defined in terms of HZ? And if it's only used in gwps-common.c (i.e. isn't a public value), it can be defined there.
13:30:16linuxstbBut "vol = 10" seems to also be reliant on TIMEOUT_UNIT...
13:30:41amiconnlinuxstb: I would like to move the (few) variables from rangecoding.h to entropy.c (they are against the "no code in .h files" rule). Do you think that's ok?
13:30:47fmllinuxstb: you mean HZ/10 in the define instead of just 10? Yes, that would be better.
13:31:36Nico_Pfml: hi :)
13:31:53amiconnbytebuffer, bytebufferoffset and rc
13:32:38linuxstbamiconn: Yes, they should be moved to a .c Maybe the whole of that .h could...
13:33:04JdGordonis anyone else seeing LD rockbox(ui/.elf) before the rockbox .c files are compiled and thinks thats wierd?
13:33:59fmlJdGordon: yes. Very weird.
13:34:07amiconnIt's a side effect of the "call PRINTS" construction when used for more than one command per rule
13:35:06JdGordoncan it be "fixed"?
13:35:21amiconnlinuxstb: You mean pasting all of it into entropy.c? That's the only file using it...
13:35:22*JdGordon hasnt tried Zagor's new Makefile yet to see if its a problem there also
13:36:02*JdGordon has an incredibly crude open file/folder widget working which could come in handy....
13:36:15amiconnJdGordon: No, the only solution would be to avoid it, i.e. have only one command per rule
13:36:38amiconnOr only have one message, printed before the commands
13:36:57JdGordonoh well... not really a big deal, a bit wierd because it used to work
13:37:18amiconnThe cause is that 'make' evaluates all internal commands (and 'PRINTS' evaluates to 'info for gnu make 3.81+) before external commands
13:37:34linuxstbamiconn: Yes
13:39:46fmllinuxstb: what do you mean by vol=10? Where is it?
13:40:22linuxstbfml: Sorry, I meant "val = 10;" in wps_parser.c
13:41:15fmllinuxstb: yes, that's also a timeout. I.e. the default for sublines is 10 units = 1 sec
13:41:34n1sJdGordon: a file/dir open widget would be a major improvement in many places of rockbox. 1) remove backdrop selection from file browser context menu 2) easier setting of recordign dir 3) less typing when saving playlists :D
13:41:35fmllinuxstb: err.... for the button_volume, I mean
13:42:22fmlAnyway. I have to go now. Bye.
13:42:29 Quit fml ("CGI:IRC 0.5.9 (2006/06/06)")
13:42:39JdGordonn1s: yeah, I've been wanting to make one for AGES... first attempt is a bad hack to dirbrowse() though...
13:42:50JdGordontrying to work out keypresses also
13:43:30n1sI had an idea of basically using the filebrowser but with a different context menu, is that similar to what you are doing?
13:43:36JdGordonatm I've got action_std_context being used to say "use this filename" so directories can be selected (or ok on files) but im nto sure it works well
13:43:55JdGordonyes, bypass the usual button and context menu handling
13:44:59n1sanother place where that could be used is the playlist catalog which uses a custom list now with an arbitrary max item count.
13:46:17***Saving seen data "./dancer.seen"
13:46:29JdGordonhmm... need to somehow add a "<new file>" to the top of the list if its being used as a save-as dialog
13:48:01amiconnlinuxstb: Standalone demac didn't compile on cygwin... already fixed it locally. The linking step also needs -mno-cygwin, otherwise it throws "undefined reference to `__imp___iob'" errors
13:48:34pixelmabluebrother: around?
13:49:26linuxstbamiconn: OK, thanks.
13:51:24 Join Nibbler [0] (n=Nibbler@
14:00:27*amiconn will now work on introducing demac_config.h
14:00:56 Join gabe565 [0] (
14:08:15 Quit simongmzlj ("Leaving")
14:09:47 Quit gevaerts (Nick collision from services.)
14:09:59 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
14:10:09JdGordondoes it make sense to allow all the usual browser buttons (i.e return to menu/wps, stop, volume, etc) in a saveas/open file dialog?
14:10:33JdGordonI have a feeling that it will just confuse things if they are allowed
14:11:43*JdGordon thinks the current filters system isnt good enough for it
14:11:52 Quit {phoenix} (Remote closed the connection)
14:12:03amiconnlinuxstb: What is ROCKBOX_PLUGIN meant for (in parser.h)?
14:12:21 Join Schmogel [0] (
14:13:16linuxstbamiconn: Hmm... I think I first implemented demac as a Rockbox plugin, so I'm guessing it's left over from then.
14:13:32linuxstbi.e. it probably should never have been committed
14:13:49amiconnOk, I'll remove it
14:14:28*amiconn wants this demac_config.h in order to be able to choose between 16 and 32 bit filters
14:15:33 Join {phoenix} [0] (
14:16:48 Join simongmzlj [0] (
14:19:06 Join webguest56 [0] (n=51386aec@gateway/web/cgi-irc/
14:19:51 Join tyfoo [0] (
14:20:41 Part simongmzlj ("Leaving")
14:24:05 Quit webguest56 ("CGI:IRC (Ping timeout)")
14:30:02 Join Nibbl [0] (n=Nibbler@
14:37:42 Join webguest56 [0] (n=51386aec@gateway/web/cgi-irc/
14:38:18webguest56Hi all, sorry to ask this, but I'm quit sure it's not in the faq, I actually get an Ipod classic (maybe named ipod video on rockbox website?), and I need to make him work as a linux live cd. Can I make two partition, one bootable and the other with rockbox ?
14:38:47domonokyrockbox doesnt work with the ipod classic...
14:39:13 Quit Horscht ("I got raided by the FBI and all i got is this lousy quit message")
14:41:36 Quit Nibbler (Read error: 113 (No route to host))
14:43:40 Quit webguest56 ("CGI:IRC (Ping timeout)")
14:45:16 Join Loules [0] (
14:45:42 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
14:47:14JdGordon isnt finished and doesnt work yet... but is the new code better than the old code? the enw code would let you choose exactly which known filetypes to display in the browser instead of the known groups or "all" or only 1 type
14:47:38 Quit BigBambi (Read error: 104 (Connection reset by peer))
14:47:55 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
14:51:53amiconnlinuxstb: Do you think the vector math choice in filter.c should go into demac_config.h too? I'm undecided... (proper channel this time)
14:53:08linuxstbThis is choosing whether to use 16-bit or 32-bit math, depending on target CPU?
14:53:53amiconnIt will centralise several choices which are now scattered
14:54:33amiconnIRAM usage, inclusion of rockbox headers, ...
14:54:59amiconnThere's a nice standard macro that allows to write a header which is usable by both C and asm files
14:55:44amiconnStuff for C needs to go into an #ifndef __ASSEMBLER__ section, stuff for asm into the opposite
14:55:54 Quit Nibbl (Read error: 113 (No route to host))
14:56:18 Join tessarakt [0] (
14:56:37amiconndemac_iram.h will be removed
14:59:03 Join __lifeless [0] (n=lifeless@
14:59:32amiconnlinuxstb: Btw, is there a reason why libdemac has two HISTORY_SIZE macros which aren't related? (one for the filters and one for the predictor)
15:01:02 Join Nibbl [0] (n=Nibbler@
15:01:14 Quit fyrestorm ("( :: NoNameScript 4.22 :: )")
15:01:16linuxstbamiconn: I don't think so...
15:01:35 Join fyrestorm [0] (
15:02:56JdGordoncan anyone think of a reason why the dirfilter would change (while the tree is open) other than by the quickscreen?
15:04:28 Quit dirwiz (Remote closed the connection)
15:05:20amiconnlinuxstb: I think both history sizes could use the same value (it's just trading memory for speed), but it's your lib after all
15:06:03amiconnI want to move this into demac_config.h because it's probably not possible to use 512 when the filters are 32 bit.
15:09:55linuxstbamiconn: What are you suggesting? Combining them into one #define, or changing the names to keep them separate, but with the same value (for now) ?
15:10:20 Nick JdGordon is now known as JdGordon|zzz (n=jonno@rockbox/developer/JdGordon)
15:10:23amiconnThat's what I want to know. Either would be possible
15:11:53 Join faemir [0] (
15:12:41linuxstbI would say to keep them separate for now - i.e. change the names to make them unique, but keep them both as 512.
