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 2009-08-17

00:00:44CIA-61New commit by jdgordon (r22364): remove the rockbox info line saying the AA size and replace it with skin RAM usage
00:01:18pixelmaI don't know how to use bloat-o-meter, don't know if I still need to set something up to make it work here, don't know what output I'll get and how to deal with it
00:01:58JdGordonhaha damn... the front page has "1 more file" instead of showing the actual .c file with the relevant changes :p
00:03:45kugelpixelma: "./bload-o-meter rockbox-old.elf rockbox-new.elf"
00:04:09pixelmacan you really just delete the complete line?
00:05:03pixelmakugel: even if I don't know if it would work at all, I doubt your line would, it's spelled bloat I think
00:06:09pixelmathat still doesn't tell me where to execute this from and what type of script support I need
00:06:16kugelrockbox.elf is created after compiling (after stripping it's rockbox.bin, after scramble it's rockbox.<target>)
00:06:33kugelboth tools are in utils/analysis, as I said
00:07:22kugelmoos: I think you referred to the wrong rev in r22362
00:08:20*pixelma is confused - where does it know which target I want to use it for?
00:09:06kugelpixelma: it doesn't know it, it just compares both files, telling you which symbols changed (a symbol is a function or a global variable, for example)
00:09:13kugel<kugel> pixelma: it doesn't know it, it just compares both files, telling you which symbols changed (a symbol is a function or a global variable, for example)
00:09:19kugeloops, damn xchat!
00:10:34kugelpixelma: compile rockbox with the rev before JdGordon's commit and with the rev after, each time copy the created rockbox.elf into utils/analysis (make sure you rename them, else the 2nd copy will override the first one), then run bload-o-meter passing both files as parameter
00:10:52JdGordonhehe same typo :)
00:10:55pixelmaok, reading helps a bit... what type of script is this? I'm asking to find out if I need to install some more components
00:10:59JdGordonhave you swapped the t and d keys?
00:11:24kugeluhm...yea :P
00:11:32 Quit bertrik ("De groeten")
00:11:41kugelbut I'm not looking at the keys anyway
00:12:25kugelpixelma: both are written in python, that's all you need (think the toolchain too, for *-elf-objdump)
00:14:29pixelmayes, got that
00:15:15pixelma(I mean that it's installed)
00:16:18bluebrotherok, played around with ccache a bit. Changing the output folder name breaks it, creating the output folder with the same name as used in the first build produces cache hits.
00:16:32 Quit petur (Remote closed the connection)
00:16:57bluebrotherso I'd say using build-<targetname> instead of build-<pid> should make it work as expected. The current state unfortunately breaks ccache.
00:17:14*JdGordon wonders if thats enough to drive someone to start our own ccache that actually works :)
00:17:40bluebrotherwell, ccache itself works. I assume it takes the output folder into account when caching.
00:17:47kugelJdGordon: going for customlist now
00:17:53JdGordonuh oh!
00:18:07kugelbluebrother: yea, we found that our yesterday
00:19:17pixelmamany changes in the same area and the blame game begins tomorrow if somehting is broken...
00:19:25 Quit evilnick_home (Read error: 113 (No route to host))
00:19:57kugelI touch a totally different area if you mean that
00:20:16CIA-61New commit by kugel (r22365): User definable UI viewport, to be able to restrict the UI into a viewport for all bitmap displays. ...
00:20:17 Quit bluebrother ("sleep")
00:20:36JdGordonsomeone didnt fix the manual! :)
00:20:53kugelgrml, I forgot to added that themes need "ui viewport: " to restore from a theme using one properly :/
00:21:58JdGordonskinned statusbar shouldnt be too far off now
00:23:03kugelw.r.t. to the manual, I was wondering were to add it. It doesn't have a settings entry (available via .cfg only)
00:23:22JdGordonI thought we were against those settings!
00:23:48 Join evilnick_home [0] (
00:23:51kugel12.1 seems good
00:23:58kugel"12.1. Customising the User Interface" :)
00:24:28pixelmait's not 12.1 everywhere
00:24:46kugelJdGordon: I actually had a setting for it, basically unusable
00:25:02pixelmae.g. in manuals for targets that don't have a radio or recording - the numbers are generated
00:25:18 Quit bmbl ("Bye!")
00:27:49 Join thegeek [0] (
00:28:04gevaertskugel: looks like r22365 is what you need on the fuze to work around this top-line-not-visible thing
00:28:28CIA-61New commit by kugel (r22366): Fix ret for buttonbar targets.
00:28:30 Quit thegeek (Read error: 131 (Connection reset by peer))
00:28:52*JdGordon slaps kugel around
00:28:57JdGordonthere goes my lovely greens
00:29:31kugelgevaerts: yea, that made me pushing it in! :P
00:29:42pixelmakugel: did you add something about the playlist viewer shortcut to the manual yet?
00:29:55kugelno, but I haven't forgotten
00:30:37kugelI don't have tex stuff on my laptop (and I really don't want to install it here), and my desktop was unfunctional until a week or so
00:40:43 Quit pamaury ("exit(*(int *)0 / 0);")
00:42:05 Join thegeek [0] (
00:42:38moosJdGordon: What is the *skin* RAM usage please? Does it count for all the theme or for the WPS only...
00:42:43 Quit domonoky (Read error: 54 (Connection reset by peer))
00:43:07JdGordonskin is the new name for the wps stuff...
00:43:13JdGordonfor now its the rwps and wps
00:43:47pixelma_buffer_start +2688 causes the biggest plus, a lot is reduced again by _wps_datas again but not all (then there are some minor + and - but negligable compared to those two)
00:44:22pixelmathis is with my baklight modded OndioFM build
00:45:20moosJdGordon: Thanks, I'm trying to update the french translation since I'm around
00:47:10pixelmawhat does that tell?
00:47:42JdGordonthat you're SOL...
00:48:37JdGordonnothing you can do about it... unless oyu manually shrink buffer_start (which is easy)
00:48:51JdGordonunless we allow that buffer to be user sizeable
00:49:01pixelmawhere does it come from?
00:49:29JdGordonskin_engine/skin_buffer.c at the very top
00:49:39 Join mcuelenaere_ [0] (
00:49:44 Quit mcuelenaere (Read error: 104 (Connection reset by peer))
00:49:47JdGordonactually.. can you paste the bloatodiff?
00:49:48pixelmaI also mean why is it different to before?
00:49:56JdGordonit shounldt be
00:50:17kugelsounds a bit strange indeed
00:50:29 Quit dfkt ("-= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.")
00:50:36 Quit ender` (" The idea that I can be presented with a problem, set out to logically solve it with the tools at hand, and wind up with a p")
00:51:30 Nick mcuelenaere_ is now known as mcuelenaere (n=mcuelena@rockbox/developer/mcuelenaere)
00:52:22kugelskin_redraw is huge function :o
00:52:49JdGordon300 lines
00:52:52 Quit robin0800 (Remote closed the connection)
00:53:30JdGordoncrappy pastebin.... still connecting
00:54:32pixelmaI could also paste it to somewhere else if you have a preference and a link
00:54:58JdGordonyeah, please
00:55:10JdGordonah neva mind
00:55:13JdGordonjust worked
00:56:27kugelJdGordon: it's probably because wps_data has new members now due to the (many skin_token_list), and maybe alignement, while the buffer is still the same
00:56:31kugelpixelma: ^
00:57:27JdGordoni dont see how that makes sense...
00:57:31kugelhm, on the other hand, they're all just pointer
00:57:47JdGordonI do want to do something about those though... not sure what
00:57:54kugelno, that can't be right
00:58:19kugelJdGordon: the problem is that it's difficult to see a diff because you didn't do the rename seperately
00:58:23 Join CeruleanC [0] (n=Cerulean@unaffiliated/ceruleanc)
00:59:50kugelpixelma: what target is that?
01:00:19pixelmaas I said, backlight modded OndioFM
01:01:15pixelmabut you can see the same effect in the build table for "stock" OndioFMs
01:03:15 Join BdN3504 [0] (n=4e3435ff@gateway/web/cgi-irc/
01:05:00BdN3504kugel: could you post the cfg file to your example in the thread on the forums? i am not getting any schlau out of "ui viewport: X, Y, [width], [height], [font], [foreground pattern], [background pattern]" esp. the last three params
01:05:16CIA-61New commit by moos (r22367): Update the french translation. ...
01:05:17BdN3504do we now have a custom font in the menu?
01:05:37kugelBdN3504: it's just like wps viewport, just with the comma as seperator (and without the | at the very end)
01:07:59BdN3504kugel: ok so font is either user defined "1" or system "0" what do the patterns stand for? are they colours in hex code?
01:08:01kugelexactly (if you're on a colour display)
01:08:32BdN3504ok, why do you call them patterns then? shall i write a manual entry for this?
01:10:08kugelpixelma: I can't explain it, wps_datas should be more than 2,7k smaller
01:10:53kugelyou may do that!
01:11:37BdN3504ok. bebacktomorrow.
01:13:53BdN3504wait, one last thing. have i understood correctly that this feature only works in the main menu as of now (i.e. not the submenus or the context menus) or does only not work with plugins which show the bg.bmp?
01:14:24pixelmakugel: might be helpful to find out why it's not, maybe there's a flaw somewhere
01:14:47kugeldefinitely, maybe someone else knows more
01:15:48***Saving seen data "./dancer.seen"
01:17:59 Quit BdN3504 ("CGI:IRC (EOF)")
01:18:39kugelJdGordon: are you going to break the other limits also?
01:19:02kugelviewports, and stuff.
01:21:58 Join sinthetek [0] (
01:24:23sinthetekhey, is there some sort of big compelling technical reason why rockbox doesn't support sony walkman daps (ie hardware drm mechanisms, etc) or is it simply a matter of finding someone who has one and is willing/able to port it?
01:27:20 Join HBK [0] (
01:27:28CIA-61New commit by kugel (r22368): Remove obsolete IMG_BUFSIZE #define.
01:28:04JdGordonkugel: remove... not break
01:28:08JdGordonand yes.. all of them hoperfully
01:28:12JdGordonviewporets is next i tinhk
01:29:29kugelI'd really like to know why wps_datas only shrinked by 1.7k
01:30:57JdGordonit makes sense...
