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 2010-06-18

00:01:11 Quit voRia (Ping timeout: 258 seconds)
00:03:16*domonoky1 sees FS #11414 and looks at app/metadata/mod.c and sees a comment: /* Copy Title as artist */ so thats intended behavior.. but why ?
00:03:29domonoky1preglow: ping ?
00:03:46jhMikeSdoes preglow exist anymore or is that just a data ghost?
00:05:19domonoky1so should we change that so that a mod title is actually the title in metadata and not the artist ?
00:05:35***Saving seen data "./dancer.seen"
00:08:43jhMikeSdoes it even have an artist field?
00:08:52kugeldomonoky1: the comment seems weird. if it wasn't there I'd say "sure go ahead"
00:10:57 Quit jgarvey (Quit: Leaving)
00:12:15CIA-85New commit by alle (r26901): Do not produce additional space after 'to the picture above'
00:12:16jhMikeSummm...just see if it works? :)
00:13:02 Quit simabeis (Ping timeout: 252 seconds)
00:13:20 Join voRia [0] (
00:14:11CIA-85r26901 build result: All green
00:14:56wodziriver_flash.c doesn't look safe to use. It introduces delays by simple C busy loop. The story with D2 shows this is not the way to go.
00:15:00pamaurydomonoky1: from the description of the mod file format I found, only the title can be stored. So putting the title in the artist field seems rather bizarre
00:15:44pamaurywodz: I think there are such loops everything. I recently saw once in something as2525 related
00:15:54bertrikI think we can make the amsv2s quite easily run at 248 MHz too, just like the sansa ams players, I'll try it out now
00:16:36wodzpamaury: This doesn't mean it is safe :-)
00:16:59kugelbertrik: wouldn't that be a bit off too?
00:17:17bertrikyes, but already a lot better
00:18:08kugelwhy not try out the 243MHz suggestion from saratoga first?
00:18:27bertrikI don't see why
00:18:33saratogawell if both give low error, might as well use the amsv1 settings
00:19:13kugelbertrik: lower error?
00:19:33saratogai meant closer to 44100 hz
00:20:08wodzhaha, the comments about busyloop timing are the same in rockbox_flash.c and iriver_flash.c. I bet running time of the loop *IS* different on sh and coldfire
00:21:12bertrik240 MHz gives 1.1% error, 243 MHz gives 0.11%, 248 MHz gives 0.15% error
00:21:55bertrikThe change from 240 to 248 is quite simple, just adjust the multiplier from 60 to 62
00:22:31saratogathen might as well do that and get this fixed
00:25:24kugelbertrik: 61 as multiplier would work as well I guess
00:25:43kugelbut 0.15% is quite good already
00:26:11kugelI have been happy with the v1 sound quality, so doing the same on v2 can't be bad :)
00:26:53 Join halmi [0] (
00:26:59saratogathe oscillator probably isn't much more stable anyway, so no sense worrying about it
00:28:49kugelare you sure? I heard that even the cheapest oscillators manage 10^-4+ %
00:29:05amiconnWell, xtals do
00:29:08 Quit halmi_ (Ping timeout: 248 seconds)
00:30:18 Quit bmbl (Quit: Bye!)
00:30:42amiconnEven cheaper RC oscillators are less precise, but those aren't used for CPU clocking in our targets. They are used for stuff like LCD controller clock though
00:31:26 Quit scorche (Disconnected by services)
00:31:36 Join scorche` [0] (~scorche@rockbox/administrator/scorche)
00:32:12wodzamiconn: have You explored the idea of using EMAC unit in phases calculation?
00:32:26amiconnYes, I tested it
00:32:43amiconnIt's not worth the hassle (dealing with EMAC state saving)
00:33:15amiconnSpeedup at 124MHz is immeasurable; at 45MHz it reduces ISR load from 37% to 35% (measured on H1x0)
00:34:26amiconnI intend to fiddle a bit with the HD200 asm soon though (using more line accesses and arranging lcd controller writes so that stalls are avoided)
00:35:47 Quit pamaury (Remote host closed the connection)
00:36:03wodzwhat do You think about spline interpolation in test_grey ?
00:36:12amiconnRegarding fs #11413, I'll maybe try it tomorrow
00:36:41amiconnI can't say much regarding the actual mathematic and assume you did it correctly if I get reasonable results
00:36:50 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
00:37:22amiconnOther than that it'll hopefully still fit on archos
00:38:00 Quit Highlander (Quit: Quitte)
00:41:30wodzamiconn: Memmory cost should not be that much - 2 arrays of 16 ints each, 1 array 17 ints, 3 ints for coefficients ~108 bytes of ram usage more
00:44:15amiconnPlus code
00:44:33 Part captainewkllllll
00:46:48wodzit can be a bit smaller at cost of lower computation efficiency (but I don't know if this matter)
00:47:52amiconnI don't know either without testing - the plugin allocates the greylib buffer dynamically, so the .map file doesn't tell whether it'll actually work
00:55:07 Quit t0rc (Quit: Leaving)
00:55:50wodzhmm at least one 16 ints array and 3 variables with coefficients can be saved
00:57:13 Quit bertrik (Quit: De groeten)
00:57:25 Quit wodz (Quit: Leaving)
01:01:19 Quit kugel (Remote host closed the connection)
01:03:28 Join evilnick_BS [0] (620ec24b@rockbox/staff/evilnick)
01:10:44 Join webguest67 [0] (
01:11:11 Part toffe82
01:12:23webguest67hi, I am looking for some help regarding the 'new' eabi compiler
01:13:07 Quit bluebrother (Disconnected by services)
01:13:11 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
01:13:31 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
01:14:04webguest67I am unsure how to set things up, i have found various things talking about re-running /tools/, however I am unsure how to go about doing that
01:14:33 Quit domonoky1 (Read error: Connection reset by peer)
01:14:43webguest67I am using cygwin and set things up according to the instructions in the wiki if that helps
01:15:20 Quit rob (Read error: Connection reset by peer)
01:15:38 Join rob [0] (
01:17:11 Quit BradC (Read error: Connection reset by peer)
01:19:04 Quit webguest67 (Quit: CGI:IRC (EOF))
01:20:34 Quit voRia (Quit: Leaving.)
01:26:53 Quit halmi (Read error: Connection reset by peer)
01:32:08 Quit evilnick_BS (Ping timeout: 252 seconds)
01:45:43 Join BradC [0] (
01:48:44 Join fdinel [0] (
01:49:18 Quit efyx (Remote host closed the connection)
01:50:11 Quit BradC (Read error: Connection reset by peer)
01:50:54 Join BradC [0] (
02:01:34 Quit BradC (Read error: Connection reset by peer)
02:05:37***Saving seen data "./dancer.seen"
02:18:33 Quit DerPapst (Quit: Leaving.)
02:29:44 Join BradC [0] (
02:47:34 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
03:04:02 Quit Xerion (Ping timeout: 240 seconds)
03:05:40 Quit linuxguy3 (Read error: Operation timed out)
03:21:58JdGordonbieber: where do you get the settings list from?
03:35:04 Join fdinel [0] (
03:41:37bieberJdGordon: From the .cfg file: they're loaded when you "Open Project"
03:44:29 Quit steve|m (Ping timeout: 276 seconds)
03:44:36 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
03:44:52JdGordonbieber: sorry, no, I meant the list of possible config options
03:44:54JdGordonin the drop down
03:47:26CIA-85New commit by funman (r26902): usb-drv-as3525: build with LOGF_ENABLE undefined
03:47:31CIA-85New commit by funman (r26903): usb-drv-as3525: use udelay() and not a C busy loop ...
03:47:36CIA-85New commit by funman (r26904): sd-as3525: enable writing, sd_enable() and card_get_info_target() in bootloader ...
03:47:41CIA-85New commit by funman (r26905): as3525: bootloader USB mode ...
03:49:10bieberThere's a file in /resources, I think I called it "configkeys", that it loads from
03:49:18CIA-85r26902 build result: All green
03:49:21bieberI got the keys from an appendix in the RB manual someone pointed me to
03:49:55bieberThe file is just plaintext, one key per line, with a - for the separator between more common and less common keys (which I've kind of arbitrarily decided for now)
03:50:46JdGordonOK, just a bit worried that is just another thing which will eventually get outdated
03:51:02CIA-85r26905 build result: All green
03:52:01bieberIt's easy enough to edit, I'm not sure how you could automate it. Is there a list somewhere in the RB sources of valid keys? If so, I'm sure I could just load that
03:52:24JdGordonno, there isnt
03:52:34 Join fdinel [0] (
03:52:34JdGordonsame problem with the manual though
03:52:58bieberAt least they're also text edit boxes, so you can always enter any key you want, if it comes down to it
03:53:45bieberbtw, the backdrop image isn't for WPS, is it? I think I botched that earlier today
03:54:43JdGordonthe backdrop worlks like this:... if one is set in the config then any skill will use it unless either the skin has the %X() tag (so it loads a seperate bmp) or %X(d) which uses no backdrop
03:55:05bieberOh, okay
03:55:12bieberAnd in the absence of a backdrop, just use the background color?
03:55:45bieberI'm going to start working on viewports tonight, hopefully I'll have at least some rudimentary display of them by tomorrow morning
03:58:17 Join steve|m [0] (
04:05:23 Quit elcan (Read error: Connection reset by peer)
04:05:38***Saving seen data "./dancer.seen"
04:05:38 Join elcan [0] (
04:16:32 Join Barahir_ [0] (
04:19:44 Quit Barahir (Ping timeout: 240 seconds)
04:23:36 Quit crculver (Quit: Leaving)
04:24:25 Join linuxguy3 [0] (
04:25:33 Quit BradC (Ping timeout: 264 seconds)
04:37:10 Quit pixelma (Disconnected by services)
04:37:10 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:37:11 Quit amiconn (Disconnected by services)
04:37:14 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:37:22 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:37:31 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:38:29 Join binaryhermit [0] (
05:00:18 Quit TheSeven (Ping timeout: 260 seconds)
05:04:13 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
05:09:49 Nick scorche` is now known as scorche (~scorche@rockbox/administrator/scorche)
05:10:20CIA-85New commit by jethead71 (r26906): For multiprocessor targets, do the thread_exit routine such that we don't need to rely on the compiler's good graces to have stack switching be ...
05:11:59CIA-85r26906 build result: All green
05:22:22 Join DerPapst [0] (
05:30:35 Quit Horscht (Quit: Verlassend)
05:37:35 Quit jordan` (Ping timeout: 265 seconds)
05:40:00 Join hebz0rl [0] (
05:50:57 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
05:52:00 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
06:02:57 Quit DerPapst (Ping timeout: 240 seconds)
06:05:41***Saving seen data "./dancer.seen"
06:15:18 Join BHSPitMini [0] (
06:22:53Dhraakellianmy FuzeV1 flickers (screen and wheel) when coming back from hold
06:23:15Dhraakelliansvn from about 5 days ago, r26834
06:23:39*Dhraakellian doesn't remember when it started
06:23:49 Join DerPapst [0] (
06:23:56*Dhraakellian goes to update, glance at the bugtracker, hope it's not a hardware issue
06:40:10 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
06:45:46 Quit JdGordon (Quit: Leaving.)