15:12:47 Quit lasser ("ChatZilla 0.9.83 [Iceweasel 3.0.3/2008092816]")
15:13:22 Quit _lifeless (Read error: 110 (Connection timed out))
15:13:53 Quit JdGordon|zzz (Remote closed the connection)
15:14:04domonokyJdGordon|zzz: loading a cfg could also change the dirfilter..
15:14:24 Join ap0 [0] (
15:14:43 Quit culture (Success)
15:17:24 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
15:17:35 Nick JdGordon is now known as JdGordon|zzz (n=jonno@rockbox/developer/JdGordon)
15:32:09pixelmabluebrother: just asking now... I'm currently looking into preparing an M3 manual (got the platform file already) and now have a question regarding the button tables. During DevCon we said that we probably need another column for targets with a second button mapping - so my question is how to define another column in preamble.tex (I guess). Also, I think that these 3 columns tables could then be a bit wider in the pdf (less margin). Any ideas?
15:35:05 Quit gabe565 ("ChatZilla 0.9.83 [Firefox 3.0.4/2008102920]")
15:42:43 Join MethoS- [0] (
15:46:20***Saving seen data "./dancer.seen"
15:47:23 Quit tessarakt ("Client exiting")
15:48:51 Join culture [0] (
15:50:22 Quit MethoS (Read error: 60 (Operation timed out))
15:52:07 Join tessarakt [0] (
15:56:36bluebrotherpixelma: now back ...
15:57:05bluebrothercheck the btnmap macro in preamble.tex ... the .75\textwidth defines the width of the table
15:57:50 Quit tessarakt (Read error: 104 (Connection reset by peer))
15:59:11bluebrotherthe argument after that defines the columns of the table. Replacing the X in there with .5X.5X might do it. Not really sure right now though
16:03:11bluebrotherpixelma: btw, what do you think about FS #9554? From my limited perspective the way Rockbox does it currently seems more logical, but I don't have an m3
16:04:44 Join meven [0] (
16:07:41 Join tessarakt [0] (
16:07:53 Quit Horscht ("Snak 5.3.3 Unregistered copy. Evaluation period is over. Program will now quit. Thanks for using Snak.")
16:12:25 Quit BigBambi (Read error: 104 (Connection reset by peer))
16:17:45 Join BigBambi [0] (n=Alex@rockbox/staff/BigBambi)
16:20:22 Quit Thundercloud (Remote closed the connection)
16:25:53 Part dany_21a_
16:27:39 Join shotofadds [0] (n=rob@rockbox/developer/shotofadds)
16:29:00bertrikhow do players 'normally' get first sound?
16:29:39bertrikthe ams sansa v2 i2s doesn't look too hard to interface with
16:30:09bertrikshould we add a "fifo stuffer" in some of the pcm functions?
16:35:30 Join homielowe [0] (
16:35:40 Quit MegafEee ("KVIrc 4.0.0 Insomnia")
16:35:59 Join dany_21a_ [0] (
16:36:52bertrikI suppose that usually no codec is used to provide the data for an initial port, but simply a raw wav file
16:37:16 Join herrwaldo [0] (
16:37:51*domonoky is experimenting with the sd-driver for sansav2...
16:40:24bertrikdomonoky, funman posted a patch last week that makes it boot correctly to the menu on my clip almost every time
16:40:45domonokybertrik: yes, if have this patche applied.. plus more changes...
16:41:23domonokythere is still corruption... but now i have more than 20 bootups without bad - checksum... and a idea what might be wrong...
16:42:24domonokythe strange thing is, that the sd driver tells us RX_FIFO overruns, seemingly without a RX_FIFO_HALFFUL flag...
16:43:48domonokyfunman tried to fix this with retrys... but that is still buggy...
16:44:32 Join MegafEee [0] (n=Linux@unaffiliated/megaf)
16:45:31*shotofadds hits commit on some nice NAND driver improvements for D2 / iAudio 7 users..
16:46:08*gevaerts updates
16:49:37*shotofadds has some more disassembling to do before the DAX will work reliably, i'm afraid..
16:50:11gevaertsI can always use it on my D2 :)
16:52:28 Nick obo_ is now known as obo (n=obo@rockbox/developer/obo)
16:53:19shotofaddshopefully those Bad Checksum errors are a things of the past ;-)
16:54:27gevaertsNice delta :)
16:55:14 Quit Loules ()
16:55:45shotofaddsgevaerts: the old static LPT buffer was only being fully used on the 16Gb model, so the static buffer was rather wasted on everything else (including the M200 with its piddly 1Mb RAM..)
16:56:06 Quit Seed (Read error: 60 (Operation timed out))
16:56:27obosoap: I think so. Do you know if it happens if you shutdown before or after the half-way point of the track? Is your start screen set to resume playback?
16:56:33shotofaddsRAM usage on the 16Gb model will be the same, but everything else should be anywhere up to 24KB smaller
16:56:40gevaertsSo now it just allocates once on boot?
16:56:49pixelmabluebrother: had a quick look at the preamble.tex but I am not sure if I'll understand that.. need to look more closely. I also try to think about how to handle the 2 different tables - am I right in thinking that we would need 2 different buttonmap table macros and then use one or the other in the single tex files (maybe opting for "remote")?
16:57:16soapobo, I'll create a reproduction recipe for you. I had assumed there was no way around it, so had not investigated deeply. Give me an hour or so?
16:57:37soapand, no, my start screen is _not_ set to resume playback.
16:57:55 Join Seed [0] (
16:58:11shotofaddsgevaerts: I'm not sure why doom doesn't start yet, though I think it needs the user timer which I haven't implemented yet. That or filesystem write access..
16:58:20obosoap: sure, sounds good.
16:58:27pixelmabluebrother: about FS #9554 - I don't have an M3 either and no remote for my M5 but amiconn chose this way on purpose, guess he should comment...
17:02:24pixelmabluebrother: he says that the way the Iaudio OF handled this felt wrong and he always moved in the wrong direction ;)
17:05:01amiconnlinuxstb: Does the standalone 'demac' program actually work for you?
17:05:35amiconnFor me it compiles, but creates strange-sounding .wav files
17:09:06 Nick fxb is now known as fxb__ (
17:12:31bluebrotherpixelma: yes, I guess we would need two buttonmap macros
17:18:07pixelmaso something with a different name?
17:19:09pixelmaor could the opting for remote already happen in preamble.tex?
17:21:59bluebrotherhmm. We could make all targets with remote use a different macro with the same name. But are there any maps that without remote maps even for targets with remote?
17:24:25 Quit shotofadds (Read error: 104 (Connection reset by peer))
17:25:05pixelmamaybe I'm slow today but could you rephrase the second part?
17:25:49 Quit Seed (Nick collision from services.)
17:25:52 Join Seedy [0] (
17:27:03*gevaerts doesn't think that slowness is required to not understand that sentence :)
17:27:20 Join webguest03 [0] (n=54e21762@gateway/web/cgi-irc/
17:27:27 Quit webguest03 (Client Quit)
17:27:42 Join webguest14 [0] (n=54e21762@gateway/web/cgi-irc/
17:27:45 Quit webguest14 (Client Quit)
17:32:00 Join funman [0] (
17:33:12*gevaerts hands funman a ☕
17:34:32funmanit's very welcome because ☔
17:35:45*domonoky seems to have now a very stable sd-driver for sansa v2 (funamns changes + no CCU_IO + busy loop in sd_poll_status() (no yield))...
17:36:13 Join saratoga [0] (n=41becb3b@gateway/web/cgi-irc/
17:36:42funmandomonoky: nice, does very stable means never fail ? (what about the delays also ?)
17:36:59domonokyfunman: as far as i can see, there could also be a problem with next_yield in the sd-driver, its global and uses in the thread and in the other functions...
17:37:30domonokyfunman: very stable, means over 20 boots without a error.. but it still does the retry sometimes...
17:37:35funmancould very well be since i copied this code without trying to understand it
17:37:36 Quit {phoenix} (Read error: 104 (Connection reset by peer))
17:37:48funmandomonoky: do you check for RX fifo overrun and DATA crc fail status bits ?
17:37:59 Quit stoffel_ ("leaving")
17:38:13bluebrotherpixelma: well, I'm wondering if a remote-enabled target also needs button maps without remote buttons, in which case we would need both non-remote and remote maps
17:38:24*bluebrother is kinda confused today
17:38:44domonokyfunman: i uses your patch, which checks for overruns, and makes a retry. i think DATA crc isnt checked...
17:39:38amiconnIs that SD?
17:40:32domonokyamiconn: yes, internal memory which is accessd with a via a sd controller (pl180) ..