01:31:19JdGordonumm.. maybe not
01:32:48kugelit doesn't, the skin buffer is ~2.7k (and so was the old image buffer)
01:33:03kugel(on the ondio)
01:33:49JdGordonthe other chunky structs
01:34:00JdGordonits still nearly 10K which suck
01:35:21 Quit HBK (Read error: 104 (Connection reset by peer))
01:35:38kugelwhat's about the other structs?
01:37:30JdGordonthats where it dropped
01:37:41kugelit doesn't change the fact that removing img_buf[] alone should've free'd 2.7k
01:39:14kugelgood thing the commit wasn't with buffer_alloc already, that revealed a massive ram waste
01:40:18kugelJdGordon: "that's where it dropped"?
01:40:34 Quit Thundercloud (Remote closed the connection)
01:41:37 Join HBK [0] (
01:42:38 Quit GeekShadow (Read error: 104 (Connection reset by peer))
01:46:22kugelpixelma: ah I found it
01:47:54pixelmakugel: yes?
01:48:01 Join ehntoo [0] (
01:48:03kugela wps_token struct is quite a bit bigger now, and there's an array of 1024 of them in wps_data
01:48:53JdGordon+4 bytes isnt *quite*!
01:49:12JdGordonmind yo.. that is 4k though
01:49:13gevaertsJdGordon: if there's an array of 1024 of this, it is
01:49:23kugelJdGordon: it was only 8 before, that's a 50% increase
01:49:55JdGordonoh wait.. no shouldnt it only be 2 bytes bigger?
01:50:02 Join hd [0] (
01:50:06 Quit hd (Read error: 104 (Connection reset by peer))
01:50:08pixelmaalso - reading the forums a bit, is your list viewport seperated from the statusbar? So is the statusbar still on top of the screen or inside you ui viewport area?
01:50:18 Quit HellDragon (Read error: 54 (Connection reset by peer))
01:50:23pixelmakugel: ^
01:50:33 Join HellDragon [0] (
01:50:44JdGordonits probably being bloated for alignment...
01:50:58 Join thegeek_ [0] (
01:51:00JdGordonit used to be 4, its now 6 but probably 8
01:51:11kugelJdGordon: 7bytes +1byte alignment before, 9+3 bytes alignmenet now IIUC
01:51:29JdGordonhow does it get to 9?
01:51:39kugela bool is 4 byte IIUC
01:51:51JdGordonisnt it a char?
01:53:25kugelI'm not entirely sure, and it may be platform specific, but I think it's 4
01:54:15JdGordonI would be very surprised if it is
01:54:17kugelpixelma: if you use the ui viewport, you need to make sure yourself that it doesn't bite with the statusbar
01:55:02pixelmaso statusbar is still at the same place not inside that viewport, that's what I wanted to know
01:56:08kugelno, the statusbar is seperate
01:56:16kugelas is the buttonbar
01:56:43JdGordonI thought the buttonbar was going to be forced inside the uservp?
01:56:57kugelwhy should it?
01:57:33kugelit needs to be aligned with the buttons below, doesn't it?
01:58:44JdGordonyes, but the argument was its easiesr and noone actually uses the bb
01:58:51JdGordonbut ok, if oyu got it working...
01:59:00kugelpixelma: that ram will be free'd if/when the 1024'ish array is "converted" to use the skin_buffer
02:00:12kugelJdGordon: the easiest is to let the user take care of positioning so that the button bar is visible
02:00:51 Quit daurn (Read error: 113 (No route to host))
02:01:10 Quit Rondom (Nick collision from services.)
02:01:21 Join Rondom [0] (
02:01:28kugelrasher: ping
02:01:38rasherkugel: yes?
02:01:55kugelI need admini access to the theme site if I want to update themes?
02:02:43kugeladmin access too
02:03:12rasherPretty much :\
02:03:20kugelupdate as in delete and re-upload, I guess
02:03:28rasherIt's not very ideal
02:03:57 Quit Lss__ (Read error: 104 (Connection reset by peer))
02:04:56pixelmakugel: I see, thanks for looking into this
02:05:19kugelno problem
02:06:07kugelit would've been more obvious if JdGordon svn mv'd beforehand (I can't stress this enough! :) )
02:06:56*kugel wonders if kkurbjun checked that the shortname really matches checkwps usage before adding the mr500
02:13:17 Join froggyman [0] (
02:15:05CIA-61New commit by kugel (r22369): Add the ui viewport to the theme settings, so that it will be in the theme.cfg created by "Save Theme Settings".
02:25:00JdGordonanyone feel like quickly testing another ram saving patch?
02:26:16kugelDo you mean another malloc patch? yea sure :p
02:26:52JdGordonmoving viewports into the new buffer
02:28:31 Join Strife89 [0] (
02:28:50CIA-61New commit by kugel (r22370): Change the default value for the ui viewport to "-" (which will give a fullscreen vp since parsing fails).
02:32:58 Quit ehntoo (Read error: 60 (Operation timed out))
02:33:39kugelJdGordon: doesn't work. freezes upon entering the wps
02:33:51JdGordonwhich target/wps?
02:33:51kugelbuttonlight doesn't turn off, so I guess it's a real freeze
02:33:55JdGordontheme i mean
02:34:08JdGordondoes it use viewports?
02:34:14kugelit has a conditional one, ye
02:35:07kugelno normal one. the conditional is whether album art is shown or not
02:35:55kugelJdGordon: another themes which has dozens of normal viewports seems to work fine
02:35:58JdGordonso only the default viewpoerr?
02:36:31kugelI guess so
02:37:31sinthetekdoes anyone know why sony walkman mp3 players are not at all supported?
02:37:32JdGordonok ta.. time to make dinner, this wont be commited tonight
02:37:46JdGordonsinthetek: noone has done anything to try to make it work
02:38:39sinthetekoki, just wondering if anyone had looked into it or if there was a solid reason why not
02:53:53kugelrasher: the author has to verify the theme? :(
02:54:15rasherkugel: Yes?
02:54:19kugelor his email rather?
02:54:31kugelgrml, I kept the other guy as author
02:55:18kugelupdating the screenshots at least without needing to do the whole delete 'n' re-upload thing would also be nice
03:02:45 Quit froggyman (Remote closed the connection)
03:11:48 Join dmb [0] (n=Dmb@unaffiliated/dmb)
03:12:59kkurbjunkugel, I assumed that it matched the modelname from the configure script
03:15:51***Saving seen data "./dancer.seen"
03:18:34 Quit mcuelenaere ()
03:21:12 Join aaron424 [0] (
03:22:31 Quit SUSaiyan` (Read error: 104 (Connection reset by peer))
03:22:53 Join SUSaiyan [0] (
03:23:34 Join SUSaiyan` [0] (
03:24:55 Quit SUSaiyan` (Read error: 104 (Connection reset by peer))
03:25:15 Join SUSaiyan` [0] (
03:25:43UnhelpfulJdGordon: it, um, would, i suppose, be possible? i had considered also multiple sizes for multi-screen WPS, one of the few beast OF bits i miss - you can go from playlist to huge AA to smaller AA and more track details, etc.
03:26:25 Quit SUSaiyan (Read error: 104 (Connection reset by peer))
03:28:10Unhelpfulwe'd need to add code to handle multiple covers in the buffer associated with one track, and obviously code to do something useful with that, but there's no reason we couldn't do it... you load an image for display just by doing a bufopen of an image file as TYPE_BITMAP
03:28:15CIA-61New commit by kkurbjun (r22371): M:Robe 500: Add Rockbox logo to small image
03:28:40Unhelpfulif we want to support multiple sizes, we'd need to add a way to pass a size argument. right now, it just looks at the WPS size
03:29:43 Quit Strife89 (Read error: 113 (No route to host))
03:30:43Unhelpfulas far as making the buffer more malloc-y, i was thinking that perhaps a somewhat-heavy compaction mechanism could be OK if 1) it's cheap when compaction is not needed 2) compaction is rare
03:31:46 Quit aaron424 ("ChatZilla 0.9.85 [Firefox 3.0.13/2009080315]")
03:31:55Unhelpfulthe easiest way would be to compact only when we can't otherwise find enough buffer, or when something that *must* move the buffer start happens - we could use the same method to move buffer start during playback
03:35:20 Quit HBK (Read error: 104 (Connection reset by peer))
03:35:24 Join KBH [0] (
03:39:34 Quit efyx ("Quitte")
03:40:51 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
03:41:06 Quit HellDragon (Client Quit)
03:42:37Unhelpfulmaybe just a "i need to compact" flag that the buffer code sets. the running codec can periodically check it, continue if not set, send an event and then wait for one if set, and reload any pointers to buffer data
03:43:16 Join HellDragon [0] (
03:43:56Unhelpfulthe same mechanism could also be used for something amiconn and i threw around a while ago, moving the audio buffer end dynamically to allocate plugin buffer on-demand
03:52:47JdGordonUnhelpful: currently, images are resized on load yeah?
03:54:28CIA-61New commit by kkurbjun (r22372): M:Robe 500: Correct UI simulator.
03:57:26 Quit martian67 (Remote closed the connection)
04:03:08 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
04:05:39 Quit martian67 (SendQ exceeded)
04:07:32 Join martian67 [0] (n=martian6@about/linux/regular/martian67)
04:11:15 Join fdinel [0] (
04:13:06 Join CaptainKwel [0] (
04:21:20kugelcheckwps is really horribly broken
04:25:58*kugel stops trying
04:26:05kugela rewrite would be appropriate
04:30:18 Quit dmb (Read error: 54 (Connection reset by peer))
04:30:44 Join ehntoo [0] (
04:38:06 Join Topy44 [0] (
04:38:30Topy44hm, before i start looking for it for ages: where in the source are possible filenames for album art defined?
04:38:46Topy44(and why doesnt it just take any file that ends on .jpg or any other compatible format?)
04:39:34gevaertswhy should it do that?
04:39:40 Quit alexja ("leaving")
04:39:50Topy44because my albums often have random filenames or something like bandname.jpg
04:39:51 Join elcan [0] (
04:40:08Topy44of course i can rename them all - but it would be a lot less effort if it would do so automatically
04:40:40gevaertswhat if you have more than one .jpg file?