06:54:29 Quit jhMikeS (Ping timeout: 240 seconds)
06:55:51 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
06:55:57CIA-85New commit by ranma (r26907): Move usb-drv-as3525 defines into header file
06:57:44CIA-85r26907 build result: All green
07:05:17Dhraakellianoh, and I got 15h5 play time when I tested my new (refurb) clip+
07:05:34DhraakellianBruckner symphony cycle in q6 Ogg Vorbis set to repeat
07:07:02CIA-85New commit by ranma (r26908): Avoid ifdefs
07:07:12Dhraakellianhold on most of the time, minimal interaction
07:07:32Dhraakellianiirc, that's better than my FuzeV1 gets, but I haven't checked it in a while
07:08:13 Quit kramer3d_ (Read error: Connection reset by peer)
07:08:33 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
07:08:34CIA-85r26908 build result: 25 errors, 79 warnings (ranma committed)
07:09:00 Join kramer3d__ [0] (
07:12:25 Quit kramer3d (Ping timeout: 272 seconds)
07:12:36 Quit binaryhermit (Quit: Leaving)
07:22:25 Join stooo [0] (
07:22:41 Part stooo
07:28:44 Nick kramer3d__ is now known as kramer3d (
07:38:10 Quit BHSPitMini (Ping timeout: 264 seconds)
07:38:38 Join Buschel [0] (
07:43:03 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
07:47:46 Quit kramer3d (Quit: Leaving)
07:53:42 Quit DerPapst (Read error: Connection reset by peer)
07:54:08 Join DerPapst [0] (
08:05:44***Saving seen data "./dancer.seen"
08:09:55 Join LinusN [0] (linus@rockbox/developer/LinusN)
08:11:03 Quit Buschel (Ping timeout: 260 seconds)
08:22:10 Join ender` [0] (
08:30:32 Join einhirn [0] (
08:32:26 Quit hebz0rl (Ping timeout: 265 seconds)
08:35:06 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
08:35:58 Quit TheSeven (Ping timeout: 240 seconds)
08:37:15 Join Zagor [0] (bjst@rockbox/developer/Zagor)
08:47:18 Join flydutch [0] (
08:48:18 Join B4gder [0] (~daniel@rockbox/developer/bagder)
08:54:59CIA-85New commit by jethead71 (r26909): Squeeze down the PP5002 cache routines a bit.
08:56:46CIA-85r26909 build result: 3 errors, 0 warnings (jethead71 committed)
08:56:52bieberJdGordon: Question about viewports, if you're here
08:57:50bieberWe have the last two parameters as ints in the tag table, but don't we need a string for RGB colors?
08:58:33Zagorwow, strange error
09:01:16simonrvnwell the sansafuzev2 port built successfully
09:01:37simonrvnsvn up\nAt revision 26909.
09:01:53 Join Rob2222 [0] (
09:02:31 Join hebz0rl [0] (
09:05:55 Quit Rob2223 (Ping timeout: 252 seconds)
09:14:58 Join petur [0] (~petur@rockbox/developer/petur)
09:17:39ZagorIgor Poretsky looks like a capable guy
09:20:07B4gderwow, decent list of patches applied
09:20:30 Join pondlife [0] (~Steve@rockbox/developer/pondlife)
09:20:38Zagorand a pile of his own stuff too. and everything very tidy and orderly.
09:21:15ZagorI'm sending him a mail inviting him to join the list and irc.
09:22:28B4gdergood idea
09:23:25bertrikthe mail by Sergei says that somehow "he is not able to subscribe to the list"
09:23:43Zagoryeah, I don't know what that means. I'll just ask.
09:26:23 Quit bertrik (Quit: De groeten)
09:28:06 Quit pondlife (Quit: Leaving.)
09:36:12 Join CGL [0] (~CGL@
09:36:40 Join Kitr88 [0] (
09:38:41 Join swilde [0] (
09:39:28 Quit Kitar|st (Ping timeout: 248 seconds)
09:40:50 Quit Kitr88 (Ping timeout: 240 seconds)
09:41:29 Quit leavittx (Ping timeout: 258 seconds)
09:42:48 Quit DerPapst (Quit: Leaving.)
09:45:20 Join Kitar|st [0] (
09:49:13 Join pyro_maniac_ [0] (
09:52:15 Quit pyro_maniac (Ping timeout: 240 seconds)
09:53:13 Join wodz [0] (
09:54:38 Join mitk [0] (~mitk@
09:56:26 Quit liar (Ping timeout: 258 seconds)
09:58:07 Quit hebz0rl (Ping timeout: 265 seconds)
09:59:44 Join Rob2223 [0] (
10:01:52 Quit linuxguy3 (Ping timeout: 245 seconds)
10:02:54 Join Buschel [0] (
10:03:26 Quit Rob2222 (Ping timeout: 272 seconds)
10:03:28 Join kugel [0] (~kugel@rockbox/developer/kugel)
10:03:46 Join linuxguy3 [0] (
10:04:43 Join leavittx [0] (~leavittx@
10:05:45***Saving seen data "./dancer.seen"
10:19:56 Quit CGL (Remote host closed the connection)
10:31:17kugelwhat happened about 3.6.1 for fuzev1?
10:32:15 Quit Buschel (Ping timeout: 240 seconds)
10:32:27 Join BradC [0] (
10:41:25 Join stoffel [0] (
10:45:40 Join DerPapst [0] (
10:49:34 Nick pyro_maniac_ is now known as pyro_maniac (
10:51:41 Join wpspb [0] (
10:54:09wpspbI want to use %pb without a image, which is the correct syntax ?; 1) %pb||x|y|width|height| or 2) %pb|x|y|width|height|
10:56:01S_a_i_n_t_(well, for the new skin syntax)
10:56:29S_a_i_n_t_if you're using 3.6 or lower, then %pb|x|y|width|height| will do the trick
10:57:11S_a_i_n_t_oh...actually. no.
10:57:20S_a_i_n_t_I forgot what I was talking about.
10:57:37wpspbI have time :-)
10:57:39S_a_i_n_t_If you just want to use the inbuild playbar, then just use %pb
10:57:42 Join efyx [0] (
10:57:45kugel%pb|image.bmp|x|y|width|height| or plain %pb for the inbuild peakmeter, no?
10:58:21S_a_i_n_t_kugel: Yep, I just remembered that mayself. (playbar btw, not peakmeter)
10:58:31kugel%pb|-|x|y|width|height| *might* work, but I don't know. if not you need to fiddle with viewports for resizing the peakmeter
10:58:38kugelerr, playback, right
10:58:46S_a_i_n_t_*bar ;)
10:58:52wpspbyes %pb without a image
10:59:05S_a_i_n_t_sweet, than just use %pb
10:59:24S_a_i_n_t_as in literally, just "%pb" no other parameters
11:00:09wpspbso, how does it know where to place itself, or is that defined by it's position in the list
11:00:58S_a_i_n_t_put it in a viewport.
11:01:26wpspb*gives Saint his best blank look*
11:01:48wodzcallibrating greylib is tricky :-/
11:03:25S_a_i_n_t_IIRC, the inbuilt playbars height is defined by the height of the font used.
11:04:06S_a_i_n_t_and, if your WPS isn't using viewports, then it is positioned by the line it corresponds to in the WPS
11:04:33S_a_i_n_t_but, you really *should* be using viewports.
11:05:26S_a_i_n_t_that link will tell you what you need to know about viewports, and anything WPS related.
11:06:00S_a_i_n_t_there is also this:
11:06:24S_a_i_n_t_wpspb: ^
11:07:59wpspbaha. Viewports is way out of my league. i use a _extremely_ simple three line wps, 'cause as a non coder it's time consuming trying to figure out where the tiny syntax change is that breaks the thing
11:08:05wpspbfor info these two pages weren't very informative on thsi subject, i.e. %pb without a image ; and;
11:09:10kugelthen just put a %pb into the wps somewhere and it'll show somewhere :)
11:09:35wpspbthat's perfect. thanks
11:09:44*pixelma already wondered if CustomWPS had been updated to the new syntax
11:09:52pixelmadoesn't look like it
11:09:56S_a_i_n_t_yes, as I said earlier...if you're not using viewports then the position is defined by the line it corresponds to in the WPS code.
11:11:04S_a_i_n_t_if you put %pb as the first line of the WPS, it will be on the top line, if you put it on the second line, then it'll be on the second "line" in the WPS and so on.
11:11:41pixelmaby the way, I still can't follow why colour is not a viewport parameter but separate
11:11:55S_a_i_n_t_a "line" is defined by the font height. for example, if your font is 10px high, and the players screen is 100px high, then there are 10 "lines" on the screen.
11:12:45S_a_i_n_t_pixelma: Niether can I really...but I think it boils down to the fact that the colour *can* change, or, *could* change rather.
11:13:03S_a_i_n_t_There's no reason it needs to be permanantly defined by the viewport.
11:13:27S_a_i_n_t_I think the intention is to be able to make the FG/BG colours conditional?
11:13:47pixelmayou could win me over if that makes it simpler to have inverse draw mode, i.e. white on black on monochrome more easily
11:14:13S_a_i_n_t_I think this is possible with the new parser.
11:14:22S_a_i_n_t_But don't quote me on it ;)
11:15:23S_a_i_n_t_wpspb: Did you see my description of line placement for non viewport WPS' above?
11:16:32wpspbyes thanks, nice info, it would be nice to hve that in the wiki
11:16:58S_a_i_n_t_I believe it is...but, not using viewports is generally frowned upon.
11:17:19wpspbjust call me simple
11:17:31*S_a_i_n_t_ calls wpspb simple
11:18:33S_a_i_n_t_Using viewports makes it less likely that doing something as simple as changing the font doesn't completely break the alignment of the theme.
11:19:16S_a_i_n_t_an example being if you have the playbar as the very bottom line, and then change the font height, it will almost definitely be pushed off the screen.
11:22:10S_a_i_n_t_wpspb: The manual for your player also has a nice description of WPS code in it.
11:23:01pixelmasince when is not using viewports "frowned upon"?
11:23:51wpspbI understand that there are advantages to the advancements in the wps, but call me old fashioned or simple, but as a non coder I really enjoy the simplicity of a three line wps, complex syntax changes and concept changes are almost impossible for me to relsove. On my H140 simple is good Artist, Title %pb, Album, that's it
11:24:08S_a_i_n_t_well, it's just...ugly. especially for the reason I mentioned regarding font change vs. alignment.