17:40:42funmandomonoky: could you post your current diff somewhere please ? (I have a lot of attempted modifications but can't remember the results for each, except that no was perfect)
17:41:24amiconnOn Onio (MMC) we neither check command crc nor data crc. I'm not sure whether the PP SD controller does that automatically
17:41:26domonokyfunman: will do, although my current code is also full of other changes.. (mostly panics to check for errors)...
17:41:45funmandomonoky: cool, i like the use of panic to debug :)
17:42:27funmanamiconn: some commands issued at least in the power-up phase of the SD card fail the (command response) CRC check, so we ignore it.
17:42:45 Join Seed [0] (
17:42:49pixelmabluebrother: I've been told that some screens on e.g. the Archos Recorders (which have a non-lcd remote) don't have a remote mapping yet (some could probably get fixed). Would it be too weird to have a " - " or "n/a" in the remote column there?
17:42:55 Join linuxstb_ [0] (n=linuxstb@rockbox/developer/linuxstb)
17:43:14 Quit linuxstb (Read error: 104 (Connection reset by peer))
17:43:16funmanit was said that the CRC bits of the response which contains the CSD register were equal (and wrong) on 2 models which answered a different CSD
17:43:32domonokyfunman: my current modifications: (remember to replace both bootloader and rockbox to test it).
17:44:12funmanI have never had any problem with the bootloader so far
17:45:09domonokybut you inherit the state which the bootloader has set. (ie the CCU_IO thing) ..
17:45:17funmanoh right
17:46:21***Saving seen data "./dancer.seen"
17:46:34 Nick linuxstb_ is now known as linuxstb (n=linuxstb@rockbox/developer/linuxstb)
17:47:00funmanmodifying the CCU_IO register might be needed with the 'original' pl180 controller (a.k.a for SD slot)
17:47:47 Join kugel [0] (n=chatzill@unaffiliated/kugel)
17:47:55 Quit Seedy (Read error: 110 (Connection timed out))
17:48:10domonokyyes, it might be needed for v2s with external sd, but on the m200v4 it conflicts with the backlight.. (backlight goes out, when you switch to SD-mode) :-)
17:49:09kugelI wonder how much code is copied from ata-sd-pp
17:49:41funmanI would say 90%
17:49:45kugelhas anyone tried to use that file instead (including #include the proper ata-target.h etc)
17:50:05funmanhu? what do you mean exactly ?
17:50:27kugeli mean to use ata-sd-pp for our ams sansas
17:50:44 Join jeffdameth [0] (
17:51:00funmanThat's what I did - except for using the PL180 registers instead of whatever controller is used on PP
17:51:45kugelso why do we have a a extra file? also I see some code in ata-sd-pp not copied to ata_sd_as3525 (btw: why underscores in that filename?)
17:52:30funman1/ because Someone didn't factorize the common code
17:52:42funman2/ because that code was not needed for the driver to run
17:52:53funman3/ why not underscores?
17:52:59bluebrotherpixelma: not sure.
17:53:07kugelunderscores are totally uncommon in rockbox
17:53:17kugelfor filenames
17:53:29soapfirst attempt at reproduction failed, obo. I shutdown @ the 30% mark in a 23 minute track, restarted and let the track complete and only the second entry counted as listened. This "double listened" issue happens to me all the time, though I'm not quite sure why. Will try another recipe.
17:53:41funmankugel: 1546 files in rockbox use them
17:54:21*bluebrother wonders if kugel is serious on that underscore claim
17:54:43kugelsorry, in the target/ dir it's uncommon
17:55:04funman10% of the files in target/ use underscores
17:55:13gevaertsRockbox has 372 c files and 338 h files with underscores
17:55:21bluebrother61 files below firmware/target is uncommon?
17:55:37funmandomonoky: indeed, very stable ..
17:55:56obosoap: currently the code should mark a track as skipped if shutdown occurs before 50%, and skip it entirely if it is resumed >50%. If the track is started afresh on the next boot that could explain it...
17:55:59kugelfunman: and the other 90% dont
17:56:08soaplinuxstb, yesterday you mentioned a test compile with the jpeg decoder in core - that does not appear to be a current flyspray entry - is this just your personal hack to check deltas?
17:56:12funmanwow, a panic in fat driver : "Updating size on empty dir entry %d"
17:56:13kugelanyway, - or _ doesn't really matter
17:56:26domonokyfunman: nice to hear... now to find out, what makes it stable :-)
17:56:51saratogasoap: is there an FS for a jpeg decoder in core?
17:57:04linuxstbsoap: It was just a test - I've since deleted the small changes I had to make to get it to compile (it didn't actually do anything, it was just compiled into the core).
17:57:06soapobo, in my normal usage pattern I would never start afresh upon reboot - that is what is confusing me.
17:57:33soaplinuxstb, that's cool - I was just curious if you had done that or if you had done more reworking.
17:58:10oboweird - I can only guess that there may be some spurious track changed events coming from the playback engine...
17:58:12funmandomonoky: we should not return 0 in sd_poll_status if the bit we were looking for was set, but instead clear this bit and see if other ones are set
17:58:34soapI'll find a recipe - I promise.
17:58:44amiconnlinuxstb: ping
17:59:18domonokyyes, and it could get confused if we get a timeout and MCI_STATUS is complety 0 :-)
17:59:42 Quit MegafEee ("KVIrc 4.0.0 Insomnia")
17:59:43funmanit shouldn't, I think 'RX active' should be set at least
17:59:45linuxstbamiconn: I haven't tried demac for a long time - but it used to work...
18:00:27 Join XavierGr [0] (n=xavier@rockbox/staff/XavierGr)
18:01:52linuxstbamiconn: I've just tested demac and it gives me a .wav with identical md5sum to the original file (-c3000 created with mac 3.99)
18:02:04domonokyfunman: if a init or something failed rc activ might not be set. It also isnt set when we get the Overrun..
18:02:10amiconnlinuxstb: On linux?
18:02:25linuxstbYes. 32-bit
18:02:30funmandomonoky: we could use -1 as an error, the 32 bits are not all used
18:03:06domonokyfunman: yes, or some bit in the unused area..
18:03:07amiconnHmm, I'll try linux next (in a VM). Cygwin produces strange-sounding .wav files
18:03:29linuxstbamiconn: demac should also do crc checks on the decoded audio data, so should report decoding errors.
18:03:43amiconnIt reported no error...
18:10:46amiconnlinuxstb: Works in linux - very weird...
18:11:27funmandomonoky: did you have a look at the DMA controller ?
18:11:49domonokyno, not till now..
18:15:26funmanI still have errors and very long delays
18:15:30 Join kugel_ [0] (
18:15:59linuxstbamiconn: Maybe a binary/ascii issue?
18:18:04domonokyfunman: yeah, it still seems to get oberruns sometime and has to retry..but its much better now :-)
18:18:14amiconnlinuxstb: Indeed....
18:18:54amiconnlinuxstb: Does open() open in text mode by default??
18:19:02amiconn(or creat() in this case)
18:19:46funmanI'd suspect it is system dependant
18:19:52linuxstbamiconn: There's no such thing in Linux, and I've no idea what Windows does. But I'm guessing it might do.
18:20:58*domonoky mentions that RTC works nicely on his m200v4..
18:21:04soapgot it obo. 10 minute track, bookmarking on. Paused the tract at the 5:30 mark. Let the device idle shutdown. Restarted, hit "Play" from main menu to resume playback, final 4:30 of track plays. Both entries marked as "L"
18:21:47oboooo, okay. I need to read up on bookmarking then :)
18:21:48funmandomonoky: does it keep a consistent setting across reboots ?
18:22:12soapThat is very likely the situation which does it to me normally, as I frequently listen to long tracks, and frequently pause in the middle when someone interrupts me.
18:22:14domonokyfunman: yes, even across AAA battery replacements :-)
18:22:23soapI'll attempt the exact same recipe w/o bookmarking.
18:22:23funmanhehe :)
18:23:28 Quit kugel_ ("ChatZilla 0.9.83 [Firefox 3.0.4/2008102920]")
18:24:41soapDifferent subject - what codepage are Linux FAT32 filenames in by default? Filenames with "extended characters" created by Windows display fine in Rockbox when using the Latin codepage option. Filenames with "extended characters" created by Ubuntu do not display correctly in Rockbox be the codepage Latin or UTF-8. I'm using unifont.