04:41:08Topy44either just pick one, or better yet, search for something like "front" or "cover" in the filename
04:41:11Topy44one example:
04:41:47Topy44i have many scene releases, in those the album art is generally called something like 00-bandname-albumname-edition-year-front.jpg
04:42:06Topy44and i like to keep the original filenames of scene releases
04:43:03Topy44it could also switch through different available files slideshow-style every couple seconds, but that might be overkill :)
04:43:18gevaertsI can see why you'd want it, but I also think it's a bad idea with lots of complicated edge cases
04:43:50Topy44i think a simple search for something like "front" or "cover" would be right a lot more often then not
04:44:20Topy44another good example, i dont know which software produces those files but i have MANY that have files formatted like this:
04:44:50gevaertsit would have to go through the entire directory list, which isn't needed now, it has to deal with deciding which to pick if there are multiple matches, and I'm not at all clear how to handle cover art in parent directories with that scheme
04:45:29Topy44parent directories? you mean, like, parent of the directory the media files are in? why would it ever have to handle that?
04:45:41kugelTopy44: laziness is not going to get you very far here :)
04:45:42gevaertsit does handle that now, and people use it
04:45:46Topy44kugel heh
04:46:18Topy44hm, i cant see why i would ever have the album art in a parent dir. only case i have some times is that its in a subdir, like "album art"
04:46:31gevaertsI have most of my cover art in the parent directory
04:47:12kugelTopy44: in the parent dir is useful for multi disc albums, so you just need 1 cover for all discs
04:47:24Topy44actually yes, that does make sense
04:47:47kugeland for <artist>.bmp/jpeg which is also supported IIRC
04:48:09kugeljpg rather I guess
04:48:13*gevaerts uses the "album" concept to model "work", and he uses composer portraits as "cover art"
04:48:29Topy44i get the argument that you need to process the entire directory, but then again, it does that already when queueing the files, it could search for mathing album art at that point and cache it somewhere
04:48:56Topy44gevaerts: for classical music you mean? that makes sense, yes
04:49:03gevaertsit doesn't have to do that when queueing the files. You could be using a playlist or the database
04:49:29Topy44still, i think the search should be a bit more flexible... at least make the file names flexible, like in a config file
04:49:39kugelTopy44: the point is, we don't want to complicate our firmware if more powerful pc apps can do it more easily and faster for us
04:50:11Topy44well, can you recommend a good program that instead of downloading new art can find randomly named jpegs and rename them properly? in a fairly automatic way?
04:50:40Topy44i have an enormous music collection and it would be a lot more effort then its worth to do it manually album by album
04:50:56kugelI use mediamonkey, it can drop folder.jpg, or you can do funny things with the filenames based on the tags
04:51:06*gevaerts would probably just write a quick shell script for this
04:51:11kugel(while syncing)
04:51:38Topy44i have 2k+ artists alone, each with sorted album subdirs
04:52:18Topy44sorting it all when i started getting serious took me around 2 months :)
04:52:50kugelwell, this is getting off-topic actually, but mediamonkey will handle that easily and automatically
04:52:58Topy44i'll check it
04:53:10Topy44thanks anyway
04:53:21kugelyou're welcome
04:53:22Topy44back to the subject: where does rockbox store the filenames?
04:53:38Topy44so i can make a custom build that supports the format at least most of my album art has
04:54:45kugelcheckout albumart.c
04:54:55UnhelpfulJdGordon: yes... the resizer is passed a callback to the image reader that it uses to get chunks of image data
04:54:56kugelin apps/recorder
04:56:58Unhelpfulthis is not ideal if you want the same image in different sizes, there is no rewinding. at *some* point i mean to reorder that part, and have the scaler's state entirely encapsulated in a struct, with the image loader calling a scaler function for each chunk of data. this could easily expand to multiple-size output by passing intiliazing more than one struct scaler_state and making more than one call per image chunk loaded.
04:58:38Unhelpfuli suspect this will see a slight performance gain (no check-if-more-data-is-needed in scaler loop, just loop until the passed data chunk is finished, then return), and will make it easier to use the scaler also.
04:59:09kkurbjunkugel, JdGordon, I have a wps that segfaults the simulator with the latest changes
04:59:27JdGordonsweet... email it to me :)
04:59:47Unhelpfulright now, you have to encapsulate enough of your image reader's state into a struct for the callback to be able to read in the next chunk
05:00:16kkurbjunJdGordon: sure, it's not exactly a complete or proper wps - it was just something that a member of the m:robe forms put together
05:00:22kkurbjunI'll send it to your gmail
05:00:37*JdGordon is off for some L4D
05:01:02 Quit JdGordon ("Leaving.")
05:02:57 Quit fdinel ("Miranda IM! Smaller, Faster, Easier.")
05:03:06kugelkkurbjun: which target?
05:03:15kkurbjunon the m:robe 500
05:03:56kkurbjunthe test wps I put together works fine, but this one kills the simulator
05:04:06kugelprobably jdgordons bug then :p
05:11:13 Join JdGordon [0] (i=1816d253@gateway/web/freenode/x-dncjdjnqlxpmfvph)
05:15:53***Saving seen data "./dancer.seen"
05:16:03kkurbjunwould it be appropriate to delete the dabo theme on the ipod videos?
05:16:45 Quit JdGordon (Ping timeout: 180 seconds)
05:17:36kkurbjunit just has the example wps and the image that is shown I'm sure is not CC-BY-SA
05:17:48kugelnot sure
05:24:27 Join FOAD_ [0] (
05:33:14 Quit kugel (Remote closed the connection)
05:35:26 Quit KBH (Read error: 104 (Connection reset by peer))
05:35:30 Join HBK- [0] (
05:41:30 Quit FOAD (Read error: 110 (Connection timed out))
05:41:30 Nick FOAD_ is now known as FOAD (
05:51:17 Quit Horscht ("Verlassend")
05:59:54 Quit Topy44 ()
06:08:21 Quit ehntoo (Read error: 110 (Connection timed out))
06:31:47 Quit toffe82 ("ChatZilla 0.9.85 [Firefox 3.5.2/20090729225027]")
06:36:05 Join JdGordon [0] (i=1816d253@gateway/web/freenode/x-fhjdfyypvukddwxf)
06:40:35 Join martian67_ [0] (n=martian6@about/linux/regular/martian67)
06:47:54 Quit martian67 (Connection timed out)
07:11:54 Quit sharp (Read error: 104 (Connection reset by peer))
07:14:17 Quit courtc (Read error: 113 (No route to host))
07:15:54***Saving seen data "./dancer.seen"
07:26:48 Quit CaptainKwel (Remote closed the connection)
07:35:28 Quit HBK- (Read error: 104 (Connection reset by peer))
07:35:32 Join KBH [0] (
07:38:10 Quit Trista901 (Remote closed the connection)
07:39:35 Join LinusN [0] (n=linus@gateway/web/cgi-irc/
07:39:47 Part LinusN
07:39:53 Join LinusN [0] (n=linus@rockbox/developer/LinusN)
07:52:43 Join Llorean [0] (
08:01:41 Join n1s [0] (n=n1s@rockbox/developer/n1s)
08:07:26 Quit mylogic (Client Quit)
08:11:17 Join Rob2223 [0] (
08:17:15 Join ender` [0] (
08:18:36 Quit Rob2222 (Read error: 60 (Operation timed out))
08:29:25 Join flydutch [0] (
08:41:06 Join JdGordon1 [0] (
08:41:32 Join Grahack [0] (
08:42:58 Quit JdGordon (Ping timeout: 180 seconds)
08:43:14 Nick JdGordon1 is now known as JdGordon (
08:47:38 Quit thegeek_ (Read error: 104 (Connection reset by peer))
08:47:51 Join thegeek_ [0] (
08:58:56*n1s wonders how one would use this user definable viewport thingy
09:02:24 Join petur [50] (n=petur@rockbox/developer/petur)
09:08:09 Join daurn [0] (i=daurn@freenode/staff/daurnimator)
09:10:26 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
09:15:54 Quit faemir ("Leaving")
09:15:55***Saving seen data "./dancer.seen"
09:17:00CIA-61New commit by jdgordon (r22373): fix the %xd<id> tag parser to complain if you try to display an image it hasnt loaded yet
09:18:20JdGordonn1s: its pretty pointless as it is... but the plan is to put the list *somewhere* in the screen and then have the rest of the display skinned...
09:19:11 Join courtc [0] (n=court@unaffiliated/courtc)
09:22:30n1sJdGordon: so you can specify an arbitrary box and the list will draw inside that?
09:23:18n1saha, so if it's pointless now, what is the future plan? ;)
09:24:44JdGordonwhat I just said :)
09:24:56JdGordonthing the current status bar on steroids
09:25:35 Join Thundercloud [0] (
09:26:06n1sskinned == wps-style stuff?
09:27:25n1sah, i read skinned as "showing a pretty picture" :)
09:27:35 Join Saline [0] (
09:27:51JdGordonI'm sure there will be themes which do that soon
09:28:08JdGordonAA + track info in the menu
09:28:14n1sanyway, a nicer statusbar will be nice
09:30:04SalineHi all
09:30:23SalineI'm trying to add a feature to rockbox.
09:30:58SalineI want to show a clock instead of the "do no disconnect", "charging", etc screens
09:31:11 Quit SUSaiyan` (Read error: 104 (Connection reset by peer))
09:31:19 Join SUSaiyan [0] (
09:31:35Salinecurrently, i'm looking at the screens.c and usb.c
09:32:07 Join pamaury [0] (
09:32:10Salinethe problem - i think that the "do not disconnect" screen is the default ipod screen
09:32:35Salinecan you give me some pointer? (other than *, and ** ..)
09:32:48JdGordon0x8765ff0 :)
09:33:18Salinethanks for the pointer..
09:33:54JdGordonumm... what clock do you want to show?
09:34:03*JdGordon is half asleep
09:34:05Salinelets start with an "hello world"
09:34:13n1syeah, the "do not disconnect" screen is in apple fw
09:34:26Salinei'm thinking about using the the clock plugin, later
09:34:42Salineis there a way to bypass the apple firmware?
09:34:46JdGordonthat should be doable
09:34:56n1sSaline: do you still get that screen with a current rockbox build, i thought we used our own usb now?
09:34:57JdGordonuse rockbox usb and you wont get that
09:35:01*JdGordon goes to bed
09:35:13Salineok, i'll give it a whirl, thanks!
09:35:29Saline(i'm using the stable build, via the installer, btw)
09:35:30 Quit KBH (Read error: 54 (Connection reset by peer))
09:35:36Salinegood night!