11:24:46S_a_i_n_t_pixelma: ^
11:24:57wpspbI don't change font at all, nimbus 12 rules :-)
11:25:22S_a_i_n_t_wpspb: The syntax change actually makes the WPS code easier to read.
11:25:35S_a_i_n_t_(and write)
11:26:23wpspbyes but code IS a language and as in any language the nuances are only appreciated by those who know how to speak it
11:26:29S_a_i_n_t_But, as you said if you're just using a three line WPS...then you probably don't really care much :D
11:27:09S_a_i_n_t_If you break it down into sections...then WPS code really is very simple.
11:27:09wpspbthat sums it up quite nicely :-)
11:27:16pixelmaS_a_i_n_t_: that sounds more like your opinion. Purely text based WPS are at least a bit more flexible
11:27:54S_a_i_n_t_pixelma: Until you change the font height...
11:28:14S_a_i_n_t_then bits start magically dissappearing off the screen.
11:29:02 Join n1s [0] (
11:29:03 Quit n1s (Changing host)
11:29:03 Join n1s [0] (~n1s@rockbox/developer/n1s)
11:29:09S_a_i_n_t_Even if a WPS is simple text, it should still use viewports if it is supported by the target.
11:29:20pixelmadepends how much space you have left. And the same problem would exist with viewports too, even making it worse
11:29:27S_a_i_n_t_Even if it is solely for that reason. (font change)
11:30:55S_a_i_n_t_with viewports (if used correctly) changing the font height will just lose some text. Not push the playbar or whatever off the screen.
11:31:07S_a_i_n_t_the placement of other items will still remain the same.
11:32:50AlexPS_a_i_n_t_: Viewports are great and all, but I don't recall not using them being frowned on
11:33:04AlexPThis all sounds like your opinion :)
11:34:16pixelmaIIRC text will completely disappear in a viewport if it would exceed its limits
11:34:22S_a_i_n_t_When I first showed up here...I got a semi-lecture from Jd about "the beauty of viewports, and why not using then is bad", and after looking into it, I happen to agree.
11:35:22S_a_i_n_t_pixelma: How do you mean "completely dissappear"? If it can't draw a full line, it wont. But it will draw all the full lines it can.
11:35:49 Join dfkt [0] (dfkt@unaffiliated/dfkt)
11:36:50n1sS_a_i_n_t_: if the height of the viewport is less than the font height
11:36:53pixelmawhere's the difference to not using viewports then? If you do use them, they are usually smaller than fullscreen, so things will start disappearing earlier
11:37:40AlexPS_a_i_n_t_: Sure, they are great, but there isn't a position on it. You just seemed to come across a bit strong
11:37:45pixelmaor just differently
11:38:16 Quit kugel (Ping timeout: 260 seconds)
11:38:54AlexPThere isn't a hould use viewports about it - if they aren't needed, then great
11:40:03*S_a_i_n_t_ wishes JdGordon were here to chime in...
11:40:35S_a_i_n_t_The talk I was given when I first showed up equated to "drawing in the default viewport is evil"
11:40:53pixelmayeah, maybe I'm more annoyed that it came across as speaking for everyone here
11:40:56AlexPI fully agree that viewports are great, and the only way to do it if you want any sort of proper control over position
11:41:23AlexPAnd in most circumstances they should be used, yes imo
11:41:38AlexPBut you can't speak for everyone
11:42:18S_a_i_n_t_I didn't...I understand how it was percieved that way, and I do wish I worded it differently now, but its down to perception.
11:42:40AlexPAnd it isn't a big deal either :)
11:42:49S_a_i_n_t_No, it isn't ;)
11:42:52 Join Highlander [0] (
11:44:26pixelmaIMO, WPSs without using viewports (or just the default viewport) still have some advantages depending on what you want and should still be "supported". If you start using viewports, drawing in the default viewport is another thing if it causes complication (in the code or for screen updates etc.)
11:45:32S_a_i_n_t_Heh...well, as we found out last night that's not even supposed to work ;)
11:45:52pixelmawe did?
11:45:54S_a_i_n_t_(drawing in the default viewport whilst another viewport is specified)
11:46:26wodzI ploted lcdlinear matrix for various targets -> this at least shows what should one expect :-)
11:46:34S_a_i_n_t_And yeah, we did.
11:51:16 Join hebz0rl [0] (
11:55:14 Nick fxb__ is now known as fxb (
11:58:03 Join Piggz [0] (
11:58:24 Quit wpspb (Quit: CGI:IRC)
11:59:22Piggzhi...a quick question, my ipod 5gen video, even after a restore, just shows the apple logo on startup. So I installed Rockbox and it works fine, any idea why?
12:02:18wodzapple logo comes for early bootloader which is not touched with standard installation method
12:03:21 Join watto [0] (~watto@
12:03:32S_a_i_n_t_Piggz: Do you mean that when you try to boot the Apple firmware, it just shows that Apple logo, and nothing more?
12:03:42S_a_i_n_t_(ie. doesn't finish booting)
12:05:24Piggzwhen I put it into disk mode, and used itunes to restore it, it reboots and sticks at the apple I put it in disk mode again, installed Rockbox, and it loads Rockbox ok.....
12:05:38 Join simabeis [0] (
12:05:46***Saving seen data "./dancer.seen"
12:05:52PiggzIf I then do do boot trick to load the apple fw with the hold button, it doesn't boot
12:09:19S_a_i_n_t_That is quite odd...
12:09:31S_a_i_n_t_At least it can still boot Rockbox ;)
12:09:53 Part watto
12:14:23Piggzyes :) I got it from ebay as faulty...didn't cost much, so I'm happy Rockbox runs on it
12:26:14 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:30:00 Join robin0800 [0] (
12:34:35pamauryWhat is this "bootloader usb mode" that funman committed ?
12:42:11 Part LinusN
12:43:01n1susb mode in the bootloader?
12:44:09ranmapamaury: AFAICS he changed the bootloader so that it will load the usb_storage driver if usb is connected while the bootloader starts up
12:44:20pamauryWhy ?
12:44:34ranmaDon't ask me, I was kind of wondering too :)
12:47:06ranmaBTW usb storage reads work, but on writes I get a panic, because the driver tries to receive dma a packet from the host, when rockbox hasn't called recv() yet. :/
12:47:50ranmaIn theory I think setting NAK on the EP should prevent that, but for some reason it happens anyway...
12:53:36pamauryI have the complementary problem on clip+, whatever I do, the usb phy doesn't trigger an interrupt on setup packet :(
12:54:08 Quit petur (Quit: *plop*)
12:54:44 Quit robin0800 (Remote host closed the connection)
12:58:10 Join DerPapst1 [0] (
12:58:46 Quit stoffel (Ping timeout: 265 seconds)
13:00:10 Quit DerPapst (Ping timeout: 240 seconds)
13:01:31ranmapamaury: BTW, did you have a look at drivers/usb/gadget/s3c-hsotg.c? I read yesterday on lwn that's also a DesignWare part. Register names seem to match Clip+
13:03:32pamauryNo, I already have the linux patch so I basically know all the registers. I'll have a look at it but currently I'm stuck and I don't know why it doesn't work.
13:07:03ranmaThe linux patch usb_otg_if.c doesn't seem to implement packet recv/transmit AFAICS on a quick glance though :)
13:08:19ranmaCompared to that the official s3c-hsotg.c driver should be quite working (used in then OpenMoko mobile phone)
13:10:09ranmaAnd since it's part of the normal vanilla Linux kernel it's GPL of course
13:10:40 Join dfkt|x [0] (
13:10:40 Quit dfkt|x (Changing host)
13:10:41 Join dfkt|x [0] (~dfkt@unaffiliated/dfkt)
13:17:33 Join jordan` [0] (
13:18:03 Quit dfkt|x (Remote host closed the connection)
13:22:08 Join teru [0] (
13:26:48 Quit B4gder (Quit: It is time to say moo)
13:41:47pamauryranma: usb_otg_if.c is the init code, the rest of the driver is in drivers/usb/gadget and it's quite big :)
13:42:19 Join watto [0] (~watto@
13:58:28CIA-85New commit by wodz (r26910): HD200 - calibrate lcdlinear[] matrix
14:00:16CIA-85r26910 build result: All green
14:01:12 Quit hebz0rl (Ping timeout: 276 seconds)
14:05:50***Saving seen data "./dancer.seen"
14:08:16 Join grawity [0] (grawity@
14:14:50 Join robin0800 [0] (
14:18:34grawityIs there a way to remove some characters from a .fnt, and only keep ASCII? (It takes a long time to load a .wps with 3 fonts...)
14:19:38 Quit wodz (Quit: Leaving)
14:21:37 Join petur [0] (~petur@rockbox/developer/petur)
14:28:34CIA-85New commit by mcuelenaere (r26911): Lua: make actions.lua, buttons.lua and rocklib_aux.c depend on it generators
14:28:36CIA-85New commit by mcuelenaere (r26912): Lua: document a bit, so it's easier to find out about it when stumbling over a CC error in rocklib_aux.c
14:29:12 Quit robin0800 (Remote host closed the connection)
14:30:20CIA-85r26911 build result: All green
14:30:55mitkgrawity: Take a look at point 4.
14:31:42CIA-85r26912 build result: All green
14:34:53grawitymitk: I see... however, in this case, one font is only used for numbers; would it not be possible to have two copies (full and ASCII)? It's mostly for my personal use, anyway.
14:36:34 Join ender1 [0] (
14:38:20mitkgrawity: For personal use: Edit *.bdf file in trunk/fonts directory with FontForge, compile rockbox and you will get custom build with your custom fonts.
14:39:59 Quit ender` (Ping timeout: 276 seconds)
14:41:01mitkgravity: IIRC font with numbers only is used to display frequency on FM screen for targets with really big displays.
14:42:13grawityWell, it is used differently here.
14:42:19grawity - time and playlist position
14:44:29mitkgravity: I sad IIRC ;)
14:45:26*grawity tries to make a nicer volume indicator
14:47:19 Join funman [0] (~fun@rockbox/developer/funman)
14:49:01funmanranma, pamaury: USB in bootloader can be helpful if the filesystem is corrupted / .rockbox is absent
14:50:04pamaurybut it makes the bootloader much more complicated...
14:50:30funmansome other targets do it, but I don't understand how mass storage works at all here, i think these bootloaders haven't been tested in some time
14:51:22funmanranma: i could write an empty file yesterday but copying rockbox.sansa didn't succeed
14:51:36pamauryYou have to include the whole usb stack
14:51:55pamauryWhich is not simple piece of software
14:52:05ranmafunman: Got the BNA panic I expect?