18:25:57amiconnIt depends on how you mount it
18:26:08amiconnOn disk, VFAT is *always* UCS-2
18:26:11gevaertssoap: mount(8) tells me that they are in unicode on disk
18:26:48linuxstbsoap: Have you tried looking at your disk in a Windows machine? Sounds like Ubuntu is doing something wrong.
18:27:06gevaertsWhat your linux system translates them to depends on the iocharset mount option, which IIRC defaults to whatever LC_CTYPE is set to
18:27:20soapI guess I can get a windows machine.
18:27:30*gevaerts suspecs a mismatch between ubuntu systemwide options and specific user options
18:27:34funmanmount it with -o utf8 on Linux
18:28:24amiconnlinuxstb: Thanks for the hint - problem fixed :)
18:29:15soapcodepage is currently cp437
18:29:26soapiocharset is iso8859-1
18:29:46 Join kugel_ [0] (
18:29:53gevaertsWhat does locale say about LC_CTYPE?
18:30:29kugel_jhMikeS: do you like the fading?
18:30:30soapnot sure how to check that.
18:30:31 Quit faemir (Remote closed the connection)
18:30:43funmansoap: 'locale'
18:30:45gevaertssoap: just run locale without arguments
18:30:50soapahh, "en_US.UTF-8"
18:30:59soap(I was getting there)
18:31:10gevaertsIn that case you probably need iocharset set to utf8 as well
18:31:14funmanI would change the mount options for vfat filesystems to include 'utf8'
18:31:31soapso mount it as UTF-8 instead of iso8859-1?
18:32:09 Join faemir [0] (
18:32:16soapcan do. Thank you. Repair = copy files off and back on? Or will I need to abandon what is currently on there?
18:32:17funmanso mount it as UTF-8 instead of iso8859-1.
18:32:23gevaertsWhat you're doing now is telling things like 'cp' to use utf8, which then gets interpreted as iso8859-1 by the vfat driver and converted to UCS-2
18:33:06*gevaerts thinks that this is actually OT
18:33:06funmandomonoky: do you sometimes reach the panic in drivers/fat.c line 1530 ?
18:33:24 Quit Seed (No route to host)
18:33:31 Quit Darksair (Connection reset by peer)
18:33:36 Quit kugel_ (Client Quit)
18:33:49 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
18:34:10domonokyfunman: yes..
18:34:15 Quit kugel (Read error: 110 (Connection timed out))
18:34:54domonokybut that might be because we dont have write access, but i am not sure...
18:35:24funmanI would expect write access to not be mandatory, for example if a hardware read-only switch is set (on other targets perhaps)
18:36:23funmanweird since I would expect as well our SD driver to panic if a data transfer failed
18:36:36 Quit Nibbl (Read error: 113 (No route to host))
18:38:05funmanI just formated my Clip to rule out a bug caused by linux' mkfs.vfat , and I reach the same pani
18:38:27 Quit saratoga ("CGI:IRC (EOF)")
18:39:25domonokyif this is caused by wrong data it might not panic...
18:39:51funmanonly if the datacrcfail bit wasn't set by a wrong data transfer
18:39:51domonokybefore some fixes, i often got "checksum wrong" in the bootloader, and it didnt panic..
18:40:00 Join Nibbler [0] (n=Nibbler@
18:40:19funmanin my diff, I check for fifo_full and fifo_half_full flags before reading
18:40:44funmanI retry on no flags set / or rx_overrun (after clearing rx_active, rx_data_avail, data_block_end and data_end)
18:41:45 Join Bensawsome [0] (n=Bensawso@unaffiliated/bensawsome)
18:42:07linuxstbamiconn: O_BINARY doesn't exist on Linux - see the way it's defined at the top of demac.c
18:42:49funmanposix makes no mention of O_BINARY either
18:43:09 Quit Nibbler (Remote closed the connection)
18:45:02domonokyfunman: directly before this panic, there is fat_readwrite(), so it might need the write access...
18:45:56 Join Nibbler [0] (n=Nibbler@
18:46:33*domonoky corrects him self, this readwrite call takes the read path..
18:46:36funmanI can't believe write access is needed - more that this routine is used in both read & write modes
18:48:22soapobo - same thing happens with bookmarking off.
18:50:18domonokyfunman: another interresting obervation: when i got the bad checksums in the bootloader, i let it count the retrys. it only got past the checksum check, if the retry_count was 0 ...
18:51:19 Join BHSPitLappy [0] (n=BHSPitLa@unaffiliated/bhspitmonkey)
18:52:46domonokylooked like we get wrong data when the retry happens. But i could not prove it...
18:52:54funmanI just reached the data_timeout bit set
18:53:19funmanquite chaotic ..
18:56:00 Join toffe82 [0] (
18:57:04linuxstbUnhelpful: The Gigabeat S uses TFAT (bringing you here from #rockbox-community) - do you have any idea what happens if a TFAT partition is treated at FAT?
18:57:44amiconnThat's documented somewhere on msdn
18:58:09Unhelpfullinuxstb: i believe MS actually can tell you a bit about that. i think their docs state that files and directories you create during FAT mount may be invisible under TFAT.
18:58:26amiconnIirc it's not dangerous, just directories created in FAT mode don't have the transaction protection even on a TFAT capable system
18:58:41amiconnThey aren't invisible
18:59:14 Quit gromit`` (Read error: 110 (Connection timed out))
18:59:55Unhelpfulamiconn: rereading now. you're right, they're just not safe. there's also this: "In addition, TFAT directories cannot be deleted using a Windows desktop OS."
19:00:32Unhelpfulthey certainly don't *seem* to, but i wonder if any of these issues have to do with the "file not found" boot problem on beast?
19:00:47*domonoky still wonders why we get overruns in rx fifo for v2s even when we busy loop and dont yield.. *scratch head*
19:03:07funmantoo slow
19:03:24funmandon't forget there is still the kernel timer interrupt running
19:04:14domonokyah, yes. interrupts could delay us...
19:05:28 Quit homielowe ()
19:05:31domonokyare there more/longer interrupts expected when we run rockbox, vs the bootloader ?
19:05:59funmanthe second timer maybe .. but afaiu it's only used by the plugins
19:06:58 Join kugel [0] (n=chatzill@unaffiliated/kugel)
19:07:02funman_looks_ much faster when disabling interrupts, but obviously the timeouts won't work anymore; and I still have this panic in fat driver
19:08:28funmanby the way the tick timer is a bit too fast, we should do a precise mesurement (one day)
19:08:45domonokydo you still get retrys/overruns with the interrupts disabled ?
19:08:55funmanI don't know :)
19:09:16domonokyput a panicf on the retry case, and you will know :-)
19:10:34funmansomething else to do: disable the 'fast' hardware power off, so we can use normally the power button (it's a bit in i2c register 0x21 iirc)
19:12:04funmandomonoky: no, it seems transfer is flawless with irq disableds
19:12:23soapOk, I know this has been answered 100 times, but why has USB not been committed for the non-sansa PP targets?
19:12:35funmanI'm not sure how thread-friendly this is however
19:12:42soapSo please forgive my stupidity.
19:14:34 Join miepchen^schlaf [0] (
19:16:08domonokysoap: does it work reliably on non-sansa-pp ? we probably need some heavy testing before enabling it for ipods..
19:16:14amiconnsoap: It is broken
19:16:29 Quit miepchen^schlaf (Client Quit)
19:16:43amiconnIt causes dozens to hundreds of bus resets depending on the port you use
19:17:09 Quit kugel (Read error: 131 (Connection reset by peer))
19:18:20soapthat's right - the bus reset issue on some ports.
19:18:51soapThough I was under the impression Some USB ports = none, most = few, and some = many.
19:19:17 Quit dany_21a_ (Remote closed the connection)
19:19:31soapAnd, last stupid question I promise, is the bus reset issue a data integrity risk?
19:19:39 Join dany_21a_ [0] (
19:19:52amiconnI have no single port where I get *no* resets. Some cause only a few, but some cause so many that windows gives up and states that data was lost during write
19:20:26amiconnIt also depends on the actual target (maybe PP chip, or maybe cable quality, or a combination)
19:20:59 Nick fxb__ is now known as fxb (
19:25:28 Join miepchen^schlaf [0] (
19:26:13 Quit miepchen^schlaf_ (Read error: 110 (Connection timed out))
19:28:52funmanThis panic in fat driver was added by Zagor 6 years ago (r2853), I want to ask him what it means exactly. It's not very explicit to someone who doesn't know (this) filesystem(s).
19:29:25 Quit bmbl ("Woah!")