09:35:38 Join KBH [0] (
09:38:40 Quit SUSaiyan (Read error: 104 (Connection reset by peer))
09:38:47 Join SUSaiyan [0] (
09:39:10n1sonce you use rockbox's own usb mode, usb_screen in screens.c is what is drawing the usb graphics in rockbox while connected so you probably want to modify that
09:45:12 Quit Thundercloud (Read error: 54 (Connection reset by peer))
09:52:21 Quit SUSaiyan (Read error: 104 (Connection reset by peer))
09:54:36 Join SUSaiyan [0] (
10:02:57 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
10:05:58 Join robin0800 [0] (
10:11:03 Join zhxk [0] (i=zhxk_@
10:11:03 Quit fyrestorm ("lamers envy me like they envy bill g -- main boot xp, just the way it should be!")
10:11:13 Join mt [0] (n=MTee@
10:11:28 Join fyrestorm [0] (
10:11:32zhxkwhats rockbox
10:11:57 Part zhxk
10:14:06 Quit CeruleanC (Remote closed the connection)
10:18:40robin0800zhxk try
10:26:32 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
10:29:11Salineafter i build (using make), where does the binaries go?
10:29:16Saline(using cygwin)
10:35:38Salinemake zip
10:35:45Salinei should rtfm mroe
10:55:53robin0800Saline: if you followed instructions you will have created a "build directory" the files will be in there
11:13:26 Join Topy44 [0] (
11:15:47 Join bertrik [0] (n=5a809af7@
11:16:00***Saving seen data "./dancer.seen"
11:17:42bertrikUnhelpful: regarding the ams sansa RTC: the rtc driver was re-used from the older sansas
11:18:04bertrikthose older sansas used 1980/1/1 as the reference date, while the ams sansas use 1970/1/1 as reference date
11:18:41 Join antil33t [0] (n=Mudkips@
11:18:47bertrikThis caused the RTC to be reset by the OF (or possibly even rockbox) and I think this reset was the reason that DRM stopped working
11:19:22bertrikThe RTC driver used for the ams sansa is now updated to use 1970/1/1 as the reference date, just like the OF
11:22:29 Quit J-23 ("wszedłem")
11:29:47 Join Isolinear [0] (
11:30:23 Quit Isolinear (Client Quit)
11:30:45 Join Rudy4Pez [0] (
11:31:08Rudy4PezAnyone awake?
11:31:28Rudy4PezI have a question about albumart filenames.
11:32:35Rudy4PezRockbox searches for certain formats such as "folder.jpg" or "cover.jpg"
11:32:48Rudy4PezI would like to know if there's a way to add one to that list.
11:33:25Rudy4PezBecause all of my albumart files are named "album.jpg", which of course, Rockbox doesn't recognize.
11:33:31 Quit Saline (Read error: 54 (Connection reset by peer))
11:33:47 Quit bertrik ("CGI:IRC (Ping timeout)")
11:35:42 Quit KBH (Read error: 104 (Connection reset by peer))
11:35:46 Join HBK- [0] (
11:37:00Topy44Rudy4Pez: actually i asked that exact question a couple hours ago
11:37:30Topy44answer is: yeah, check out apps/recorder/albumart.c line 124 onwards
11:38:22Topy44in essence, duplicate lines 199-205 and replace "cover" through "folder" or whatever
11:38:37Topy44compile, run, have fun :)
11:39:42 Join Saline [0] (
11:41:44 Join BdN3504 [0] (n=4e3435ff@gateway/web/cgi-irc/
11:43:23 Join J-23 [0] (
11:43:23markunif it's a very common naming scheme I guess we should add it
11:43:23robin0800Rudy4Pez: you could always do a find and replace outside of rockbox unless you need album.jpg for something else
11:43:53Topy44it is, some (unknown to me) software produces "Folder.jpg" files, i have tons of those
11:44:41Topy44nearly as common is front.jpg
11:44:45BdN3504which section of the manual would you deem appropriate for the new ui viewport: Theme Settings, Advanced Topics, WPS Tags or Config file options? i'd go with Advanced Topics and the latter
11:44:54robin0800Topy44: windows media player does that
11:45:52Rudy4PezGah, "compile"!
11:45:56Rudy4PezSuch a dirty word! :P
11:46:33*Rudy4Pez was hoping for an easy-to-edit xml file or something.
11:46:57robin0800Rudy4Pez: I gave you an alternative to compiling
11:47:22Rudy4PezAh, sorry. It's 3am here. I'm a bit loopy. :)
11:49:52robin0800Rudy4Pez: think I used infranview a couple of years ago
11:50:20Topy44hm yeah, some hexeditor-hacking might actually do the trick... %scover%s. %sfolder.jpg /.rockbox/albumart/%s-%s%s. /.rockbox/albumart/ <== shows up in my rockbox.iriver
11:50:58Topy44(or non-destructive texteditor)
11:51:18Topy44but keep the size the same
11:51:49robin0800Topy44: bulk renamer can change filenames
11:51:58Rudy4PezI'm sorry, what exactly are you renaming?
11:52:06Rudy4PezErr.. Editing?
11:52:11Rudy4PezSorry.. Brain is fried.
11:52:38Topy44i just realized you were talking about a search/replace on the music dir, not inside the binary :) my brain is fried already too, though my version MIGHT actually work aswell
11:53:09Rudy4Pez(And yes, I'd rather not rename the .jpg files.)
11:53:23robin0800Rudy4Pez: album.jpg to cover.jpg
11:54:03n1sRudy4Pez: then your last option is to convince a dev to make the change for you ;)
11:54:20Topy44Rudy4Pez: just download the vmware development envoirement, edit the albumart.c file i mentioned before and compile it, its really not hard and there is a good guide in the wiki
11:54:43Topy44its a big download, but the actual editing and compiling bit is a matter of a few lines
11:54:45robin0800Rudy4Pez: or keep two copies album and cover
11:55:26Rudy4PezWhat about an XBMC type solution.. Anyone familiar with their "advanced.xml" file?
11:56:20Rudy4PezIt's been a long time since I've played with it, but the way I remember it, there was an option to add a line..
11:56:24Topy44Rudy4Pez: i suggested more intelligent albumart filename handling this morning - i got some good arguments against it, some bad ones, and a lot of general disinterest :)
11:56:47robin0800Rudy4Pez: If you do compile it please make a patch and post it to the tracker
11:57:21Topy44robin0800: actually i will do that, though i am still making a list of what to add
11:58:15robin0800Topy44: Not too much if you want to see it committed
11:58:51LloreanWhat program creates the name album.jpeg?
11:59:00LloreanOr album.jpg rather
11:59:20Rudy4PezFor example, the system recognizes many extensions as playable audio files.. .mp3, .wav, .flac...
12:00:28Rudy4PezAnd if you want to add a new extension to the list of extensions that get recognized as music, you would add the tag <musicextensions> .new </musicextensions> to the xml file.
12:00:47n1swe *could* add a setting for user specified filenames to look for
12:01:02Rudy4Peznls: Exactly.
12:01:13Lloreann1s: Why add that complexity?
12:01:17n1sRudy4Pez: although you can probably forget xml files :)
12:01:19LloreanThere can't be that many "common" ones.
12:02:04n1sLlorean: it wouldn't be much complication, just a setting string allowing the user to specify a name they want and add that in the search
12:02:38robin0800perhaps just album.jpg but to aswer your question iv'e never seen this
12:02:57Lloreann1s: I asked why though. There's not really much need to support it.
12:02:59Rudy4Peznls: Yeah they had a lot of other stuff in their to warrant setting up that xml system.
12:03:03LloreanWe're still going to have to support all the common names anyway.
12:03:06n1sbut sure, if there are only very few common ones i guess the setting would be kind of pointless
12:03:10LloreanOnce all the common ones are supported, why do you need a user value?
12:03:14Topy44i dont think its needed either - what would make a lot more sense IMHO is having some intelligent code searching for the file even if it has a random filename
12:03:19 Join bertrik [0] (n=5a809af7@gateway/web/cgi-irc/
12:03:19Topy44but that adds complexity of course
12:03:23n1sLlorean: IMO there is no need for AA at all :)
12:03:45Topy44hehe, you could argue about that alright i guess :)
12:03:51Lloreann1s: Basically, once you support all the common names, a "custom" value really only exists if a user wants to use a unique filename incompatible with any other program, which would be a bit odd right there.
12:04:11LloreanTopy44: How on Earth do you identify which jpeg in the folder belongs as album art?
12:04:27LloreanThere's no such thing as an "intelligent" algorithm, and we already match things such as album and track name.
12:04:58Rudy4PezHmm.. I recall a program that created a ridiculously long string of random characters as the jpg filename..
12:05:04Topy44Llorean: first of all, if there is just one jpeg, use that, no matter what its called. second, if there are several, and one contains "cover", "front", the albumtitle, etc, anywhere in the filename, use that. and if nothing works, just use the first one.
12:05:06n1sLlorean: sure, i have no idea how these files are named usually
12:05:10Rudy4PezI wanna say Windows Media Player but I'm not sure.
12:05:12 Join einhirn [0] (
12:05:18LloreanTopy44: And what do you do if that jpeg *isn't* supposed to be album art?
12:05:39LloreanWhy include a system that can get it wrong, when humans are capable of simply getting it right in the first place?
12:05:44Topy44then you get a wrong jpeg displayed. doh. :) why would you have anything else then album art or a bandphoto in a music dir?
12:06:04LloreanYou're making an assumption that users don't do silly things.
12:06:05Topy44because i have 10k or so albums, and will not start sorting them by hand
12:06:06Rudy4PezTopy44: That makes sense, if none of the .jpg files match the format, then just pick the first .jpg available.
12:06:21Topy44it could all be optional of course
12:06:26LloreanTopy44: A simple mass rename will do, it's not hard. If you have a single jpeg in each folder, just mass rename or copy all to cover.jpg
12:06:29Rudy4PezOptions are good.
12:06:30Topy44in general, what i suggest isnt much effort, but requires parsing the directory, and thats where the actual problem lies
12:06:57LloreanYou're suggesting introducing a system that can introduce errors, and requires the player to do more work, because you're willing to do a simple file rename?
12:07:12Topy44yes :)
12:07:28Topy44well, the errors are not catastrophic - worst case you get a wrong image displayed
12:07:48LloreanAnd the current problem is not catastrophic either
12:07:54LloreanWorst case, you actually need to organize your files slightly.