14:52:23funmanscreen was off
14:52:35funmanpamaury: well the code is there already
14:52:37pixelmamitk: the font which only contained numbers was taken back first (and then made into a more complete font)
14:53:07ranmafunman: Anything against calling _backlight_on() in system_exception_wait()
14:53:08pixelmaand it's not used anywhere by default yet
14:53:13funmannow i just need to see why the disk isn't reported correctly
14:53:15mitkpixelma: 35-Nimbus?
14:53:47funmanranma: yes we talked about this with amiconn: backlight can be more complex than just setting a GPIO pin
14:54:10 Quit bieber (Ping timeout: 260 seconds)
14:54:13pixelmamitk: yes, IIRC
14:54:42funmanalso we would need to enable the LCD again
14:54:51funmanif it works on all sansa ams why not
14:55:43mitkpixelma: Thanks. Btw IIRC is lifesaving abbreviation ;)
14:55:47funmanafaiu, all bootloaders using USB (but AMS) will report a device with 0 interfaces
14:56:11grawityAugh. "*PANIC* Stkov usb" when trying to use the screendump feature.
14:56:25gevaertsI'm not aware of any bootloader other than gigabeat S that uses the rockbox USB stack
14:56:37ranmaWell on C200v2 at least it works. _backlight_on() already calls lcd_enable() for C200v2 at least
14:56:57pamaurygrawity: on which target ? I thought it was fixed...
14:56:59 Quit Piggz (Ping timeout: 240 seconds)
14:57:18grawitypamaury: ipodnano2g, yesterday's build
14:58:03pamauryprobably a sector buffer hanging around on the stack
14:58:21funmaniriver h1x0, iriver h300, mrobe500, mpiohd200
14:58:23 Join DannyA [0] (
14:59:05 Join detaos_ [0] (
14:59:05gevaertssee? :)
14:59:19 Quit detaos (Ping timeout: 240 seconds)
14:59:36gevaertsmrobe500 uses the rockbox USB stack, but it boots from the OF, so I doubt if anyone has ever tried USB from the bootloader
14:59:47gevaertsthe other three use a bridge chip
15:00:37funmanisn't it the gigabeat *F* which uses usb ?
15:00:47gevaertsalso a bridge chip
15:01:17gevaertsgigabeat S uses USB from the bootloader mainly because it
15:01:21DannyAGreetings. My server is a Rockbox build client. It was working OK until yesterday, but now I just get "Your build client has been temporarily blocked by the administrators due to: has weird issues. Please go to #rockbox to enable your client again." Does anyone have any ideas?
15:01:24gevaerts's needed for a reasonably easy install
15:02:45gevaertsDannyA: it failed all builds submitted to it
15:03:11gevaertsHave you tried a manual build in the build client's checkout?
15:04:36funmanranma: just curious: what's wrong with ifdefs ? (we already use and abuse them a lot in rockbox)
15:05:28ranmaIt makes the code less easy to read if there ifdefs within a function
15:05:38n1sgevaerts: it's also very handy if you manage to break rockbox since the OF isn't helpful when you just want to transfer some files :)
15:06:03gevaertsn1s: I'd consider that to be the same case as installing actually :)
15:08:33 Quit kramer3d_ (Ping timeout: 245 seconds)
15:10:16CIA-85New commit by funman (r26913): FS #11347 by me: *dir LUA functions: luadir module ...
15:10:30ranmafunman: If the bootloader is uncached, making invalidate_dcache() and friends noops in the bootloader would also be a nice solution IMHO (and require neither ifs nor ifdefs in the usb-drv-as3525 code)
15:11:04funmanperhaps we should just enable MMU in the bootloader
15:11:47funmanwe'll need a new release for all AMSv1 when USB works so now is a good time to make bootloader changes
15:12:08CIA-85r26913 build result: All green
15:12:45 Quit petur (Quit: *plop*)
15:14:54 Join pyro_maniac_ [0] (
15:15:34 Quit pyro_maniac (Read error: Connection reset by peer)
15:18:47funmanranma: btw udelay(1000) works fine on c200v2 ? (different timer settings)
15:19:04ranmaSeems to work fine
15:19:52pixelmamitk: the Ondio build you once provided - which gcc version was that made with? (As my impression was that the ajbrec.ajz was bigger for no apparent reason)
15:20:41funmanpamaury: D2 works with eabi ?
15:20:45DannyAgevaerts: The builds directory is empty. Does that look right?
15:21:08gevaertsDannyA: that bit is probably right
15:21:27DannyASo how do I do a manual build?
15:21:43gevaertsthe way you do any build
15:21:50pamauryfunman: I think so. I didn't try the commit made by Rob but I tried a similar one I wrote and it worked. I didn't try the plugins hozever, perhaps some of them are broken
15:21:55gevaertsMake a directory (or use builds/), run configure, and run make
15:22:32 Quit pyro_maniac_ (Remote host closed the connection)
15:23:08mitkpixelma: IIRC ;) host gcc was 4.4.3. Do you still have this build? rockbox-info.txt was included in it.
15:25:19pixelmaI know, but I don't have it anymore. Host gcc doesn't matter for target builds, just for sims IIRC ;) . The gcc-sh (or however it is called exactly) is important
15:26:07amiconngevaerts: In fact many targets with usb-storage bridge do bootloader usb
15:26:55amiconnmitk: Did you build your toolchains using
15:27:01mitkpixelma: then it was gcc version used these days for standard builds.
15:27:10amiconnIf not, your sh-elf-gcc might be unpatched, which is a bad thing
15:27:46mitkpixelma: yes. version was 4.0.sth
15:28:50DannyAgevaerts: D'Oh! Now I realise what was wrong... I reinstalled Linux, but 32bit instead of 64bit. So the Rockbox tools no longer worked. I emptied the tools directory and restored it from svn. Seems to be working now.
15:30:27CIA-85New commit by funman (r26914): Sansa AMS: fix bootloader USB mode
15:31:06CIA-85New commit by gevaerts (r26915): re-allow frederico-dannya
15:31:21gevaertsDannyA: great! We'll see after the next build :)
15:31:27DannyAThank you
15:31:59CIA-85r26914 build result: All green
15:33:25CIA-85New commit by teru (r26916): use int instead of enum.
15:34:11funmanfuzev1 bootloader with usb: 104kB too large for IRAM (same number with thumb + -Os?)
15:35:01CIA-85r26916 build result: All green
15:35:12funmanah no, 104kB with thumb/Os, 116kB with arm/O
15:37:22mitkpixelma: Ooops! I sad 4.0.sth, former version for arm targets. But I'm sure it was version used for Ondio these days
15:38:09n1sheh, inserting the charger cancels a yes/no question, i guess it counts as abuttonpress
15:39:56 Quit Highlander (Quit: Quitte)
15:40:57 Join evilnick_B [0] (0c140464@rockbox/staff/evilnick)
15:43:03funmando we need full unicode support in bootloader?
15:43:13ranmafunman: How about this:
15:44:30funmanlooks good
15:45:27CIA-85New commit by ranma (r26917): Enable display and backlight on panic.
15:47:39CIA-85New commit by nls (r26918): libmad: Optimize away 2 instructions from coldfire III_imdct, no measurable speed difference.
15:49:38CIA-85New commit by teru (r26919): reorder apps/plugins/lib/SOURCES.
15:51:22CIA-85r26919 build result: All green
15:52:57CIA-85r26918 build result: All green
15:56:28Zagorhuh? out-of-order build results
15:57:23 Quit Strife89 (Quit: Reboot to Windows.)
15:58:07 Quit mitk (Quit: Leaving)
15:58:31pixelmaand r26917 hasn't been announced (and had some red in it)
15:59:28Zagor917 was aborted due to dropped network connection
16:00:43Zagorthe server actually did run 919 before 918. the question is why/how...
16:01:40 Quit guymann (Quit: brb guys)
16:01:58BagderHAL is taking over!
16:05:45 Join jgarvey [0] (
16:05:48n1si think it feels it hasn't gotten enough attention lately
16:05:53***Saving seen data "./dancer.seen"
16:06:00Zagorahh, now I see it
16:06:14funman < bss of fuzev1 bootloader sorted by size
16:07:18Zagor918 was committed before 917 was done, thus putting 918 in the "next build" slot. but then 917 was aborted due to all clients being disconnected. and when the clients came back 919 was committed, which triggered a new round for 919. and once 919 was done, the server looked in the "next build" slot and started 918...
16:08:06n1snot the most common case perhaps :)
16:08:22Zagorno :)
16:08:38n1sfunman: is the codepage thing used in the bootloader?
16:09:50funmani think not
16:10:21n1sah it comes from unicode.c
16:10:50n1si think not supporting unicode in the bootloader would be fine
16:10:58 Quit Dark_Rak3r (Ping timeout: 258 seconds)
16:11:04funmanbut fat code uses it so it needs some tweaking
16:11:32amiconnZagor: What happened to the occasional wrong build assignment?
16:11:43CIA-85New commit by ranma (r26920): C200v2 lcd controller still gets stuck sometimes, do a full controller init in lcd_enable()
16:11:46funmanfat doesn't support unicode but it supports a lot of iso encodings
16:12:01pixelmaamiconn: which wrong build assignment?
16:12:03amiconnvfat is *always* ucs-2
16:12:18funmanwhat's vfat?
16:12:28pixelmaif you mean the wrong packaging then that was fixed
16:12:35 Join Strife89 [0] (
16:12:39ranmafunman: long filename support on fat
16:13:03funmani think we don't need lcd scrolling either
16:13:11CIA-85r26920 build result: All green
16:13:31amiconnfunman: Some things are included in the bootloaders in order to avoid too much ifdefing
16:14:09funmanthat means there is some things which can be removed then
16:14:28amiconnThe question is whether they should be removed
16:14:46funmanwe must look at each thing one by one
16:15:36n1sthe player's codepage_table buffer is only 640 entries long while it's 32k entries for bitmap targets so you should just be able to shrink the buffer
16:15:40funmanhm supposedly the bootloader only needs to look at '.rockbox' + the ascii filename of the rockbox file
16:15:46funman(or OF)
16:16:23amiconnThe bootloader does need unicode support in order to decode ucs-2
16:17:00 Join Dark_Rak3r [0] (
16:17:34amiconnThere is no such thing as an ascii filename
16:17:46n1samiconn: but that works on the player with the small buffer, no?
16:17:51grawityvfat is UTF-16, not UCS-2.
16:18:22amiconnUTF-16 and UCS-2 is the same except the former also supports characters outside the BMP
16:18:37amiconnBut rockbox does only support the BMP, so it's essentially the same
16:19:05CIA-85New commit by zagor (r26921): Cumulative commit of various build server changes.
16:19:47funmanscrolling is ~40kB
16:20:18pamaurygrawity: UCS-2 is older than UTF16 iirc, it has been superseeded by UTF16 but still, this is the one used in FAT
16:20:46amiconnHuh? The whole bootloader is little more than that...
16:21:05funmanshould I use #ifndef BOOTLOADER or put a #define HAVE_LCD_SCROLL in lcd.h ?