19:30:32amiconnIf your file system is ok, this panic only comes up if some data processed by the fat driver is incorrect, i.e. the storage driver is buggy
19:31:20funmanThat, I had understood. I'd like details on what is incorrect in the filesystem
19:38:00amiconnIt means that some call to the fat driver tries to update a dir entry which is empty.
19:43:33Unhelpfulok, 8 covers scaled with the area scaler i coded last night. #5 and #8 show pretty bad aliasing - the half-tone dots in #5 have regular lines across them, probably at area boundaries, and the lines on the dog's hat in #8 are completely lost. natural images without high-contrast edges, or images with primarily vertical or horizontal lines, do a little better - #4 and #6 are good examples.
19:43:59 Join toffe82_ [0] (
19:44:10Unhelpfullink here, will cause a warning if you don't have startcom's CA cert installed:
19:46:22***Saving seen data "./dancer.seen"
19:46:48 Join kugel [0] (n=chatzill@unaffiliated/kugel)
19:47:50 Quit BHSPitLappy (Remote closed the connection)
19:52:40 Join {phoenix} [0] (
19:55:09 Join saratoga [0] (n=9803c6dd@gateway/web/cgi-irc/
19:55:26saratogaUnhelpful: is that using the same algorithm as the rockbox patch?
19:55:53Unhelpfulsaratoga: i've not examined the rockbox patch, but i believe it was stated last night to be an area average scaler
19:57:02saratogaUnhelpful: it does, but i'm sure theres a few ways to implement one
19:57:14Unhelpfulthis is my attempt at an area scaler with low memory footprint.
19:57:51saratogaah ok
19:58:00Unhelpfulit processes one input line at a time, using two buffers of 32-bit ints, which can be made the length of the input line, and one buffer of 8-bit ints, the size of the output line.
19:58:04 Quit ap0 ("Ba")
19:58:29saratogaperhaps you can clarify this for me:
19:58:33Unhelpfularea edges at 1/16-pixel-accurate (in the input image)
19:58:45saratogaif you do area sampling with only two lines, how is that different then raster sampling or whatever theyc all it
20:01:10Unhelpfulsaratoga: that's why two of those lines are 32-bit, not input-sized. they accumulate values from more than one row of input - one line accumulates scaled pixel values, the other accumulates the total of the scale values. when an entire line of output has been brought in, it scales it back down to 8 bits, and outputs it
20:01:21Unhelpfuland, sorry, actually those buffers are all input-width size
20:01:45Unhelpfulalso, i think a see a bug with respect to area edges - perhaps that will clean up some of the aliasing.
20:02:35 Quit toffe82 (Read error: 110 (Connection timed out))
20:02:57saratogaUnhelpful: how many rows of input are used to calculate each pixel of output?
20:03:42Unhelpfulsaratoga: however many would cover that output pixel, if you layed out an output-sized grid over the input image.
20:06:34Unhelpfulit *should* be able to handle scale factors down to 1/256 each direction without overflows. smaller would need you to scale more than once, but wrapping it in a function that does that might be better code-size-wise than trying to fix it in this one
20:06:48saratogaso in the example above where its 455 -> 240 that would be two pixels?
20:07:46saratogaerr two rows
20:07:58Unhelpfulapproximately 2+14/16 pixels, since it uses partial pixels when the area edges don't align perfectly
20:10:48saratogaah that makes sense
20:12:53Unhelpfulbut if you look closely, there's a bit of an issue, in that with the partial pixels, i apply them correctly in one output row/column, but in the next one, i apply the scale for the "other portion" to the *next* input pixel
20:12:59Unhelpfulthat's the bug i'm fixing now :)
20:16:19 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
20:17:22 Quit bmbl (Client Quit)
20:19:44 Quit miepchen^schlaf ()
20:25:05Unhelpfulfixed, and replaced the old images with the corrected ones. aliasing is not really improved, the bug was not the cause of the aliasing, anyway.
20:26:20Unhelpfullanczos would probably do much to improve the aliasing. but i need to read up to be able to write one, and it will need more context.
20:27:18 Join esthar [0] (
20:28:03gevaertssoap: if the USB resets are severe enough, the OS can decide to stop trying. In that case you have the equivalent of unpluging without warning
20:30:27 Quit reacocard (".")
20:31:06bluebrotherhmm. Is Track Artist used in place of the Album Artist if the latter is empty (in the database)?
20:32:45kugelbluebrother: I think so yes
20:32:51pixelmathought so (but don't know)
20:33:54bluebrotherso then FS #7129 can be considered done?
20:35:09kugelsomeone should double check before closing
20:35:36*bluebrother considers closing an old feature request for yet another strange playlist format
20:35:40kugelI'm fairly sure it already works like that. Yet I don't know why he reports that
20:36:40bluebrotherbecause the entry is almost a year old?
20:36:58bluebrotherI tend to assume the behaviour has been changed since ...
20:37:24kugelah yea true 1,5 years already
20:38:13 Part pixelma
20:38:55 Join pixelma2 [0] (n=marianne@rockbox/staff/pixelma)
20:39:11*gevaerts wonders if he should now also build windows sims, or if his torturing of FS #9534 has gone far enough
20:48:54 Join AndyI [0] (i=AndyI@
20:54:31 Nick __lifeless is now known as _lifeless (n=lifeless@
20:54:42kugelbluebrother: I'm just reading the rbutil wiki page. Can the 3rd issue not be worked around by installing the build before the bootloader?
20:55:41bluebrotherwell, that would still break in automated install when other stuff like fonts is about to get installed
20:55:42funmanforum admins: is it possible to allow attachment of .diff/.patch files ?
20:55:48 Quit toffe82_ (Read error: 110 (Connection timed out))
20:55:58amiconnlinuxstb: Interesting test results: (1) Using 32 bit ints for the filter works. (2) 32 bit filter in C is only a little bit slower on ARM7 than 16 bit filter in asm
20:56:10kugelbluebrother: then install fonts etc before the bootloader to
20:56:14bluebrothera better solution is to wait until diskutil remounts the player
20:56:53bluebrotherwell, it's only a workaround. It could also happen with manual install if the user is faster than diskutil
20:57:11kugelwhy is it unproper to install the bootloader last?
20:57:14bluebrotherso IMO the only real solution is to wait until the player gets remounted
20:58:00bluebrotherit is not improper to install the bootloader last. It doesn't fix the issue that installing the bootloader unmounts the player, thus leaving in a somewhat unknown state
20:58:09linuxstbfunman: The general view is that patches should go on the tracker, not in the forums, which is why those extensions aren't allowed.
20:58:34 Join toffe82_ [0] (
20:58:41 Nick toffe82_ is now known as toffe82 (
20:58:55bluebrotherso imagine a user installing the bootloader last and then trying to copy stuff. He won't find his Ipod anymore ...
20:59:54 Quit dany_21a_ (Remote closed the connection)
21:00:08kugelI think the typical user would reboot the ipod to see if it worked before copying any stuff
21:00:17 Join dany_21a_ [0] (
21:01:10bluebrothermaybe. Changing the order still doesn't fix the real issue but works around it
21:01:10 Quit AndyIL (Read error: 110 (Connection timed out))
21:01:12kugelor, you could tell them to (wait for the) remount
21:01:32gevaertsbluebrother: In linux you should be able to find the mountpoint from the USB id
21:01:34linuxstbamiconn: Interesting... I assume 32-bit asm filter functions are next?
21:02:19bluebrothergevaerts: linux is not an issue as that doesn't require the unmount
21:02:26amiconn92.6% realtime (up from 88.4), and that's only the first version
21:02:32amiconn(for -c2000)
21:02:55gevaertsbluebrother: I'm looking at the reds in
21:03:00funmanI have found a bug which makes plugin not link on swcodec targets with 2MB of RAM or less
21:03:08 Join robin0800 [0] (
21:04:16funmanin codecs/SOURCES libmusepack is built conditionally on MEMORYSIZE > 2
21:04:34kugelbluebrother: adding more complexity and delay doesn't sound like the proper solution too
21:04:41funmanin (plugins|codecs)/Makefile plugins are linked against libmusepack unconditionally
21:05:04funmanSince we can't do variable comparison in Makefile, should we create a LOWMEM define ?
21:05:14linuxstbfunman: Zagor is rewriting all the makefiles, so it's probably not worthwhile spending much time fixing the current system.