12:08:06Rudy4Pez10k albums is a lot of organization legwork.
12:08:11LloreanNo, it's not.
12:08:15LloreanIt's a single mass operation.
12:08:26robin0800Llorean: On the AA page are programs that find album art and can rename it to cover.jpg
12:08:35LloreanIf an algorithm in Rockbox can identify the right jpeg, a rename algorithm can too
12:08:42Topy44not true actually - it simply means finding a program that does what i just said on the pc instead of putting it into rockbox
12:08:44LloreanIf it *can't* then the problem can't be solved in Rockbox either.
12:08:56LloreanSo it's not "a lot of legwork"
12:09:22Lloreanrobin0800: Exactly.
12:10:01LloreanI'm all for improving the hardcoded list of filenames to match those used by common PC-side applications, but once you try to develop a method to "detect" album art by guessing things you get problems like the database had trying to deduce track names from filenames.
12:10:13Rudy4PezWell I think in this case the easiest solution is to add more formats.
12:10:20Rudy4PezLlorean: Agreed.
12:10:24LloreanAdding formats is pretty difficult.
12:10:24Topy44i think the clean solution if you were to integrate it into rockbox would be to find the coverart file when creating the dynamic playlist (as the directory already gets parsed at that step), or possibly when creating the database if you use it, and save it in the playlist/database
12:10:28LloreanDo you mean adding more filenames?
12:10:32Rudy4PezErr, yes.
12:10:41Rudy4PezFried brain, remember? :)
12:10:53LloreanTopy44: The clean solution is to have your files properly named.
12:11:04Topy44i said "if you were to integrate it"
12:11:07LloreanIf your file is properly named, Rockbox already finds it just fine.
12:11:17Rudy4PezHaha, my files are all properly named. Rockbox just doesn't recognize that particular name.. :P
12:11:29LloreanIf there were going to be some funny algorithm, it'd just take place when the current finding takes place.
12:11:52Topy44yeah, and if all your music files have perfect tags and meaby even are all the same format and bitrate, hey, you could eliminate half of rockbox!
12:11:57LloreanRudy4Pez: You never did mention what program names them that way, did you?
12:12:29Rudy4PezHmm.. No, I don't remember..
12:12:44LloreanTopy44: Just rename your files and quit complaining. Seriously. There's not a good reason to have Rockbox try guessing things.
12:12:55Llorean"I want it to" isn't a good one, by the way.
12:12:59Topy44really what it comes down to is the compromise between putting effort into maintaining your local collection and having rockbox do more
12:13:01Rudy4PezI'm trying to figure out if Winamp had something to do with it.. I've been using that filename for a long time.
12:13:13Topy44its like "retagging all local files" vs "have player guess tags"
12:13:18Topy44its all an ancient discussion
12:13:22LloreanAnd we don't have the player guess tags, either.
12:14:12Rudy4PezBut yes, I think simply adding "album.jpg" as well as whatever other somewhat common filenames are out there as they pop up, is probably the simplest solition, and adds to the flexibility of Rockbox, and this is a run-on sentence. :)
12:14:58Topy44album.jpg, front.jpg and folder.jpg will be in a patch i'll create shortly then
12:15:08LloreanWhat programs use those?
12:15:27Topy44folder.jpg <== windows media player?
12:15:40Lloreanfolder.jpg is already supported
12:15:41robin0800Llorean: wmp uses folder.jpg
12:15:45Topy44front.jpg <== common manual naming scheme (about 50% of my coverart)
12:15:52Topy44Llorean: ah, ok
12:15:59Llorean"common manual naming scheme" doesn't say what programs use it.
12:15:59Topy44album.jpg no idea
12:16:03LloreanIt's not common just because you do.
12:16:08Topy44i dont
12:16:32BdN3504this is settled. go get yourself The rename search for files called *.jpeg. drag and drop into rename. modify prefix. end of discussion.
12:17:02Topy44i download stuff. the files are called front.jpg, and i dont usually care as my pc music player doesnt display it anyway
12:17:03CIA-61New commit by mt (r22374): Put ATRAC3Context in IRAM, 2.5% speedup on PP502x, 20% on ColdFire.
12:17:21LloreanTopy44: So you basically want an arbitrary filename supported.
12:17:46Rudy4PezJust because it's arbitrary doesn't mean it's not insanely common.
12:17:57Topy44Llorean: all i know is that its common, i coulndt care less why it is
12:18:07LloreanExcept you don't even have a program that uses it either.
12:18:24Topy44because i dont use album art on my pc
12:18:39BdN3504do as i say. here, i even have the download link for you:
12:18:47LloreanBdN3504: You're missing the point of the discussion.
12:19:12LloreanRudy4Pez: But if it were insanely common, you'd think PC software would support it too. Or that we'd have actual other people with such a thing.
12:19:27Topy44i said "i dont know as i dont use any", not "there isnt any"
12:19:28LloreanIt may be just the music store he gets his music from, or the band, or whatever.
12:19:48LloreanTopy44: I didn't say "there isn't any." I said "tell me some."
12:20:18BdN3504Llorean: no. there is no discussion. rockbox displays albumart files if they are named
12:20:19 Join bertrik_ [0] (
12:20:48LloreanBdN3504: And when we put in jpeg album art, we decided we wanted to try to make the list contain common filenames, expanding it if there were other common names used by PC apps
12:20:53LloreanWhich is exactly why I'm asking these questions.
12:21:05Topy44so because pc software might not explicitely search for that filename (most pc software will parse the dir for jpegs anyway) that is a killer argument not to use it in rockbox?
12:21:06LloreanThat list is definitive of what it does - it does not limit future expansion of the list.
12:21:23LloreanTopy44: No, most pc software won't parse a folder for random jpegs.
12:21:27Topy44also, i cannot tell you any pc software that supports it because - again - i dont use a single pc software that supports album art at all
12:21:46Topy44(except for album art downloader of course)
12:21:55LloreanSo *find* some
12:22:01LloreanYou're the one claiming its common, you should be able to.
12:22:17LloreanI'm not making a huge demand.
12:22:29BdN3504Llorean: that's seven options. if users are to lazy to pick one of them and conform it's their problem, imo
12:22:32 Quit bertrik ("CGI:IRC (Ping timeout)")
12:22:36LloreanJust show that we'd be making Rockbox compatible with a documented layout of some sort.
12:22:37Topy44i can give you proof that its common by giving you a grep on my music dir... and i have not created any of these front.jpg files myself nor do they all come from the same source
12:22:51BdN3504doensn't it slow down the process of loading aa if we expand that list?
12:22:53LloreanBdN3504: Yes, but your opinion doesn't matter to what was decided when the feature was added.
12:23:04LloreanTopy44: That's proof of nothing at all.
12:23:15LloreanTopy44: If it's common, why are you unwilling to even try?
12:23:24LloreanOr is this another "I'm lazy and just demand it my way" situation?
12:23:26BdN3504if it does not do that, then i think it
12:23:33robin0800 this needs to be updated to reflect use of jpg conversiom to bmp not needed now
12:23:45Topy44i dont demand anything, i think thats a point you are somewhat missing :)
12:24:03Lloreanrobin0800: "where to put the images" doesn't say anything about conversion.
12:24:34LloreanTopy44: You're being rather insistent that a filename should be supported that has nothing speaking for its commonality except that you've downloaded a bunch of music from a source you won't even specify that has that file in it.
12:24:38Topy44i dont demand anything, as i can fix it myself with a 2-minute patch, and if you dont want to commit it on the grounds that i cant "prove" to your satisfaction that i am not the only person using that filename then thats not really my problem :)
12:24:56robin0800Llorean: further down the page it does
12:25:05Lloreanrobin0800: Ah, I just read the anchor you linked.
12:25:35LloreanTopy44: What have you got against renaming them anyway, if none of your PC software supports it?
12:25:54LloreanThe idea is for Rockbox to support album art that is already being used with PC software, not "every filename someone thinks is common enough" anyway
12:26:13LloreanIf the files aren't named to be compatible with something, there's no reason they can't be renamed at that point.
12:26:15Rudy4PezI've lost count of how many times I've come across "front.jpg"... Usually accompanied by "back.jpg".
12:26:23Topy44some sources: soulseek, rockbox (the bittorrent music tracker :), private music archives of friends - but no specific source as in software or online shop
12:26:30Topy44correct, and cd.jpg etc
12:27:07Topy44correct, they can be renamed - but it might save a lot of users some effort if its supported directly
12:27:10LloreanSo it's not a compatibility issue, it's just a personal preference issue?
12:27:25LloreanOr rather, it's not a software compatibility issue.
12:28:17LloreanI mean you have to understand, there's going to be plenty of people with favorite naming schemes.
12:28:56Tornethe names we support already are designed to accomodate people who have *already* renamed everything to match the convention of some *other* piece of software; i.e. to prevent them from having to break compatibility with the software they already use.
12:29:27Topy44i could get all my mp3s copied to cassettes to make them compatible with my walkman - instead i bought an mp3 player - get my point? understand, this is not my personal favourite naming scheme, it is one that both i and others (rudy4pez) have seen many times, and its source seems to be pretty obvious: common sense (front as in the opposite of back)
12:29:57LloreanExcept your comparison to a walkman is silly
12:29:59Topy44i am not saying "change this because it will make it compatible to software X"
12:30:03Topy44it is really :)
12:30:05LloreanRenaming the files doesn't reduce their quality, or usability.
12:30:18LloreanYou're basically just trying to make something up to justify your statement now.
12:30:27Topy44let me try:
12:30:36Llorean"I don't want to rename them."
12:30:37 Join josh___ [0] (n=5312dd93@gateway/web/cgi-irc/
12:30:40Lloreanthat's basically your whole argument.
12:30:51Topy44doh, let me at least try (and no, it is not.)
12:30:56LloreanYeah, it is.
12:31:09LloreanYou're going to give reasons why it's good that you shouldn't have to - ease of use, simplicity for users, commonness of the naming
12:31:11LloreanBut in the end, that's the reason.
12:31:45LloreanMeanwhile you're going to willfully ignore that we don't want the list infinitely long, and that a very practical line to draw is "making it compatible with existing software" so that we *don't* have to make value judgments on "how common" a naming scheme is
12:32:02LloreanBut, feel free to prove me wrong.