16:21:17funmanamiconn: bss is large
16:22:36ranmaHmm, I think the udelay(1000) in usb-drv-as3525.c is likely to run into panicf("%s(): %d too high!", __func__, usecs) for very low backlight brightness
16:23:01funmanranma: perhaps we could extend udelay to support arbitrary high argument
16:23:36ranmaWith my current brightness level I haven't seen any problems, but with the lowest on the low BGLOAD value should be about 210us
16:23:47amiconnDidn't know you were talking about ram usage.
16:23:51ranmas@lowest on@lowest level@
16:24:02amiconnDoes that matter on some target(s)?
16:24:19funmanamiconn: fuzev1 384kB of iram is full when enabling USB
16:24:29funman105kB or so overflow
16:24:38funmanmostly due to the 128kB USB storage buffer
16:24:59ranmafunman: You can also reduce the usb storage buffer I think
16:25:07funmansure but that's too easy!
16:25:34funmanand afaiu it could impact USB performance (but we need to test it at some point)
16:25:58grawityWill Rockbox have some kind of "sleep" mode like the original iPod firmware does? (Rockbox does boot a lot faster, but still not instant.)
16:26:03 Join guymann [0] (~charlie@
16:27:29n1sfunman: but usb speed is maybe not the most important thinng in the bootloader ;)
16:27:40n1sgrawity: if someone implements it, sure
16:27:51n1sand it's not trivial
16:30:15 Quit DannyA (Quit: CGI:IRC (EOF))
16:33:46funmanhm some bootloaders use scrolling :/
16:34:10n1sthey do !?
16:34:13funmanjust for debug though
16:34:25n1si thought that required threads
16:34:33funmanthreads work fine in bootloader
16:35:09funmanthere's just one thing in lcd-charcell.c : lcd_put_cursor uses lcd_scroll_info.ticks
16:36:35 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
16:37:02funmandoes it blink faster when scrolling ?
16:39:02ranmaI think we could use the watchdog timer for udelay
16:41:50CIA-85New commit by mcuelenaere (r26922): Simplify check in gui_synclist_do_touchscreen(), no functional changes.
16:43:13CIA-85r26922 build result: 21 errors, 0 warnings (mcuelenaere committed)
16:44:05funmanhm archos needs scrolling for rolo
16:44:37*funman curses archos!
16:45:32funmanminimum width is 14 characters?
16:45:46CIA-85New commit by mcuelenaere (r26923): Fix error: assignment of read-only variable 'list_width'
16:46:51amiconnfunman: No it doesn't blink faster when scrolling, but the scroll tick function handles both scrolling and cursor blinking
16:47:11funmanyep i noticed that
16:47:25CIA-85r26923 build result: All green
16:47:33amiconnThe newlcd has a hardware cursor, but we don't use that because it's unknown for the oldlcd
16:47:58amiconnI have a strong suspicion regarding the oldlcd controller, but no device to verify :(
16:48:01funmani want to modify rolo to not use scrolling
16:49:13*amiconn wonders why funman does suddenly want to change stuff that's been working for years, and there has been consensus not to use special handling for the bootloader if it means too much hassle
16:49:23 Join toffe82 [0] (~chatzilla@
16:49:37funmanamiconn: read backlog
16:50:27amiconnAny reason why the bootloader uses iram only?
16:50:51funmanit's been working for years, and there has been consensus to not use special handling for the bootloader if it means too much hassle
16:52:11amiconnAnyway, rolo needs scrolling because otherwise you have no chance to read the various error messages
16:52:58funmanall messages fit in 1 or 2 lines
16:53:53funmanarchos player lines*
16:54:51amiconnNo they don't, especially not considering that rolo_error() effectively uses only one line for the actual message
16:55:22funmanlongest message is 28 characters so surely it will fit on 2 14 chars line?
16:55:34amiconnThere are 11 chars per line
16:55:37amiconnNot 14
16:56:28funmanhm i was counting 112 pixels & 8 pixels per char
16:57:14amiconnThe player is charcell, and has two lines of 11 characters each
16:57:47amiconnThe recorders and Ondios are no problem if the application part of rolo is moved to apps/ and changed to use splash() or similar
16:57:51pixelmafunman: 112 pixels are the bitmap archoses and sysfont is 6 characters wide, but there is teh Player
16:57:53funman2 lines :/
16:58:23funmanondios / recorders use bootbox too ?
16:59:11amiconnAll archoses do
16:59:21CIA-85New commit by funman (r26924): rolo is only needed in SH bootloaders, not other bootloaders
16:59:25CIA-85New commit by bluebrother (r26925): Fix source string spelling.
16:59:25amiconn(when flashed, of course)
17:01:07CIA-85r26924 build result: All green
17:01:37 Join stoffel [0] (
17:02:43CIA-85r26925 build result: 25 errors, 79 warnings (bluebrother committed)
17:03:35gevaertsThat one again :(
17:03:45*gevaerts wants that build issue to go away!
17:04:59 Quit stoffel (Remote host closed the connection)
17:06:46*bluebroth3r wonders what the build system did
17:07:53 Part Zagor
17:07:57 Quit grawity (Quit: Leaving.)
17:08:19 Join hebz0rl [0] (
17:08:38gevaertsbluebroth3r: a dependency issue, probably
17:08:45 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
17:09:35bluebroth3rdefinitely not related to my commit though :)
17:10:59gevaertsIf you see "25 errors, 79 warnings", it's likely to be this one
17:16:30 Nick ender1 is now known as ender` (
17:17:55 Join Buschel [0] (
17:23:40 Join Zigtown [0] (
17:23:46 Quit teru (Quit: Quit)
17:27:53jhMikeSamiconn: did you ever check 2g? I'm fairly sure eabi should work now.
17:39:16 Join panni_ [0] (
17:45:16 Quit Strife89 (Quit: Reboot back to Linux.)
17:48:07 Join Strife89 [0] (
17:49:47 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
17:50:41 Quit Buschel (Ping timeout: 260 seconds)
17:51:38jhMikeSinteresting. if I compile 3g using arm9tdmicc for cpu, the strange pictureflow lines go away
18:01:41 Join Highlander [0] (
18:05:57***Saving seen data "./dancer.seen"
18:07:11 Join stoffel [0] (
18:12:15 Quit panni_ (Read error: Connection reset by peer)
18:19:05 Join panni_ [0] (
18:19:16 Join robin0800 [0] (
18:20:14 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
18:23:13 Quit Zigtown (Remote host closed the connection)
18:31:56 Quit hebz0rl (Quit: Ex-Chat)
18:37:27 Quit swilde (Quit: ERC Version 5.3 (IRC client for Emacs))
18:38:04 Quit n1s (Quit: Lmnar)
18:39:50funmanhow are you avoiding the race condition?
18:40:19 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
18:41:03 Join CGL [0] (~CGL@
18:41:11ranmaBy reading the period value twice. Hmm, now that I think about it again I think if it is detected I should also read now again...
18:41:13 Quit rob (Quit: ( :: NoNameScript 4.22 :: ))
18:41:54funmancompiler won't optimize this out since the register is volatile
18:42:09ranmaIt's not a register on !C200V2
18:42:27funmanoops ok the TIMER2_ confused me
18:42:28ranmaIt's defined as (KERNEL_TIMER_FREQ/HZ) in that case
18:42:50 Quit robin0800 (Remote host closed the connection)
18:43:12 Join robin0800 [0] (
18:43:48funmancan the period change anytime while we are running the loop?
18:44:14ranmaIf pwm is enabled it changes on every timer underrun
18:44:48 Join rob [0] (
18:45:08funmani'd just disable interrupts, get the period value, and handle 1 possible period change
18:45:57ranmaIf you disable interrupts it can't change in between
18:46:39ranmaBut if we disable interrupts we don't really have to muck about with the timer at all
18:46:51ranmaAnd could just use a delay loop instead :)
18:47:29funmani mean disable it while reading the period, enable it afterwards, and handle at most one timer wrap (and at most one period change)
18:48:13funmanor modify the code so we can delay for more timer wraps (then no need to check if asked delay is too high)
18:48:52ranmaIf you disable it while reading the period you can still get a timer wrap that would screw up the calculation
18:48:54funmanbtw if you built the USB code with gcc-eabi, the delay loop had been optimized away so it's perhaps not needed
18:49:16funmanranma: it can't happen while you're reading the period?
18:50:20funmanwhat about watchdog, wasn't that simpler to use ?
18:50:42ranmaThe problem with watchdog is that the frequency is not very nice.
18:50:51ranmaIt's fixed at PCLK/256
18:51:08ranma=> 242187Hz
18:51:30funmanhow is that not nice?
18:52:08ranmaSo converting between usecs and watchdog ticks can not be easily done with shifts.
18:52:44funmanwe could switch the PLL to 384MHz
18:53:08ranmaWhat's PCLK then?
18:53:22funmanand fclk 240
18:53:50funmanbut we need test_codec + battery_bench results
18:54:10ranmaMaybe use watchdog for timer-as3525.c?
18:54:23ranmaAnd then use TIMER2 only for udelay
18:55:24ranmaSo for now fixing udelay seemed easiest for me :)
18:55:27funmanwe could just have a simple busy loop depending on fclk setting (we only have 2 settings)
18:56:36funman4 cycles per loop => (FREQ/4) loops per sec => FREQ/1000000/4 loops per µsec
18:56:38ranmaThat would probably be a bigger function though (if you want it to be inlined)
18:56:54ranmaI don't really see the point in forcing udelay to be an inline though...
18:57:09funmantrue, and i think the function would be qutie small anyway
18:57:54funmanloops = ((cpu_frequency == CPUFREQ_MAX) ? 248 : 62) / 4; + 2 instructions for the loop
19:00:39ranmaBut you have to disable interrupts during the udelay.
19:00:55ranmaThat the timer based udelay doesn't have to disable interrupts is kind of a nice property
19:00:57funmanwe can wait a tad bit longer, not a problem
19:01:23ranmaThe problem comes when someone changes the cpu frequency in between :)
19:01:32ranmaHmm, or just check dthe cpu freq in each inner loop
19:01:41funmantrue but we won't give back thread control
19:01:43ranmaThat could work quite well
19:01:57funmanno because it could change after you have read it
19:02:26ranmaBut it doesn't change that often, so the error would be small
19:03:04funmanwe want a minimal delay
19:03:59funmanonly threads change the cpu freq, not code that runs in interrupts so it's ok
19:04:35 Join liar [0] (
19:04:36ranmaBut udelay could be called from a thread?
19:04:54funmanyes, and since it blocks other threads cpufreq won't change
19:05:13ranmaHow does it block other threads if it doesn't disable interrupts?