21:05:45funmanOk, I'll make him aware
21:05:52linuxstbIt's here -
21:06:52bluebrotherthere isn't more complexity by waiting for diskutil
21:06:59kugelregarding that wps tag discussion from earlier today: I agree with Llorean and the others removing the possibility to go quickly to the filebroswer is bad
21:07:54kugelI'd like to see a button combo, which doesn't only allow you to go forwards through the view modes but also backwards. I could imagine something like play+left/right
21:08:00bluebrothergevaerts: ah. Thought you were talking about the other issue ;-)
21:09:08Unhelpfuli wouldn't mind seeing left/right as mode switch on targets that *also* have next/prev buttons... but maybe that's only because that's exactly what beast OF does.
21:10:28Lloreankugel: On many players holding "Play" is "Stop" so that wouldn't work well at all.
21:11:11kugelLlorean: wasnt't that (hold play to stop) removed from the e200 keymap because it's inconsistent to other targets?
21:11:15Unhelpfuli see definite issues if a button that has a meaning when held is to be used as a modifier. :/
21:11:51bluebrotherkugel: all ipods have hold play to stop as there is no other button
21:12:21kugelanyway, it doesn't need to be play necessarily. Any free modifier will work
21:12:55bluebrothername a free modifier for all supported players ;-)
21:13:43Unhelpfulit would pretty much have to vary player-to-player
21:13:50kugelwhy does it need to be free for all? sometimes consistency can't be achieved. Refusing that just because a modifier isn't available on all targets is stupid
21:14:50bluebrotherwell, you need a button or a combo.
21:15:00bluebrotherso you need one ...
21:15:12Lloreankugel: e200, sure. But it's still used on the iPods where there are so few buttons it can't be put elsewhere.
21:15:22Lloreankugel: Basically, the iPods don't have a free modifier.
21:15:51LloreanI think there's going to, generally, be few enough that a "cycling" button can be used, rather than trying to make explicit controls
21:16:06LloreanButton combos are a *very* unneeded complexity here.
21:16:58UnhelpfulLlorean: perhaps use one of the duplicate buttons on players that have enough, which would satisfy those who want it in one click, and put it in a context menu on all other players?
21:17:12kugeluh? I think switching the modes through a menu brings even more complexity
21:18:03Unhelpfulkugel: testing for a button combination, in a way that is consistent and doesn't behave unexpectedly in terms of "natural" user expectations, is pretty difficult.
21:18:16kugelI don't think targets which are short on buttons should cause decreased flexibility on targets with enough buttons
21:18:45Lloreankugel: My current suggestion is "players with a spare button, use it to cycle" rather than "use it in a combo"
21:18:58LloreanPressing the button is "next screen" and when you get to the last, "next" is "first"
21:19:05*gevaerts agrees with Llorean
21:19:10bluebrotherwell, the arguing was that it should be quick. A button combo isn't quick ...
21:19:11LloreanThat's "short-press record" on a lot of players
21:19:15 Quit Schmogel (Read error: 110 (Connection timed out))
21:19:16LloreanIt may be a different button on Gigabeat S
21:19:23LloreanAnd on players without a spare button, you use the context menu
21:19:48*Unhelpful agrees with Llorean as well, if we can allow cycling either ways on players with two spare buttons
21:19:56 Join Seed [0] (
21:20:01kugelbluebrother: a combo isn't quick? I regulary use the combo to get into the pitchscreen. I find it VERY quick, especially compared to going the way through the context menu
21:20:09LloreanUnhelpful: I don't think we need to waste that many buttons.
21:20:30gevaertsUnhelpful: how many modes do you expect>
21:20:51*bluebrother agrees with Llorean on this
21:20:55Lloreankugel: It isn't quick compared to a single button press.
21:21:19kugelquick enough to not call it "not quick"
21:21:24*gevaerts expects no more than three or four different modes, in which case one button is plenty
21:21:29linuxstbWhat will the context menu contain if there is only one mode in the WPS?
21:21:40UnhelpfulLlorean: it's not exactly a "waste" on beast, and i'd be worried about what the opposite direction does if it's assigned to a directional
21:22:18LloreanUnhelpful: Don't assign it to a directional
21:22:20kugelusing rec would interfere with my build using rec to quickly access the playlist though :/
21:22:27LloreanThe Beast is likely to have recording, assign it to its "Recording" button
21:22:37Lloreankugel: Unsupported builds are not our problem.
21:22:42bluebrotherkugel: do we care about your build? ;-)
21:22:45kugeli know i know :)
21:23:01gevaertskugel: we move that to the context menu :)
21:23:15bluebrothergevaerts: we?
21:23:22Lloreanlinuxstb: I'm suggesting it as a conditional option in the context menu. If your theme contains modes, there should be a "Mode List" option in the context menu that then shows all the mode labels (if we use strings) or a list of integers if we don't.
21:23:41gevaertsbluebrother: I mean Mr Someone, in this case represented by kugel
21:24:03kugelmaybe I should just post that patch to the tracker, which utilizes the rec button for playlisting and get it committed before %mo :P
21:24:13UnhelpfulLlorean: fair enough, the <- button doesn't seem to do anything in wps now
21:24:34gevaertsI'd go for two entries in the context menu : "next mode" at the top, and Mode List under that
21:24:36linuxstbLlorean: Why not just a single option that when selected does the same as the button shortcut
21:24:45Lloreanlinuxstb: I suggested that first.
21:25:00gevaertsThat way mode switching doesn't require thought, and it's only two button presses
21:25:14LloreanIt *seems* to be more widely preferred that you can jump to an arbitrary one (if we can "name" them within the WPS) since that can save a little time vs switch, check, context, switch, check, context.
21:25:24*bluebrother bets rec-to-playlist will loose against %mo
21:25:24LloreanI think if we can't name them, a "Next mode" option would in the end be preferred.
21:25:41*gevaerts agrees
21:25:48LloreanThe "Next mode" option that does what the button does is *my* personal preference too.
21:26:02kugelbluebrother: also if it adds the possibility to quickly insert files into the current playlist in the filebrowse/database?
21:26:04bluebrotherbesided, how should rec-to-playlist work on players without a rec button? *g*
21:26:18kugelnot at all of course
21:26:37gevaertsbluebrother: easy. Take another button that you personally don't use ;)
21:26:44kugelin the same way %mo won't work on targets w/o rec button
21:26:56bluebrothergevaerts: like "stop", that I don't use on the Ipod? ;-)
21:26:59 Join miepchen^schlaf [0] (
21:27:02Lloreankugel: I find it pretty quick to insert files into the current playlist already.
21:27:18Lloreankugel: It will work, it'll just be slower.
21:27:19bluebrotherkugel: wrong. %mo would work through the context menu. As the playlist insert / view does right now
21:27:50 Quit robin0800 (Read error: 104 (Connection reset by peer))
21:27:51kugelbluebrother: of course. And I didn't remove to remove that
21:27:51linuxstbI've got a great idea, why don't we make the record button configurable?
21:27:56*linuxstb runs, very fast...
21:28:19kugellinuxstb: lol, there was actually a task doing that which got closed
21:28:25 Join fml [0] (n=4fd3ecb1@gateway/web/cgi-irc/
21:28:26bluebrotherlinuxstb: you should ping Rincewind about that!
21:28:36*Llorean brings out the pitchforks and (flaming) torches
21:28:37*gevaerts has a better idea. Button configuration in the WPS, so that each %mo mode has different button bindings
21:28:42kugelthat rhincewind is still sad about the rejection
21:28:46*gevaerts runs even faster
21:29:21Lloreankugel: I don't know why. He was pretty much told the first time the idea was even brought up, we're not fond of configurable buttons.
21:29:35LloreanI guess we left the task open too long, so he got his hopes up?
21:29:44fmllinuxstb: have you looked at FS #9557? The fact that nobody has noticed the bug until now shows that either all users use the default or that the tag isn't used at all. I assume the latter :-)
21:30:24 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
21:31:08gevaertsfml: your fix isn't entirely correct
21:31:20*gevaerts looks again
21:31:22gevaertsIgnore me
21:31:48bluebrotherLlorean: I remember him getting really excited after devcon 2 years ago where a consensus was that we should do something with the rec button ... he even wondered if his patch was accepted by that
21:32:24fmlgevaerts: because it's correct? Or simply ignore?
21:33:01kugelbluebrother: he also doesn't understand the dev's view
21:33:10linuxstbfml: I'm not sure if your patch is an improvement though - I would be tempted to simply do "HZ*10" in wps_parser.c...
21:33:29linuxstbOr rather, something else...