12:32:06LloreanOffer your reasons.
12:32:13josh___hi! anyone tried playing pokemon red/blue on rockboy and noticed game crashing when you get killed or when trying to heal?
12:32:27Lloreanjosh___: Turn sound on.
12:32:35LloreanOr possibly off. I can't remember. It's a gnuboy bug.
12:32:46Topy44renaming them is effort (just a little, but still) for every user that has files with that naming scheme. adding the filename to rockbox instead is a little effort for one person. and...yeah, it is the reason really :) but not because i'm "lazy" (i have already offered to create the patch), but because i think rb should facilitate things for the users/make using their player to the full extent as little effort as possibly
12:32:57josh___w8 i can check this out...
12:33:18Topy44therefore i have 2 possible solutions:
12:33:23LloreanTopy44: You can't facilitate *all* things. Having lines you can draw that don't require value judgments is very handy.
12:34:17LloreanThe vast majority of users who use cover art will have renamed it to an existing scheme. You're probably quit uncommon in a desire to use cover art in the mobile case, but not the PC.
12:34:26Topy441. collect actually common naming schemes (some that are already supported i have never seen for example), see how common they are among a certain selected usergroup with large collections, add ones that are obviously common
12:34:52Topy442. make it configurable, allowing everybody to add their own personal favourite naming scheme, though that increases complexity a fair amount
12:35:04LloreanTopy44: You going to be around ad infinitum to update the list, and field complaints about a favorite scheme not making the list?
12:35:08Topy44of course there is also 3. which is "call me an idiot and ignore me" :)
12:35:15LloreanOption 3) Make it compatible with software, rather than people.
12:35:16josh___Llorean: TX a lot! it helped!
12:35:19LloreanNobody called you an idiot.
12:35:38Topy44its software for people, no? an entertainment product, not a scientific tool for example
12:35:58TorneTopy44: Code is not free
12:36:10Torneputting more choices in, or making it configurable, takes memory
12:36:15Topy44Llorean: we really are going around in circles - both you and me, i believe to have valid points
12:36:17Torneputtin gmore choices in takes execution time
12:36:29Tornewe have a pretty sensible criteria to keep the list down to a manageable size
12:36:58 Quit josh___ ("CGI:IRC 0.5.9 (2006/06/06)")
12:36:59Topy44torne: agreed, which is why personally i would reevaluate the ones already in the list (also, the current method of checking which files exist looks horribly inefficiant to me)
12:37:15LloreanWhich ones already in the list seem wrong or bad to you?
12:37:25TorneTopy44: what's wrong with the ones in the list? "named after the file" "named after the album" folder.jpg and cover.*
12:37:30Tornethat all seems very logical :)
12:38:24Topy44personally, i find "compatible with common naming schemes" more important then "compatible with existing software" (and the first largely includes the second)
12:38:51TorneThe second is what technically *prevents people from using the feature* though
12:38:59Torneif, say, our list was just front.jpg
12:39:01Topy44well for example i dont have any "filename.ext" ones (files named the same as the current music file)
12:39:13Tornethen there would be no way to rename the art such that it worked in rockbox *and* windows
12:39:17LloreanSo because *you* don't have any it should be reconsidered?
12:39:20LloreanThey're all either known to be used by other software, or necessary to offer the feature set we do (song-specific images, and album-specific images from a pool folder)
12:39:31LloreanWe can't offer song-specific album art without filename.jpg/bmp
12:39:36Tornethus, supporting windows' scheme (folder.jpg) is *vital* so that people aren't stuck in an impossible situation
12:39:40Topy44Llorean: so because *you* dont have the front.jpg one it should NOT be considered?
12:39:49LloreanTopy44: No, because nobody can name software that uses it.
12:39:55TorneTopy44: When I have files called front.jpg I rename them to folder.jpg
12:39:57LloreanSee, established criteria here, personal opinion over there.
12:40:02Torneso that windows displays them, and so they work with rockbox :)
12:40:29LloreanSo why don't you try again, maybe explain why filename.ext is bad?
12:40:57LloreanIt meets an actual need, rather than a simple preference.
12:40:59Topy44i wont because i agree that if someone can name reasonably common software that uses that it should stay in.
12:41:30LloreanSo you objected to it a moment ago, but don't now?
12:42:02 Join alexbobp_ [0] (
12:42:57Topy44all i said is i dont know any software that uses it nor do i have any files that match it - of course someone else might
12:43:29LloreanAs I said, that filename exists because a Rockbox feature requires it for the feature to work.
12:44:07Topy44ok, software that uses front.jpg: foobar2000 supports it apparently
12:44:35LloreanSee, not too hard, was it?
12:44:36LloreanThat's all I asked
12:44:51Rudy4PezLlorean is stubborn. :P
12:44:56Topy44he is isnt he :)
12:45:04TorneLlorean's stubbornness is a feature, not a bug
12:45:08LloreanThough I'd like to see your source
12:45:10Rudy4PezI haven't been on Rockbox IRC in about 2 years.
12:45:16LloreanMine says foobar2000 is simply configurable and you can *add* front.jpg
12:45:21Rudy4PezStill stubborn.
12:45:29 Join kugel [0] (n=kugel@rockbox/developer/kugel)
12:45:46 Quit Rudy4Pez ("Java user signed off")
12:45:52Topy44let me check, this forum post i found suggest its default but a forum post is of course not a reliable source
12:45:58 Join Rudy4Pez [0] (
12:46:05LloreanI mean, if it's there by default, that's great.
12:46:06LloreanGo for it.
12:46:21LloreanBut if it's configurable, and just something someone added to their personal configuration, different story entirley.
12:46:41Rudy4PezWell there must be more to it than that, because it's very very common.
12:46:54LloreanRudy4Pez: Commonness isn't really a factor though.
12:47:13*kugel thinks it would be fine to support front.<ext>
12:47:25kugelI've also seen it often
12:47:26 Nick bertrik_ is now known as bertrik (
12:47:26Rudy4PezLlorean: Shouldn't it be?
12:47:52Rudy4PezThe idea is to make Rockbox work for you, not to make you work for Rockbox.
12:47:56kugelRudy4Pez: it is, commonness is the only reason we support folder.jpg
12:48:09Lloreankugel: No, we support folder.jpg to maintain compatibility with other software.
12:48:13kugeland we don't support folder.jpeg because it's uncommon
12:48:47LloreanCollections that show album art in other players should show it in ours, it's more or less that simple.
12:49:18Llorean"commonness" doesn't really factor into that - just what works elsewhere.
12:49:45LloreanBesides, if it's so *very* common, I'm amazed at least a moderate amount of FLOSS doesn't support it natively
12:49:52BdN3504who has access to the themes site? you better remove this theme. it contains a copyrighted graphic, the wps is simply copied off the wps creating guide.
12:50:20 Quit Rudy4Pez (Client Quit)
12:50:49Topy44Llorean: confirmed, foobar2000 loads front.jpg by default
12:50:57Topy44i just installed and tested it
12:51:07LloreanTopy44: There you go.
12:51:46kugelBdN3504: the theme itself doesn't contain any graphic
12:52:48 Quit n1s ("Lämnar")
12:54:06LloreanAn option for custom filenames for AA might be to do it in the WPS instead.
12:54:24LloreanOr rather specifically, allow images in the WPS to be relative to the played file path, rather than the WPS path (through some alternate tag)
12:55:01*linuxstb thinks front.jpg/back.jpg is a sensible convention, but also that we don't want to expand the search for album art any more than it already is, and simply renaming front,jpg to cover.jpg would solve it.
12:55:30Lloreanlinuxstb: Well, if foobar2000 already supports it, we probably should. I think if collections work with a PC player, we shouldn't really ask for renames.
12:55:31bertrikso the conclusion of this is that since foobar2000 loads front.jpg, rockbox could do it too? or is this not enough?
12:55:42Topy44my personal preference would still be not having to rename anything at all and have rb parse the dir for any jpeg it can find - seems to me like a perfect patch project to keep uncommited
12:56:18Topy44and i am sure i can find some pc software that will parse the dir and guess the right cover :)
12:56:23 Quit alexbobp (Connection timed out)
12:56:31linuxstbLlorean: I don't think changing the wps is the solution - it would be better as a global option.
12:56:34LloreanTopy44: And you're sure I can't create a folder that stumps it?
12:56:48Lloreanlinuxstb: I was thinking it'd offer some other interesting flexibility too though
12:56:55linuxstbLlorean: foobar2000 does lots of things Rockbox doesn't, such as supporting odd tag/container-format combinations...
12:57:12Lloreanlinuxstb: Like, if image and backdrop tags were allowed to be relative to the song, you could actually have artist-specific variants of a WPS.
12:57:14Topy44Llorean: well you can of course make it pick the wrong image, but that doesnt really matter much. other then that, i dont see a risk
12:57:33LloreanTopy44: What's the "risk" in not showing an image, then?
12:57:44LloreanI'd say not showing it is better than showing the wrong one.
12:57:58LloreanAt least the former just looks like a lack of a feature, rather than a bug.
12:58:04Topy44i dont agree, as the correct hit percentage would be very high
12:58:22Lloreanlinuxstb: Well, as I said, the discussion when we added folder.jpg basically went "if we find other filenames programs support that we don't, we should probably add them"
12:58:24BdN3504Topy44: the winamp toaster plugin parses the folder for specific filenames you throw at it, you can also use *.jpeg
12:58:33Topy44i think having 99.5% correct and .5% wrong is better then having 100% not there at all
12:58:36LloreanTopy44: Ah, "good enough is good enough"
12:58:46linuxstbTopy44: I partially agree with you, but it could cause problems for people that store their music in a single directory, rather than neatly split by album.
12:58:54LloreanTopy44: You think 100% of album art isn't shown right now?
12:59:59Topy44Llorean: no, it depends on the test group of course - 100% of incorrectly named album art :)
13:00:36linuxstbLlorean: I just think the code already offers enough flexibility. But I'm not strongly against adding front.jpg - it's a sensible convention.
13:00:46Topy44which in the case of my imho rather representable (because large and sourced through many different ways) collection would be roughly 50% of all albums containing album art
13:01:05Lloreanlinuxstb: I do too. I was just saying that if custom filenames were going to be added, maybe it could be made more interesting than that by making it "bigger" as it were.