19:05:35funmanblocking == not explicitely giving control back with yield/sleep/functions with timeout
19:06:02ranmaOh, I thought we are using preemptive multitasking using the timer interrupt :)
19:06:05funman(thanks to cooperative threading)
19:06:05 Quit liar (Client Quit)
19:06:51funmanyou want to test this function? i can do it later
19:07:07ranmaGreat, so let's just switch to a simple udelay like that one :)
19:07:15funmannow i'm trying to change bootloader memory map
19:07:32funmansorry if you had more complex things in mind ;)
19:07:42ranmaI'll change it so it doesn't need the multiply and test it
19:08:18funmansubs %0, %0, %1 ?
19:09:25funmanwe need (cycles_per_usec+3)/4 also, to make suer we don't wait too little time
19:09:47funmanand CPUFREQ... + 999999 ?
19:12:37 Join Buschel [0] (
19:14:09 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
19:15:58 Join JohannesSM64 [0] (
19:21:00 Quit pixelma (Quit: .)
19:21:22 Join pixelma [0] (quassel@rockbox/staff/pixelma)
19:21:35bertrikfunman, I think I found out how to configure the PLL on the AMSv2 sansas
19:22:07bertrikwe can set the PLLA to 248 MHz, just like the AMSv1 sansas, for example
19:24:54 Join bieber [0] (
19:31:38 Join Xerion [0] (
19:32:31 Join Jaykay [0] (
19:33:56CIA-85New commit by funman (r26926): as3525*: enable MMU in bootloader ...
19:34:59funmannow bootloader with USB links fine
19:35:25CIA-85r26926 build result: All green
19:36:47 Quit DerPapst1 (Quit: Leaving.)
19:36:52 Join DerPapst [0] (
19:37:37 Quit pamaury (Remote host closed the connection)
19:39:24CIA-85New commit by funman (r26927): as3525*: enable USB stack in bootloader (but not USE_ROCKBOX_USB yet)
19:40:50 Quit DerPapst (Ping timeout: 240 seconds)
19:41:27CIA-85r26927 build result: 7252 errors, 1155 warnings (funman committed)
19:42:47topikjust a couple of errors
19:43:02 Join fml [0] (
19:43:26fmlHello. Have the manuals been built today?
19:47:08CIA-85New commit by funman (r26928): fix endif wrongly removed (should have been moved) in r26927
19:48:10 Quit linuxguy3 (Ping timeout: 245 seconds)
19:48:36CIA-85r26928 build result: All green
19:48:58 Quit S_a_i_n_t_ ()
19:49:13funmanranma: i don't see µSD drive on fuzev1
19:50:04*ranma hasn't tried the slot on his C200v2 yet
19:50:09 Join linuxguy3 [0] (~timj@
19:50:56 Join S_a_i_n_t [0] (S_a_i_n_t@
19:52:18funmanalso lsusb -v gives some errors (if that helps)
19:53:21funmanif i do that, rb stays stuck on USB screen after unplug
19:54:24 Quit Buschel (Ping timeout: 252 seconds)
19:54:50funman% sg_luns /dev/sg6
19:54:50funmanLun list length = 16 which imples 2 lun entries
19:55:24funmani guess that means the sd slot is around
19:55:52funman1 lun entry on clipv1
19:56:52 Quit powell14ski_ (Read error: Connection reset by peer)
19:57:23 Join powell14ski [0] (
19:57:42funmanabout USB bootloader mode, since it's intended as a recovery tool, perhaps we could enable it only on a certain keypress
20:02:24 Quit powell14ski (Ping timeout: 276 seconds)
20:03:10 Nick fxb is now known as fxb__ (
20:03:29funmanbertrik: what gives most accuracy: 248 or 384 ?
20:04:39bertrik384 MHz gives most accuracy
20:06:01***Saving seen data "./dancer.seen"
20:06:53 Quit fml (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
20:08:01funmanif we use 384MHz PLL we could use it also for USB
20:09:12funmanwe would need to max out the CPU at 240MHz, not a big difference, and perhaps the faster pclk on AMSv1 would lower this difference
20:11:15bertrikChanging to 248 MHz is a small change, just change from 4 MHz * 60 to 4 MHz * 62. For 384 MHz we may have to experiment a bit to make sure we don't exceed limits of the internal PLL signals (Fvco and Ffbk)
20:11:27ranmaTo not have to change voltage for cpu frequency switching 200MHz would better IIRC?
20:12:00funmanranma: this doesn't work on devices with sd slot and gives very limited battery savings (last time i tested on clipv1)
20:12:34funmanbertrik: we can make 2 changes: 1/ switch AMSv2 to 248MHz so they use the same code than AMSv1
20:12:58funman2/ see if using 384MHz PLL has no drawback on AMSv1, and see if we can use a 384MHz setting on AMSv2
20:15:05bertrikfunman, we can generate the 240 MHz CPU clock from the 384 MHz? (using the pre-divider for example)
20:16:14funmanyes, 384*5/8 == 240
20:17:08bertrikBy the way, I realised we can just use the 2nd PLL for USB only, because the extra 2.5 mW power consumption is not important when we're connected to a PC anyway (we're probably charging then)
20:18:17 Join kugel [0] (~kugel@rockbox/developer/kugel)
20:18:40bertriknice idea
20:20:45funmanyes that's what the code does now, so using 384MHz pll would just give more accurate playback freq and rounder frequencies
20:20:49 Quit stoffel (Remote host closed the connection)
20:21:10funman(the latter helps with the planned new udelay() )
20:22:13 Quit t0rc (Remote host closed the connection)
20:24:11ranmaNot tested yet, but this one doesn't need the mul, but may be not that accurate :/
20:25:40funmanwhy not using mul?
20:25:51bertrikWhat do we need super-accurate udelay for?
20:26:04funmaniiuc it's not as terribly slow as division
20:26:18funmanbertrik: not super-accurate, just guaranteed minimal delay
20:27:04funman-> to avoid surprises when changing compiler/whatever like it happened on the D2 (at least)
20:30:03 Join petur [0] (~petur@rockbox/developer/petur)
20:31:16CIA-85New commit by funman (r26929): arm/crt0.S: comment why the reset vector doesn't use absolute addressing
20:31:57 Join binaryhermit [0] (
20:31:58ranmaOk, so just this simple one then:
20:32:32funmanranma: i agree with what you said earlier: no need to inline it
20:32:39CIA-85New commit by bertrik (r26930): as3525v2: document PLL bits and show current PLL frequency in the debug menu
20:32:45CIA-85r26929 build result: 207 errors, 62 warnings (funman committed)
20:33:11funmanah it's not me :)
20:33:18ranmaInterestingly the +3 is optimized away completely :)
20:33:34ranmafunman: Put it in system-as3525?
20:33:36*kugel thinks udelay should be as accurate as possible, so inline if doable
20:33:46funmanranma: if usecs is always a multiple of 2, yes
20:34:10CIA-85r26930 build result: 207 errors, 62 warnings (bertrik committed)
20:34:32bertrikkugel, that depends on whether you want some *guaranteed minimal* delay or super-accurate
20:34:44ranmaBut udelay can be used in quite a few places, so forcing it inline bloats the code
20:35:12ranmaOne could probably add a udelay_inline() if that's needed anywhere
20:35:30bertrikthat would just call udelay? :P
20:35:34kugelbertrik: both ideally :)
20:35:49funmankugel: well with this change we don't take in account time spent in interrupts
20:35:57funmanbut it should be quite small anyway
20:36:03kugelranma: the current udelay is tiny on !C200V2 because of constant parameter optimization
20:36:06 Join DerPapst [0] (
20:36:16kugelI checked disassembly
20:37:04bertrikThe delay we currently use in firmware/target/arm/as3525/fmradio-i2c-as3525.c is probably way off
20:37:06kugel5-6 insn basically
20:37:22ranmakugel: Well, as I said, if we would change timer-as3525.c to use the watchdog we could use TIMER2 exclusively for udelay :)
20:37:49ranmaThis udelay is 11 instructions
20:38:04kugelor the other way around :)
20:38:14kugelboth have 1.5MHz so it doesn't really matter does it?
20:38:49ranmakugel: No, watchdog is not 1.5MHz
20:38:59ranmaIt's PCLK/256
20:39:09kugeland PCLK is?
20:39:16funmanranma: btw did you check if usb_delay() was removed when built with eabi ?
20:39:24funmansdram_delay() wasn't so i removed it completely
20:39:33funman(forgot to mention that in the svn log)
20:39:33ranmafunman: No, did not check.
20:39:50funmani mean it *was* removed (it wasn't present)
20:40:21kugelranma: but does watchdog really work for timer? timer restarts automatically independantly of how long the tick tasks run (so the timer doesn't get off by the time the ticks need to run)
20:40:34ranmakugel: Yes, it has automatic reload
20:41:08ranmaI poked at it using JTAG and verified that :)
20:41:47funmanbertrik: where does the 12MHz freq comes from ?
20:41:59kugelyou could also use watchdog for backlight on the c200v2 and the current udelay could stay as is, couldn't you?
20:42:05funmanwhen clk_sel == 0, the source is 24MHz like on AMSv1
20:42:13kugelit works well where TIMER_FREQ is a constant
20:42:19bertrikfunman, I don't know
20:42:33funmandoes the other 240MHz setting match with these bits?
20:43:13funmanhum i see a 384MHz setting in clip+ OF
20:43:21ranmaNot as easy, the watchdog timer has no bgload register, could possibly work well enough anyway
20:43:41funmanalso 232 and 234MHz (table is at 0x8CC0)
20:43:54 Quit JohannesSM64 (Quit: WeeChat 0.3.3-dev)
20:44:13funmankugel: what's wrong with using a busyloop for udelay?
20:45:01kugelwhat's wrong with the current one?
20:45:16funmandoesn't work on c200v2
20:45:22funmanand it's long & complex
20:46:15funmanbertrik: do you need me to test on fuzev2/clipv2 before committing the change to 248MHz ?
20:46:26kugelfunman: I disagree
20:46:50kugelit works on c200v2 if you call it in a loop with smaller delays, doesn't it?
20:46:59bertrik232 MHz would result in exactly the sample rate error that the OF makes that dfkt posted an image of
20:47:16 Quit flydutch (Quit: /* empty */)
20:47:41ranmafunman: With my small changes it should work reliably :)
20:47:56kugelfunman: it's optimized and inlined very few instructions
20:48:04funmanranma: which one?
20:48:18kugeland it could be cleaned up if the c200v2 can get a fixed TIMER_FREQ
20:48:20funmankugel: perhaps it's small in binary but the source is hard to read
20:48:40kugelfunman: you re-wrote it a bit, I found it very readble in my first version
20:49:08funmani fixed a bug
20:49:25bertrikfunman, yes please. Shall we change it to 248 MHz now, and consider changing/trying the PLL to 384 MHz in the coming weeks?
20:49:27funmanranma: well without locking it can't work..
20:49:32funmanbertrik: sure
20:50:07ranmaWhy do you think it can't work?