21:33:38gevaertsfml: I didn't read the patch correctly
21:33:49funmanis there any reason not wrapping lines the panicf() way in splash() ?
21:33:56linuxstbfml: In fact ignore me as well...
21:34:07 Quit amiconn (" bbll")
21:34:32 Part Llorean
21:34:59fmllinuxstb: I tried to unify how the timeout is stored for the sublines and for the volume button. For the sublines (old code), it's stored in 0.1 sec units. But the volume button (new code) did it in 0.01 sec units.
21:35:02 Nick Bensawsome is now known as GrammarNazi (n=Bensawso@unaffiliated/bensawsome)
21:35:31fmlI can ignore you both but the code should be corrected :-)
21:35:36 Nick GrammarNazi is now known as Bensawsome (n=Bensawso@unaffiliated/bensawsome)
21:35:45funmanwell it seems there is wrapping code ..
21:36:25kugelwhat's the "panicf() way of line wrapping"?
21:36:47funmanjust look at panicf() in firmware/panic.c
21:36:51kugel\n didn't work for me
21:36:56gevaertskugel: look at a line, panic when seeing how long it is, and start wrapping
21:37:05kugelfunman: I can always look in the code thanks!
21:37:44*kugel doesn't find that very helpful
21:37:54kugelthat answer that is
21:38:08funmanwell the code is explicit .. when we reach the end of line we go to the next line
21:39:00funmansplash() seems to truncate and dump what was out of the screen
21:39:46linuxstbfunman: splash() should split lines - what happens?
21:40:06linuxstbYou also have a splashf()
21:40:27 Join draft [0] (
21:40:44funmancan't open "/.rockbox/rocks/games" - perhaps I was wrong thinking the path was longer than what I see
21:41:04draftif i turn my iPod off with pressing PLAY/PAUSE button for long time and i'm in iPod mode.. how can i boot the machine completely so rockbox loader loads?
21:41:39funmanwell the directory exists
21:42:28linuxstbdraft: The Apple firmware has no "off" option. You have to force a reset by holding MENU+SELECT
21:43:18funmanI'm trying to launch chessbox, since all the plugins couldn't be built I may miss a file in .rockbox, but I don't know which one
21:43:23draftfunman: thanks
21:43:29draftthat was just what i was looking for
21:43:48funmandraft: you're welcome ;)
21:44:16 Quit draft (Client Quit)
21:45:34funmanlinuxstb: splash() calls splashf() and splashf() splash_internal() so both should split too long lines
21:46:10linuxstbfunman: Looks like splash* only wraps on spaces though...
21:46:16 Quit balou (Remote closed the connection)
21:46:22 Join balou [0] (
21:46:23***Saving seen data "./dancer.seen"
21:46:31funmanoh ..
21:46:42kugeldoesn't splash* also center the text?
21:50:53 Part dany_21a_
21:52:21 Join Horscht [0] (n=Horscht@xbmc/user/horscht)
21:52:40funmanis chessbox meant to be built on another target than simulator, archos, and iriver ? See line 32 of apps/plugins/chessbox/Makefile
21:52:45linuxstbfunman: chessbox is a special case (like rockboy) - it has a small "loader" plugin which is built on Archos, and then that loader plugin loads an overlay into the audio buffer. (the plugin buffer is only 32KB on Archos).
21:53:22funmanThe Makefile mentions 'iRiver target' but I suppose it means 'all but archos'
21:53:27linuxstbfunman: I guess "iriver" just means "all other targets". At the time the plugin was written, there was probably only archos + iriver
21:53:40 Quit meven (Remote closed the connection)
21:53:47linuxstbSo yes to your question.
21:55:38funmanThanks. I suppose the problem is that the 'archos check' in plugins/SOURCES is 'MEMORYSIZE <= 8'
21:56:55kugelso my fuze will count as archos?
21:56:57 Quit fml ("CGI:IRC")
21:57:54linuxstbfunman: Maybe the Clip will also need a loader for chessbox - we still need to decide how to allocate RAM between plugins, codecs, and core Rockbox.
21:58:14 Quit esthar ("KVIrc 3.4.0 Virgo")
21:58:25funmanI was wondering the same, but wouldn't make any decision without the advice of experienced rockboxers :)
21:58:31*domonoky has at moment 214k buffer for audio on m200v4 and no plugin buffer :-)
21:59:01*funman will not tell anything except that he's looking at a starfield
21:59:25domonokyfunman: how big is your plugin buffer ?
22:00:15funman512k and 512k for codecs (they seem to load but there is no progress so I'm not sure if decoding happens)
22:01:31domonokyits difficult to see if codecs work without working audio. :-) how much buffer is left ?
22:01:38linuxstbfunman: Decoding probably happens, but the pcm buffer never empties, so nothing plays?
22:02:08kugelwell, seems we'll need Nico_P's patch
22:02:11 Join gromit` [0] (
22:02:13funmandomonoky: how could I tell so ? (by the way the 'stack levels' are out of screen)
22:02:20funmanlinuxstb: s/?/./ ?
22:02:41domonokyfunman: system-> rockbox info
22:02:43linuxstbfunman: Implement the pcm driver? ;)
22:03:15funmandomonoky: 'Buffer: 209KB'
22:03:22funmanlinuxstb: when the starfield will collapse :)
22:03:51 Join reacocard [0] (n=reacocar@WL-112.CINE.HMC.Edu)
22:04:50gevaertsImplement a dummy pcm driver, and use the oscilloscope plugin?
22:05:54funmanI would prefer to implement the correct one first, we have loads of docs
22:06:06linuxstbfunman: Have you looked at the "debug audio threads" debug screen?
22:07:46funmanif you mean 'debug -> view buffering thread' it's empty
22:08:38linuxstbI'm sure it used to have "audio" in the name....
22:09:01*kugel is not so sure
22:09:12linuxstbYes, I think it's "View audio thread"
22:09:30linuxstbShould appear after "view buffering thread" (but not in the sim)
22:13:15 Quit bmbl ("Woah!")
22:14:09 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
22:15:45 Join jhulst [0] (n=jhulst@unaffiliated/jhulst)
22:16:38 Quit PaulJam (".")
22:20:06 Quit mrkiko ("Lost terminal")
22:21:25 Quit nplus (Remote closed the connection)
22:22:19 Quit miepchen^schlaf (Remote closed the connection)
22:22:59 Join miepchen^schlaf [0] (
22:24:13 Quit kugel ("ChatZilla 0.9.83 [Firefox 3.0.4/2008102920]")
22:26:18jhMikeSTry this patch. It uses hardware fading for the S:
22:27:49funmanI just found a 'bug' in minesweeper: discovering an already discovered position doesn't discover all adjacent mines if the position is next to X flags, where X is the number of adjacent flags
22:28:30linuxstbfunman: Stop playing with the plugins ;)
22:30:04 Join Llorean [0] (
22:30:17UnhelpfuljhMikeS: who is supposed to try this patch? ;)
22:30:37*linuxstb can't access for some reason...
22:31:10funmanthe reason is "you shouldn't use because some people can't access for some reason", I suggest another pastebin site
22:31:35jhMikeSUnhelpful: anyone that wants to. it's probably better if they own the beast as well though.
22:31:52funmanAn helper to find why access to the FAT filesystem is buggy on sansa ams could be a plugin though..
22:34:11UnhelpfuljhMikeS: probably would help if i used fading. but i can turn it on to test.
22:35:03funmanooh .. the sansa ams filesystem contains 'weird' folders. I wonder if this is our problem
22:35:09 Quit pixelma2 ("-")
22:35:24 Join pixelma [50] (i=pixelma@rockbox/staff/pixelma)
22:35:36funmanIf you can decode dosfsck output:
22:36:29funmannothing about 'dir entry' (I don't know what is a dir entry if it's not a 'DIR' C type)
22:37:36gevaertsfunman: you mean the ##* ones? IIRC the sansa v1 also has things like that
22:37:58funmanyes these ones, and perhaps the 'SANSA CL.IP' I hadn't noticed
22:39:12 Quit bmbl ("Woah!")
22:39:51linuxstbfunman: You have a partition table on your Clip?