13:01:20LloreanTopy44: Which, if there's a reliable algorithm, could be mass renamed rather quickly anyway
13:01:32kugellinuxstb: looking up filenames isn't really a wasteful operation, is it? it doesn't even require a disk spinnup with dircache, and the buffering is way more demanding
13:01:56Topy44correct - again, a "sort your collection" vs "let the player guess" discussion, and as we agreed, the difference is "effort for one person" vs "effort for the users"
13:01:58Lloreanlinuxstb: I think though it's good enough to add front.jpg and... was it album.jpg?
13:02:11LloreanTopy44: I didn't agree to that.
13:02:48TorneTopy44: you are, again, asusming that adding features to rockbox only costs developer time
13:02:54Topy44Llorean: you wouldnt agree if i argued that the sky was blue :)
13:02:56LloreanTopy44: And you forget the "hidden cost to users who don't make use of the feature" aspect of it.
13:03:04BdN3504kugel: i think people shouldn't be able to post the example wps to the themes gallery. it's merely a copy of the theme on the wiki site. posting multiple instances of it is a waste of space.
13:03:21LloreanTopy44: You may not be aware of it, but Rockbox is already running up against some pretty strict size limitations.
13:03:25Topy44i dont forget that at all, no
13:03:35kugelBdN3504: ideally we aren't going to need the wiki sites anymore
13:04:17LloreanTopy44: If you didn't forget it, I assume you chose it ignore it when simplifying it down to "effort for one" vs "effort for many"?
13:04:54BdN3504kugel: i don't mean the wps gallery wiki. i mean
13:05:00Topy44there are of course always many more things to consider. one of them being how much actual ressources would be used
13:05:15Topy44as in "is it worth it"
13:05:21LloreanTopy44: So if those are actually worth considering, why are you factoring them out of the discussion?
13:05:22Tornethe runtime cost in cycles/memory/disk access is *far* more important than developer time, mostly :)
13:05:51Topy44ok, lets talk about runtime cost:
13:06:02bertrikMy guess is that the impact of adding another filename is very limited
13:06:06LloreanSimplicity has significant benefits, and the simplest method is a set list of possible filenames.
13:06:17LloreanUnless the list grows to rather large size.
13:07:03Topy44well, the list would still be needed to the most part
13:07:37LloreanSo, it would be added complexity on top of the list to benefit users who have album art that they're not actually already using?
13:07:51Topy44umh...yes :)
13:08:41kugelBdN3504: alright, removed
13:09:21BdN3504^^ thanks finally some consent :)
13:09:28 Join intrados_ [0] (
13:09:39BdN3504kugel: you read my earlier post?
13:10:53BdN3504kugel: quote: "the ? remainz"
13:10:53Topy44as i see it, it is a feature that would actually benefit a significant usergroup (i wouldnt want to guess percentages), by lessening their effort pre-sorting their collection, while "costing" relatively little in the way of cpu cycles and memory, and nothing in the way of disk access (no extra spin up)
13:11:03Topy44also, no cpu cycles at all if made optional
13:11:24Topy44(well ok, some for the "is it on" check :)
13:11:30kugelBdN3504: ??
13:11:34LloreanTopy44: The "significant usergroup" who wants to use album art on their DAP, already has it on their PC, but doesn't already use it on their PC?
13:12:01kugelBdN3504: Advanced Topics -> Customising the User Interface is appropriate I think
13:12:02Topy44or uses software that parses the dir (allowing wildcards)
13:12:04Topy44but yes
13:12:20Topy44go ahead and make a poll :) i am sure that i am not the only one
13:12:29kugelarguing about cpu cycles is useless imo
13:12:42Topy44(i do btw use my album art, i just dont let my media player display it, i actually look at it with an image viewer sometimes)
13:12:56Topy44(therefore the filenames are completely irrelevant as i click on them manually)
13:13:24LloreanTopy44: You're also looking at diminishing returns. To use your own approach, the vast majority of users are already served by what we have.
13:13:35LloreanWhat you propose adding would cost significantly more than what we have, and yet serve a significantly smaller number of users.
13:14:02bertrikLlorean, I'm not so sure about that
13:14:07linuxstbbertrik: Are any of your S5L8700 ports close to running a real Rockbox build? I've been trying to get it working on my Nano2G, but it's failing with odd problems, and I would be curious to know if it's just on the Nano.
13:14:07Topy44are you so sure about that vast majority?
13:14:14Lloreanbertrik: About which part?
13:14:23kugelI wasn't away that gevaerts has put up samsung yh binaries in the test builds forum
13:14:50bertrikLlorean, that it would cost significantly more and server a significantly smaller numbers of users
13:14:59kugelLlorean: are you sure it's "significant"?
13:15:07GaloisI find these discussions a bit wrong-footed. "Vast majority" arguments can and are often used to justify supporting only mp3, or only windows, or any number of support decisions that in other contexts we would dismiss.
13:15:27Lloreanbertrik: You think it would serve even relatively similar numbers of users?
13:15:35Lloreankugel: In what sense?
13:15:46kugel"What you propose adding would cost significantly more than what we have"
13:15:53LloreanGalois: But this isn't whether or not to add support for something or leave it out entirely
13:15:54Topy44you need to do that assession about every single feature in rockbox really - cost in ressources vs amount of users actually benefiting from it
13:16:03***Saving seen data "./dancer.seen"
13:16:04LloreanGalois: This discussion is whether to allow people an extended degree of disorganization.
13:16:13bertrikLlorean, no, I just doubt your statement about the significance
13:16:38Lloreanbertrik: I'm reasonably confident that we would gain very few new users of album art relative to the number we already have.
13:17:03kugellooking for another filename is rather really insignificant
13:17:05Topy44i am not, but of course thats all open to discussion
13:17:18Lloreankugel: If it needs to check every filename in the folder before attempting to decide which one to use, then yes in many circumstances it will cost a lot more time than the current method
13:17:32 Quit intrados (Connection timed out)
13:17:32LloreanWe're not talking about looking for *one* other filename now
13:17:42bertriklinuxstb, the meizu m3 can almost build to a real rockbox image, I haven't really tried for the samsung yp-s3. Overall I haven't put effort in it yet.
13:17:48Topy44no, we are talking about the "parsing the directory" variant
13:17:49LloreanWe're talking about going through the whole folder and trying to *guess* which file from any present jpegs is the album art.
13:18:10kugelit's still insignificant
13:18:28LloreanRelative to the existing method?
13:18:38Topy44yeah, if the directory structure is already in the dircache it would be a rather quick search
13:19:12kugelTopy44: oh, I'm still at "looking for front.jpg additionally"
13:19:14bertrikso, what exactly has been decided on now? or haven't we decided about anything yet?
13:19:23Topy44kugel: ah, ic
13:19:25Lloreanbertrik: Add front.jpg and album.jpg I think
13:19:44Lloreankugel: I'd say a full directory scan and decision making process is hardly insignificant *relative* to what we already do.
13:19:50bertrikLlorean, ok
13:20:10LloreanIt may take a very short time, but it's still a lot longer than just checking for a short list of filenames if the folder it's going through is big.
13:20:37kugelLlorean: it is. a directly scan is relatively cheap, especially with dircache. It's just a FAT operation, and has nothing to do with looking at the real folder
13:20:48kugeldirectory scan*
13:20:48Topy44Llorean: btw, did that rudy4pez guy mention any software that uses album.jpg? (none in my music collection btw)
13:20:49bertrikI am in favour of adding front.jpg but have doubts about adding a more elaborate heuristic scan
13:21:01Lloreankugel: Do you know what "relative to" means?
13:21:08Topy44(correction: found ONE :)
13:21:49Lloreankugel: So you don't think it could take two or even three times as much time if this were added as it does now?
13:22:16Topy44Llorean: the question is: is "much longer" any significant amount of time/cycles?
13:22:31kugelif we're just talking about looking adding 1 or 2 fixed filenames to the list, then yes
13:22:31Topy44or is it still "so quick it doesnt matter"
13:22:46Topy44kugel: we are not, already decided on that :)
13:22:46Lloreankugel: We're talking about the search
13:22:52Lloreankugel: I already told you that once.
13:23:09LloreanIt needs to find all .jpg, .jpeg, and .bmp files, then decide which one is most likely the cover via some algorithm.
13:23:10linuxstbbertrik: OK, I'll try and clean up what I've got so far, and commit it anyway. It's mostly just adding stubs, plus implementing and some changes to crt0.S
13:23:35kugelLlorean: I don't think we want that
13:23:58bertriklinuxstb, fine with me. If something breaks, I'll yell :)
13:24:05linuxstbbertrik: I'm sure you will!e
13:24:07Lloreankugel: Yes, as I said, it would take much more time (relative to the amount we already spend) and serve a relatively small additional group of users (people who don't already use their album art in some PC program)
13:24:11Topy44kugel: and thats what the whole discussion is about really: does anyone want that :)
13:24:15*linuxstb didn't mean to type that extra e... ;)
13:24:32LloreanTopy44: I *thought* I knew a program that used album.jpg but on second look it doesn't, so maybe not.
13:24:41BdN3504quick question: is the manual compeiled with files that are only found in the manual directory of the source tree? can i copy that folder onto another machine and then still be able to build the manual?
13:24:54BdN3504compeiled -e =compiled
13:25:32 Join ehntoo [0] (
13:26:01Topy44completely irrelevant if you follow Llorean's argumentation, but still interesting:
13:26:54GeekShadowbertrik, can rockbox meizu access to files yet ?
13:27:26bertrikGeekShadow, no, that's still quite far off in my opinion
13:27:52kugelnot sure I have ever seen album.jpg
13:28:17markunbertrik: and the code in openiboot is not compatible with our FTLs?
13:28:23linuxstbbertrik: Is the keymap going to be straightforward on your Samsung target?
13:29:26 Quit GeekShadow ("The cake is a lie !")
13:29:52bertrikmarkun, for the samsung yp-s3 I looked a bit closer and it didn't seem to be compatible with the openiboot FTL, I haven't looked that close at it for the meizus yet
13:30:41linuxstbbertrik: Was it the samsung where you wrote some initial NAND code to read the ID?
13:30:42bertriklinuxstb, yes, I think so, it has pretty normal up/down/left/right/select buttons, a "menu" button and a "back" button
13:31:08bertrik(I meant the samsung yp-s3)
13:31:28markunbertrik: do you think we should use some other FTL and forget about the OF?