20:50:18funmanranma: also what makes me feel bad is that we need a full page of comments to explain the code
20:50:36funman"avoid race condition" <- how exactly is it avoiding a race condition if you don't lock?
20:50:38ranmaThe double-read of TIMER_PERIOD should catch the wrap-around race
20:50:49funmanunless it wraps again after you have read it
20:51:47ranmaThat shouldn't happen unless you get a (non-timer) interrupt between the first read and the correction
20:51:53funmanudelay() is a 'somewhat inaccurate delay', writing it in 10 lines is better than in 60 lines
20:52:08ranmaAnd then that would probably add so much additional delay that it doesn't matter
20:52:22funmanbetter write a 10 lines function which needs no special case and not 30 lines of comments
20:53:14funmanwe just need something which will gives at minimum the same delay, not depending on current cpufreq or compiler used
20:53:41funmanif simple busy loop achieves that, why not going the simple way?
20:54:56ranmaI agree that a simple busy loop is probably quite sufficient
20:55:48funmanalso we can limit parameter to unsigned short and use 16*16 MUL ?
20:57:00ranmaThe disassembly I see here is doing a simple "mul r3, r1, r2"
20:57:25ranmaFollowed by "mov r3, r3, lsr#2" for the /4
20:57:26funmanthat works but if we use longer than 1<<16 argument the compiler would warn
20:58:14funmani'll test on the devices i have
21:01:39simonrvnhow do you enter a newline in the text editor?
21:04:08CIA-85New commit by bluebrother (r26931): Make System and Utils class based on QObject. ...
21:04:14CIA-85New commit by bluebrother (r26932): Rockbox Utility translation updates. ...
21:05:38CIA-85r26931 build result: 207 errors, 62 warnings (bluebrother committed)
21:07:11CIA-85r26932 build result: All green
21:07:40mc2739funman: Can you explain why the r26118 changes were necessary? They are preventing charging from being enabled when booting with USB connected.
21:08:18 Part watto
21:08:33funmanmc2739: it prevents entering USB mode if usbstack is enabled and used
21:08:50bluebroth3rhmm, jakorasia-cg seems to announce sdl but misses at least SDL.h
21:08:50funmanwell, before the change
21:11:15funmanmc2739: does charging work with these changes removed?
21:12:36kugelfunman: busy loops are the things that are usually cpu freq and compiler dependant
21:13:57ranmaWell, the loop itself is inlince assembler and it's taking cpu freq into account
21:14:11CIA-85New commit by funman (r26933): as3525*: make udelay() be a simple busy loop ...
21:14:35ranmaGiven that the multitasking is cooperative the freq shouldn't change while the delay loop is running
21:14:41 Join piggz [0] (~piggz@
21:15:33funmanmc2739: works?
21:15:52CIA-85r26933 build result: 207 errors, 62 warnings (funman committed)
21:16:42funmanmc2739: IIUC the problem is that when we boot with USB/charger inserted, there is no edge detection so we need to get the status at init
21:16:50CIA-85New commit by bluebrother (r26934): block jakorasia-cg build client
21:18:22funmanmc2739: btw about your mkamsboot patch for docking, if we harass ranma often enough he will make the USB driver full-featured and we will be able to remove USB check from mkamsboot ;)
21:18:35piggzdoes rockbox support any features of an ipod dock? i have a sony dab radio with ipod dock....sound works ok, but no controls
21:20:23bluebroth3rpiggz: Rockbox has basic accessory support, but the results differ between accessories.
21:21:28 Join 50UAAFOUW [0] (
21:21:48bluebroth3rIIRC there's a wiki page listing accessories that have been tested.
21:22:18bluebroth3rah, here:
21:24:24funmanbertrik: got a diff for me to test, or shall i make it myself?
21:25:02piggzat the moment, when i plug into my usb port, its not switching to disk was yesterday, any ideas?
21:25:43bertrikfunman, in clock-target.h: -#define AS3525_PLLA_SETTING 0x113B
21:25:43bertrik+#define AS3525_PLLA_SETTING 0x113D
21:26:08bertrikwait, that's not all sorry, I'll make a patch
21:26:42bluebroth3rpiggz: try holding Menu while plugging it in. That will make it not change into USB but only charge.
21:27:13bluebroth3rthough I don't understand why it changes into USB mode −− this would mean that the dock acts as USB host.
21:27:31piggzbluebroth3r: im talking about my laptop now :)
21:29:04bluebroth3rpiggz: err ... what do you expect when connecting the Ipod to a PC?
21:29:31piggzbluebroth3r: what is supposed to happen?
21:29:55funmanbertrik: hm that doesn't give a round pclk for fuzev2
21:30:19bluebroth3rpiggz: well, I would expect it to change into USB connection mode :)
21:30:33bluebroth3raah, misread. It's _not_ going into usb mode :o
21:30:59piggzbluebroth3r: yes :) ....... well, actually, it is when i plug it into a different port
21:31:08bluebroth3rthe usual things to test: try a different usb port and cable.
21:31:25bluebroth3respecially front usb ports have been reported to be flaky at times.
21:31:34piggzso, i one port, it goes into disk mode, the other says something about usb presentation multimedia
21:31:35mc2739funman: Sorry, had to step away for a bit. Yes, charging works with r26118 reverted. It also works with
21:31:59bluebroth3rhmm. You will always get USB HID mode unless you turned that off.
21:32:02funmanmc2739: ok i will commit this diff (with the ';' added :p )
21:32:41bluebroth3rhowever, if the connection is flaky I could imagine that the PC gets a bit confused by the multiple functionality. Turn HID mode of if you don't use it (IMO it should default to off anyway)
21:33:37AlexPOn the defaults note, is there a reason why dircache isn't on by default?
21:34:05bluebroth3rAlexP: afaik there are some bugs with dircache. Though I can't remember observing them in the last years.
21:34:24AlexPI've not seen any personally, but that doesn't mean much :)
21:34:28bertrikpamaury fixed a couple of bugs with dircache IIRC
21:34:30mc2739funman: is USB detection still working with that patch?
21:34:34bluebroth3rpiggz: HID mode is known to cause problems on OS X. In case you're using a Mac :)
21:34:35AlexPAnd pamaury has been working on it not too long ago
21:34:41funmanmc2739: i didn't test but it should
21:34:43bluebroth3rAlexP: indeed :)
21:35:02AlexPI'll bug pamaury next time he drops by
21:35:26bluebroth3rbtw, what do people think about making ipodpatcher / sansapatcher and company use a version.h file instead of using -DVERSION=...?
21:35:54piggzbluebroth3r: no, linux
21:36:45AlexPbluebroth3r: I could even send a mail asking, inspired by the recent mail exchanges :)
21:37:36funmanbertrik: <- it will give 41.333333... MHz pclk on fuzev2
21:37:42bluebroth3rAlexP: sounds reasonable :)
21:38:18bluebroth3roh, and another thing I'm considering: how about moving rbutil to the utils folder? It's a bit strange to have this separation.
21:38:59bertrikfunman, what's the usual PCLK for as3525v2 targets?
21:39:20piggzso what is the point of hid mode? surely, when you plug it in, you want to be able to copy media to it?
21:39:25funmanbertrik: ~line 89 of clock-target.h
21:39:27pixelmaAlexP: if I recall dircache chokes on too many small files (main example often mentioned was the HVSC)
21:39:29ranmafunman: Why limit udelay to unsigned short?
21:39:49funmanranma: multiplication result fits in 32 bits
21:39:58bertrikoh, just 21 MHz!?
21:40:05funman24MHz on clipv2/clip+
21:40:21ranmaWith cycles_per_usec == 248 for 248MHz it should work very well for more than that
21:40:35AlexPpixelma: OK, I'll try that, cheers. I've just sent my mail, so maybe you could respond there too if you get a min
21:40:39funmantrue but that way no need to check for overflow
21:41:11ranmaNot quite true, it doesn't warn on usb_delay(100), does it?
21:41:32ranma1000*100 gets truncated to 34464...
21:41:37funmanhm i didn't see this
21:41:50pixelmaAlexP: it's also quite RAM hungry and would need a reboot anyway
21:42:27AlexPMaybe only on as default on HD targets
21:42:34pixelmaor not? Not sure as I don't use it... but the former is the reason why it's not even there on lowmem targets
21:42:39funmanon my test it doesn't get truncated but it doesn't warn either
21:42:41AlexPIt increases performance really quite a lot there
21:43:05funmanit warns only when assigned to a short
21:43:12ranmaWith unsigned it should work for delays up to 17318416 (17 seconds)
21:43:31ranmaSo I don't think a check for overflow is needed there
21:43:45funmanthen just remove the short and perhaps add a comment to give the higher limit
21:44:19funmannot that we'd want to sleep for 17 seconds anyway :)
21:44:33bluebroth3rpiggz: normally the Ipod should do HID and UMS at the same time.
21:44:53bertrikfunman, is it a problem that PCLK is non-integer for fuzev2?
21:45:09bertrikAlso PCLK on as3525v2 seems quite slow compared to as3525
21:45:14funmanno i think it's ok, it justs shifts udelay a bit
21:45:42funmanbertrik: slower = less power needed = better ?
21:46:47CIA-85New commit by bluebrother (r26935): Log filesystem free value to system trace.
21:47:39piggzbluebroth3r: ok..that seems to not be happening here....on one port...the ipod seems to reboot into some standard disk mode, with flashing stop the other port it stays in rockbox with a usb icon, but doesnt show as a disk
21:47:51bertrikI thought you set some DRAM clock stop bit some time ago, so maybe faster is no longer more power consuming
21:48:00CIA-85New commit by funman (r26936): as3525: cache first read of enrd0 register ...
21:48:16funmanit was for DRAM but pclk keeps running no?
21:48:16 Join fdinel [0] (
21:48:26CIA-85r26935 build result: All green
21:48:28bertrikyes, I think so
21:48:48bluebroth3rpiggz: that is strange. Rockbox does reboot into the Apple firmware for disk access, but this is only true for releases. Current builds of Rockbox use the Rockbox usb stack.
21:48:53bertrikI'd expect that the CPU wastes cycles waiting for the slow DRAM, but maybe only when boosted
21:49:01bluebroth3rI'd try disabling HID.
21:49:06 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
21:49:19piggzbluebroth3r: oh, im using the 3.6 release
21:49:43bertrikfunman, is this easy to change? I could do a battery bench with a different DRAM clock.
21:49:47bluebroth3rah, ok. Then there's something wrong with the connection detection on this one port.
21:49:54funmanbertrik: sure, clock-target is here for that
21:50:03gevaertsAlexP: IIRC the main dircache issue is that it fails non-gracefully if you have too many files and/or directories. I don't know the details
21:50:10bertrikok, so just change AS3525_DRAM_FREQ, right?