22:40:44funmanI had, but I formatted it with the OF. It used to have a superfloppy partition when I bought it, but now it shows a partition table with 1 FAT partition; I'm not sure if it's the one I had created however
22:41:05funmanit's not. It's fat16 and I had created fat32
22:41:11pixelmaabout the button combo discussion earlier - I also find combos slow and there are some targets where even button combos are limited and won't give you more ways to control (thinking of the Iaudios as worst example where they are basically impossible)
22:41:26 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
22:41:54 Join kugel [0] (n=chatzill@unaffiliated/kugel)
22:42:04 Join tyfoo2 [0] (
22:42:49funmanI'm not sure what fat_file and fat_direntry structs are. fat_direntry has a 'name' member but not fat_file; and only a fat_file is used in update_short_entry()
22:44:54funmanstrangely enough, update_short_entry() is only called from fat_create_dir() , fat_closewrite() and fat_rename(). And we have no write support for sansa AMS; perhaps at creation of the settings file ?
22:46:49gevaertsWhat do the tcc ports do? They don't have write support either?
22:47:57funmanthey have a dummy nand_write_sectors() returning -1
22:48:24funmanwhile sansa ams' write() returns 0
22:49:34funmanwhich probably means 'ok write succeeded, now if some bug happens just crash'
22:49:50gevaertsprobably :)
22:50:36funmanlet's consider this 'fixed'
22:51:02funmannow the hard question: can I disable all interrupts when transferring a sector from the SD card ?
22:51:42gevaertsI guess you can for now, but it probably isn't a good idea to do that in a polished driver
22:51:53 Join wingmanz [0] (n=wingmanz@
22:52:32amiconnfunman: Why is there no write support?
22:52:46funmanSomeone didn't write it
22:53:00*amiconn would think that for SD it should be pretty straightforward
22:53:14funmanshould be, yes, especially with the PP code
22:53:25amiconnThe nand driver is different, of course. Wear levelling...
22:53:42wingmanztoday set up rockbox to my sansa player and learned how to make themes, now working on some.. kinda fun
22:53:48domonokywrite acceess shouldnt be too hard, but first we should try to get read support really fixed :-)
22:54:09linuxstbAlthough bugs in writing could potentially brick the device...
22:54:24amiconnWell, isn't it fixed? The panic is due to rockbox trying to write iiuc
22:54:48martian67hey is usb on portal player ever going to get sorted?
22:55:05martian67i dont understand why they are holding it back for all targets, even though its only causing issues on one target
22:55:08*amiconn takes out his crystal ball
22:55:23gevaertsmartian67: it's causing issues on all portalplayer targets
22:55:24amiconnmartian67: It causes problems on *all* PP targets
22:55:32 Join MegafEee [0] (n=Linux@unaffiliated/megaf)
22:55:35martian67the wiki misleads me :(
22:55:36domonokyamiconn: i still get fifo overflows when reading (although thy are fixed by retrying) so something is still not right..
22:56:03funmandomonoky: with interrupts disabled ?
22:56:17amiconnToo bad that [IDC]Dragon never made a patch from his further cut-down work for bootloaders. One part of it was making the file system read-only with a #define... :/
22:56:24 Quit tyfoo (Connection timed out)
22:56:26domonokyfunman: nope, didnt try that...
22:56:31wingmanzgot some usb problems as well. when i plug in player while it`s turned on with rockbox, my win Vista just dont recognize the player at all. can access player only when plug in then turn on to default firmware
22:57:12linuxstbwingmanz: That's expected - Rockbox doesn't have USB support, so you have to use the default firmware.
22:57:25linuxstb(USB support on the Sansa)
22:57:29funmandomonoky: I would like to commit a patch with interrupt disabled tonight, until someone writes DMA support. But I'm not sure how to handle CCU_IO register for targets with SD slots (e200/fuze/c200)
22:57:52wingmanzlinuxstb: tnx for info
22:57:55amiconnDisabling interrupts can't be more than a debug solution
22:58:07domonokyfunman: put it into a #ifdef HAVE_MULTIVOLUME ?
22:58:17 Quit Nibbler (Read error: 113 (No route to host))
22:58:22amiconnYou won't be able to have sound with that... and imprecise tick is also undesirable
22:58:28funmandomonoky: sure, but we need to check drive value for each transfer then ?
22:58:32pixelmawingmanz: it should automatically reboot into the original firmware though but it doesn't work in some builds (known bug, unknown how to fix it yet)
22:58:36funmanamiconn: I agree
22:59:02domonokyamiconn: yes, but better that non-working... it allows us, do develop easier the other needed drivers...
22:59:05wingmanzwell my sansa e280 just crushes when i plug in with rockbox.
23:00:17domonokyfunman: why would we need that ? the CCU_IO thing is only in the sd_init() ...
23:00:18amiconndomonoky: Not really.
23:00:28wingmanzand got latest release from page.. eh whatever.. anyway rockbox is so much better than default firmware
23:00:44funmandomonoky: how sure are you it's not needed for data transfers ? I didn't test this code on a SD slot
23:01:00gevaertswingmanz: yes, that's what pixelma described
23:01:46funmanif it doesn't harm transfers from the embedded SD, we can leave it in this position and modify it when we use the button led only
23:01:54domonokyfunman: what do you mean ? CCU_IO is only set once, and it only may be needed on targets with external sd...
23:02:36funmandomonoky: it's only set once in my _untested_ code, it may be buggy; who knows ?
23:02:45domonokyfunman: ah, now i understand... no, i dont know if it harms the access to the internal sd.. it might...
23:03:31 Part linuxstb ("Leaving")
23:03:33domonokyat least the first time when i remove it, i got a big improvement in stability.. but i am not really sure...
23:03:35funmandoesn't seem, but perhaps button light support on e200 will be buggy
23:04:05domonokyi think e200v2 dev will have solve this :-)
23:06:12funmanI want to commit this: (full of panicf)
23:06:14jhMikeSUnhelpful: it is backlight fading but using the hardware instead of threads
23:07:21domonokyfunman: you could reduce the sd-stack a bit... i pushed it a bit because of a stack overflow.. but now its too big :-)
23:08:07 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
23:08:38funmandomonoky: is the previous value fine ? anything else about this diff ?
23:09:18domonokyfunman: previous value was a bit too low, so maybe something in between ? :-)
23:09:37domonokyabout anything else, im stil reading :-)
23:10:23funmanif you agree, you can fix it and commit yourself (just mention that I used s/MMC/MCI/ for the registers to avoid confusion with a MMC and not SD controller)
23:10:53 Join RomanGwizdz [0] (n=4556dc69@gateway/web/cgi-irc/
23:11:03 Quit merbanan (Remote closed the connection)
23:11:15domonokyi have to go to bed now...
23:11:29domonokyso please commit it your self, seems to be fine..
23:13:54 Join Thundercloud [0] (
23:14:22domonokymaybe also comment out the yield thing in wait for state. this would now yield() while interrupts disabled, which might not be too god.
23:15:50wingmanzis there any info when jpgs will be supported on rockbox. no album arts :( guess have to convert them all to bmps
23:16:22 Quit RomanGwizdz ("CGI:IRC (Ping timeout)")
23:16:37gevaertswingmanz: as soon as someone implements jpg support in a way that's acceptable for the rockbox core
23:17:32 Join RomanGwizdz [0] (n=4556dc69@gateway/web/cgi-irc/
23:18:25RomanGwizdzand have some questions
23:19:51advcomp2019RomanGwizdz, what questions
23:20:58 Quit domonoky (Read error: 104 (Connection reset by peer))
23:22:54RomanGwizdzi created new theme for rockbox and I want to share, but I cannot attached my work, anybody can explain how to get access to upload page
23:24:32gevaertsRomanGwizdz: I'll add you to the list
23:24:53wingmanzRomanGwizdz: same here
23:25:26wingmanznow searching for info about sharing my themes
23:26:13gevaertsRomanGwizdz: done. Now please don't spam :)
23:26:41RomanGwizdzthat means that I can attached this what i did?
23:26:57gevaertsYou can now edit the wiki, yes
23:27:45RomanGwizdzis that means that now I have access to upload page?
23:28:06RomanGwizdzthank you
23:33:10 Quit RomanGwizdz ("CGI:IRC (Ping timeout)")
23:35:04 Join jeffdameth1 [0] (
23:37:52 Join tvelocity [0] (
23:38:02 Quit {phoenix} (Read error: 104 (Connection reset by peer))
23:40:11 Quit jeffdameth (Read error: 60 (Operation timed out))
23:40:22 Quit miepchen^schlaf ()
23:46:25***Saving seen data "./dancer.seen"
23:47:00 Nick JdGordon|zzz is now known as JdGordon (n=jonno@rockbox/developer/JdGordon)
23:51:25 Quit bmbl ("Woah!")
23:57:45 Quit culture (Read error: 110 (Connection timed out))

Previous day | Next day