21:50:10bluebroth3rnevertheless, try disabling HID :)
21:50:28piggzok...with hid disabled, what is the expected bahaviour?
21:50:29CIA-85r26936 build result: 207 errors, 62 warnings (funman committed)
21:50:47bluebroth3rbut this pretty much sounds like that USB port on the PC is acting up a bit.
21:50:54bluebroth3rpiggz: reboot into USB mode.
21:51:01piggzk, ta
21:51:16ranmafunman: Did you test the udelay? Seems off (too slow) here...
21:51:25funmani didn't time it
21:51:32piggzwould you reccomend i install the latest build?
21:51:33*bluebroth3r remembers his Windows box telling about overcurrent on USB the other day but no USB devices were attached to the machine
21:52:00mc2739I thought jakorasia-cg had been disabled?
21:52:01 Quit jgarvey (Quit: Leaving)
21:52:04AlexPgevaerts: I've just downloaded the HVSC to try - it must be a chiptune thing as I have a 120 GB drive full of mp3/ogg and that isn't enough files to get it to die
21:52:17AlexPwell, not chiptune specifically of course
21:52:41AlexPBut only if hou have a huge numbers of very small files as chiptune collections tend to be
21:52:44ranmaDid a for (i=0; i<10; i++) { backlight_on(); udelay(500*1000); backlight_off() udelay(500*1000); } at the end of system_init()
21:52:47funmanranma: it *will* be slower, what is the difference you are measuring?
21:52:51ranmaAnd it seems to take about 40 seconds
21:52:53bluebroth3rit depends if you want a "more stable" build. Releases are generally considered more stable, but I'm using current builds all the time. However, there was a change in the theme syntax since 3.6, so your old themes won't work anymore (in case you use themes that aren't shipped with the default Rockbox install)
21:52:58pixelmaAlexP: well it's called "High voltage *sid* collection" for a reason ;)
21:53:07AlexPindeed so :)
21:53:29funmanranma: cpu_frequency might be uninitialized
21:54:05funmanbertrik: works on all AMSv2, should I commit it ?
21:54:33ranmaYeah, but if it's != CPUFREQ_MAX it should use the lower cycles_per_usec value, which shouldn't take longer than expected
21:54:33saratogaFWIW higher pclk means faster codecs and less boosting
21:55:03bertrikfunman, yes, let's do it! :)
21:55:25bluebroth3rgevaerts: did I something wrong when trying to block jakorasia-cg? It got an SDL build again
21:56:23 Join einhirn [0] (
21:56:31CIA-85New commit by funman (r26937): as3525v2: use 248MHz PLL (reverse engineered by bertrik) ...
21:57:35ranmaOk, never mind, it was initialized to _MAX, but actually running at _NORMAL
21:57:52funmani think it's only set correctly after the first cpu_boost()
21:57:59CIA-85r26937 build result: All green
21:58:16ranmaAdding a set_cpu_frequency(CPUFREQ_NORMAL) before the test fixed it.
21:58:21saratogacan these PLL settings be used on amsv1? having unboosted = pclk = 41MHz would be nice for battery life on those players
21:58:37funmansaratoga: sure, we just need to make sure display/wheel is fast enough
21:59:12saratoga40mhz is about what a lot of the codecs use, so I think it should be fast enough for display/ui
21:59:14funmansaratoga: ok for reverting r26396 ? (clip+ button change) : it would make quickscreen the same on all the clips and make some people happy (myself i only care about common keymap across clips)
21:59:25saratogafunman: yeah go for it
21:59:50saratogawhen i get a chance i'm going to software hold to hotkeys and then see if i can get them on the clip+
21:59:56 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
22:00:15saratoga'going to add software hold to hotkeys'
22:02:00pixelmathat sounds weird
22:02:53bertriksaratoga, the display runs at 1/16 of PCLK, so PCLK at 41 MHz would limit framerate to 66 fps on the fuzes (could be enough, but I think kugel likes it faster)
22:03:36kugelbertrik: I think it's not too slow, but the code could be smarter to make better use of it
22:03:41CIA-85New commit by funman (r26938): clip+ keymap: revert r26396 ...
22:04:04funmanwe need another DMA channel so we can update the screen with DMA !
22:04:16funmandfkt: ^ now you 'owe' us a keymap patch ;)
22:04:19CIA-85New commit by ranma (r26939): With a max delay of 17 seconds before overflow happens limiting to unsigned short shouldn't be necessary.
22:04:23saratogabertrik: thats only when unboosted right?
22:04:28bertrikhm, I think DBOP does support DMA indeed
22:04:59bertriksaratoga, no, I thought PCLK was basically fixed and we only increase FCLK when boosting
22:05:09CIA-85r26938 build result: 204 errors, 62 warnings (funman committed)
22:05:30funmanis he banned only of odd revisions builds?
22:05:45saratogaideally we should always have pclk as high as it can go for a given fclk, so that would mean changing pclk too
22:06:00saratogasince boosting the CPU but leaving the ram underclocked is not the best idea
22:06:03***Saving seen data "./dancer.seen"
22:06:38funmanpclk is 65MHz max, and ram freq 90MHz max but we could never use ram freq higher than 65 (well, 64)
22:06:45CIA-85r26939 build result: All green
22:07:40bluebroth3rdoes the build system allow banning based on revision numbers? :o
22:07:57funmanthere is a lot of code we should battery benchmark, the code in AMSv1 sd_enable only had very minor impact
22:08:07funmanbluebroth3r: perl can do anything ;)
22:08:43bluebroth3rmagic! ;-)
22:08:53funmanlooks like it's trying to draw a curve on our build table
22:09:35funmani think i never tested the dma_retain() / dma_release() also
22:11:20 Join DataGhost [0] (
22:11:20 Quit DataGhost (Changing host)
22:11:20 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
22:13:39 Quit funman (Quit: ++)
22:15:20dfkt<funman> dfkt: ^ now you 'owe' us a keymap patch ;) <−− hug kiss kiss hug!!!!!
22:15:35dfktand didn't i deliver a decent keymap already? ;)
22:24:43gevaertsbluebroth3r: I don't see anything wrong. Maybe Zagor or Bagder can check if the build server uses the correct block file
22:28:22piggzbluebroth3r: i installed the latest build...with hid mode on, when i plug it in, it switch to 'usb keypad mode', but the kernel says hub 2-0:1.0: unable to enumerate USB device on port 3
22:28:42 Quit Jaykay (Quit: ChatZilla 0.9.86 [Firefox 3.6.3/20100401080539])
22:30:20piggzbluebroth3r: i take it eventuall showed up
22:32:41 Join Buschel [0] (
22:34:18bluebroth3rpiggz: nice to hear
22:36:29bluebroth3rgevaerts: ok, thanks for checking
22:40:19 Quit Buschel (Ping timeout: 272 seconds)
22:48:52 Join dfkt_ [0] (~dfkt@unaffiliated/dfkt)
22:49:31 Quit dfkt (Disconnected by services)
22:49:36 Nick dfkt_ is now known as dfkt (~dfkt@unaffiliated/dfkt)
22:51:25 Quit MagusG (Ping timeout: 276 seconds)
23:05:04 Quit petur (Quit: Zzzzz)
23:10:02CIA-85New commit by bieber (r26940): Theme Editor: Working on rendering viewports, display will now show %V(...) viewports as red rectangles over backdrop or background color
23:11:33CIA-85r26940 build result: All green
23:12:48 Join anewuser [0] (anewuser@unaffiliated/anewuser)
23:19:09 Quit evilnick_B (Quit: Page closed)
23:23:07 Join liar [0] (
23:24:54 Join hamish_ [0] (~hamish@
23:27:29 Join kramer3d_ [0] (~kramer@unaffiliated/kramer3d)
23:27:35 Quit hamish_ (Client Quit)
23:35:07 Join dfkt_ [0] (dfkt@unaffiliated/dfkt)
23:35:34 Join funman [0] (~fun@rockbox/developer/funman)
23:36:48funmandfkt: did you make a patch for keymap-clip.c or only described a better keymap?
23:37:29 Quit dfkt (Disconnected by services)
23:37:35 Nick dfkt_ is now known as dfkt (dfkt@unaffiliated/dfkt)
23:37:55 Quit komputes (Quit: I haven't slept for ten days, because that would be too long.)
23:40:47funmandfkt: ^
23:41:56dfktfunman, i think you saw my response in the logs? :)
23:42:18dfktand no, i doin't know how to code C, i didn't make a keymap, i only described it to you
23:43:21 Join stripwax [0] (
23:44:28 Quit piggz (Remote host closed the connection)
23:44:45funmanok just to be sure, i'll look at it again to make the patch
23:46:17funmani found for recording but iirc you had devised something for FM too
23:46:38funman'move menu from down to center' ok
23:46:52dfktthe only thing that was distracting in the FM keymap was that it used the down key for menu function instead of the center button
23:47:14funmanbtw apps/keymap/keymap-clip.c requires no C knowledge (IMO)
23:47:55dfktmaybe i could cobble something together, but don't take my word for it ;)
23:48:06funmancenter is already used for 'FM_PRESET'
23:48:16dfktcome to me when it's about usability, or measurements, but not coding :)
23:48:29amiconnjhMikeS: Not yet
23:48:37dfkti didn't perceive any function on the center button? lemme check
23:48:44funmancan't see any either :/
23:48:58amiconnThe strange lines probably have to do with signed vs. unsigned shifts or similar. Need to compare disassemblies
23:49:26dfktno function in either preset or scan mode
23:49:43bertrikthe thing that sticks out most for me on the clip fm screen is that you need POWER to exit
23:50:11funmanisn't POWER the exit function to pretty much everything on clip?
23:50:18funmanplugins, playback, etc
23:51:35dfktthe home button is unused in FM mode too, one could use it for exit as well
23:52:07dfkti'm not sure how the consensus on redundancy of buttons is in rockbox
23:52:34bertrikwe usually have more functionality than we have buttons :)
23:52:43funmani don't know, but the consensus on consensus is you should avoid talking about them ;)
23:53:56funmanFM menu uses context settings so SELECT is (also?) ACTION_SETTINGS_RESET iiuc
23:55:44 Quit stripwax (Quit:
23:55:54amiconnWhat annoys me about the clip keymap is that there doesn't seem to be a button to resume/ return to the wps
23:56:34dfktamiconn, that's what i think too
23:57:05dfktreturn to wps could be (once again) mapped to the power button, no matter where in the menu one is
23:57:16bertrikif you are in the WPS and press HOME, you go to the menu, press HOME again and you're in the WPS again
23:57:20 Part toffe82
23:57:21dfktor the home (sic) key)
23:57:26 Join stripwax [0] (
23:58:02dfktbertrik, when you go to the file/folder list, there isn't a fast way to return to wps, afaik?
23:58:34bertrikdfkt, indeed there isn't, HOME just seems to go one screen back

Previous day | Next day