13:31:50bertriklinuxstb, yes that code is somewhat specific for the samsung yp-s3, the specific thing is the way the NAND chip select is controlled
13:32:32linuxstbbertrik: Ah, OK. But presumably, now that's working, reading/writing NAND pages should be relatively straightforward?
13:33:44bertriklinuxstb, yes I think so, I got the read id routine from the OF, but it's very similar to the one in the data sheet, the data sheet also has sequences for reading the NAND
13:33:59bertrikI wouldn't dare to erase/write NAND pages yet
13:34:46bertrikmarkun, for the short term I hope have least read-only access to the FTL of the OF
13:35:44 Quit HBK- (Read error: 104 (Connection reset by peer))
13:35:47 Join KBH [0] (
13:36:48bertrikmaybe in the long term we could have our own rockbox flash file system (and access it over MTP for example)
13:37:02bertrikI'm still not sure what would be best / easiest / possible at all
13:43:31linuxstbWell, I've always seen Rockbox as a _replacement_ firmware...
13:45:50linuxstbAnd speaking personally, I never use OFs, so our own FTL would fully meet my needs. But of course, it would make Rockbox far less interesting for casual users.
13:46:23 Quit gevaerts (Nick collision from services.)
13:46:35 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
13:46:59 Join AndyIL [0] (i=AndyI@
13:59:08 Join GeekShadow [0] (n=Antoine@reactos/tester/GeekShadow)
14:01:00 Quit AndyI (Read error: 110 (Connection timed out))
14:10:18bertriklinuxstb, what do you think about implementing an FTL vs. using a flash file system without any FTL at all?
14:10:52bertrikwith an FTL we could still be using fat and access it over USB MSC
14:11:08Torneit's *probably* less effort to implement an FTL, also
14:11:18Torneespecially if you nick the algorithm from linux
14:11:21Torne(or similar)
14:11:24bertrika direct flash file system leaves out the FTL layer, but probably requires MTP support over USB to access the file system
14:11:37Torne(versus trying to port jffs/ubifs/etc or *shudder* actually design one)
14:11:58bertrikI don't have any ambition to design my own :)
14:12:24bertrikmaybe adapt an existing one
14:12:27 Quit ehntoo (Read error: 110 (Connection timed out))
14:14:34Tornei think using an actual flash fs would be a long way off
14:14:46Tornethe advantages are not so big probably when you're tlaking about a DAP
14:15:11Tornean FTL is not so bad for a filesystem you mostly read, and when you do write you tend to create/delete entire files.
14:17:32BdN3504have i done something wrong, or is a description of viewports really missing in the manual?
14:18:07bertrikTorne, indeed I think a typical disk access scheme is to write several large files on it over USB (e.g. copy mp3s to it), read those files from within rockbox (playing music) and write smallish files to it from within rockbox itself (e.g. updating settings)
14:18:25BdN3504do i really have to write it myself?
14:18:49BdN3504or can i just point a reference to the wiki?
14:18:55TorneYah, writing large files to a minimally fragmented disk works pretty well in an FTL
14:19:04BdN3504i think i'll go that route
14:19:10Torneyou tend to get the case where you are in fact writing a whole erase block sequentially quite often.
14:19:17Torneas long as the FTL has enough buffering
14:19:29Torneupdating settings and so on is probably little enough that we can ignore it
14:19:40Torneif you were really concerned, actually..
14:19:41bertrikTorne, do you have any suggestion for an FTL we could port?
14:19:47Torneyou could reserve some flash blocks to be the nvram
14:19:55Torneand write that to flash as raw sectors without the ftl
14:20:02Torne(you'd need a versioning mechanism or something but that'd be fine)
14:20:30TorneI'd suggest you just look in linux-mtd and nick whatever seems to be suitable for the flash type you are working with, if you really didn't care about OF compatibility
14:20:41TorneI'm not sure what the state of the art is there, though
14:20:49TorneI have onlt really worked with patented proprietary FTLs
14:21:10Torneand with flash filesystems directly. Never actually used the linux ftls.
14:22:24 Join LambdaCalculus37 [0] (i=44a0430d@rockbox/staff/LambdaCalculus37)
14:24:26bertrikI read about ONFI, open-nand-flash-interface. It looks like this is the route future flash-based DAPs will take for their NAND memories (and probably some current ones already do)
14:25:17bertrikit would be nice if we could find some FTL that offers a block interface on top and uses an ONFI interface at the bottom
14:26:30TorneThe actual details of how you talk to the nand are kinda boring and a minor porting detail, tbh :)
14:27:04Torneit's minor effort to change the actual calls you make to the storage driver slightly, at most
14:27:29Tornethe important logic in the ftl only depends on what the subpage/page/eraseblock sizes are, and whether there are page write order restrictions
14:29:22bertrikoh, I read some things about copyback for example that is a bit more advanced than just read/write/erase
14:30:42BdN3504where do i find the names of the targets in the manual? for example RECORDER_PAD, to which models does this apply?
14:30:55Tornebertrik: copyback is just a smarter way of doing read-modify-write.
14:30:58Tornethe cycle is the same
14:31:43Torneyou erase the target block, then use copyback for an internal data move of the pages that haven't changed
14:31:50Torneand you only issue actual writes for the pages that you are changing
14:32:06Tornes'just quicker and a little safer
14:39:40bertrik*someone* should put this info (where applicable to rockbox) in a more organised form on the wiki ... :) have the logs :)
14:46:24CIA-61New commit by mcuelenaere (r22375): Skin engine: add "pitch" action to touch regions
14:47:46BdN3504how can i differentiate in the manual between the different ipod models? are they not defined?
14:48:39BdN3504i am looking for the \opt{ command for 1g 2g 3g 4g 5g and 5.5g
14:48:50CIA-61New commit by mcuelenaere (r22376): Enable the pitch action in the M:Robe 500 Cabbiev2 WPS
14:49:09 Part LinusN
14:51:39bertrikTorne, on I found this statement "Unfortunately it is a rather difficult task to create a good FTL layer and nobody still managed to implement one for Linux. But now when we have UBI (see here) it is much easier to do it on top of UBI."
14:52:09bertrikI guess that means we can't that easily nick something from linux
14:52:54Tornethere are at least two FTLs in linux
14:54:47Tornethey may not count as *good*
14:54:53Torneand i dunno that any of them run on MLC NAND
14:54:57Tornebut they're there :)
14:55:32CIA-61New commit by mcuelenaere (r22377): Onda VX747: don't enable software volume control in sims
14:57:41bertrikmeh, easier said than done, I see warnings about not being allowed to use the FTLs on anything other than PCMCIA hardware ("FTL)" or on diskonchip hardware ("NFTL") for example
14:57:58TorneThat would be patents
14:58:03Torneas we have previously discussed
14:58:25Torneit is very likely that any FTL you port or implement is covered by someone's patent somewhere
14:58:25BdN3504is there a way to \opt different screensizes in the manual?
15:01:43bertrikwell, *I* didnt make up that statement from 14:51, I just quoted it and it seems fairly recent and from the linux mtd people
15:02:53Tornebertrik: Yah, i know
15:02:59TorneI don't know what they mean, anyway.
15:03:08linuxstbBdN3504: I don't believe so. It's been discussed, but I don't think anyone has implemented it.
15:03:10TorneThere are some FTLs but they are possibly crappy and probably patent-encumbered.
15:03:24Tornehowever any other ftl someone might implement is probably patent infringing as well
15:03:43Tornebut of course they aren't going to look into that, because of the whole thing with knowing infringement in the US :)
15:05:04bertrikok, I appreciate your knowledge on this subject
15:05:22Torneyou are rapidly exhausting my knowledge on this subject :)
15:05:28Tornea broad introduction is about the best i can do
15:06:46bertrikat least to me, this whole FTL thing makes the rest of the samsung-based players porting effort pale in comparison ...
15:07:03Torneit's not something that's come up before i guess
15:08:17CIA-61New commit by mcuelenaere (r22378): * Onda VX747: add browse screen, pitchscreen, context menu, quickscreen, rewind, fast forward, previous song & next song actions to cabbiev2 ...
15:10:15 Join Ubuntuxer [0] (
15:15:17*kugel has a updating progressbar in the main menu :D
15:16:06***Saving seen data "./dancer.seen"
15:16:58bertrikdon't we have the WPS for that?
15:17:29bertrikwhat kind of progress does this bar show?
15:17:56kugelplayback progres
15:18:18kugelbertrik: I'm playing around a bit with a skin'ified statusbar
15:19:26kugelthat can use all wps tags
15:22:18 Join muesli [0] (
15:22:41mueslia h120 user arround?
15:22:46mueslior 140
15:23:21linuxstbmuesli: Best to just ask the question.
15:24:34mueslijust plugged in the charger (using pre-bootloader 7pre4) and the green light is on. does it turn off when its done? (havent had a h120 for ages and cant remember)
15:26:31linuxstbI seem to recall it does, but haven't used my h120 for ages either. Best to wait for someone with more recent experience...
15:26:38 Join evilnick [0] (i=0c140464@gateway/web/freenode/x-chelptnnsjieyteb)
15:27:29 Quit kugel (Read error: 60 (Operation timed out))
15:27:32muesliok...but was it normal that the screen remains empty? can remember that there was a charging sign?
15:29:11moosIIRC yes that's normal, the only indication is the green light
15:29:51moosbtw /me remenbers an old muesli user :)
15:30:22muesliyepp, thats me ;)
15:30:46mooshola then, and welcome back to rockbox usage ;)
15:31:08mueslihola moos :-)
15:31:26mueslinever gave up using rockbox...was just happy with it and no question left ;)
15:33:10mooswhen you use rockbox for years, hard to use any OF after ;)
15:34:14muesliexactly :)
15:34:38mueslithats why i didnt consider any other player than the old good irivers
15:35:44 Quit KBH (Read error: 104 (Connection reset by peer))
15:35:53 Join KBH [0] (
15:37:07 Nick martian67_ is now known as martian67 (n=martian6@about/linux/regular/martian67)
15:37:34muesliim only desperatly waiting for audible seeking :o
15:38:22moosmuesli: imho, don't count on it ;)
15:39:14mueslibooh ;(
15:40:56mueslii remember some pre-recording setting. was is it that way, that you could set up ie 2minutes until it records?
15:42:09moosyes you can
15:42:31moosan you can even check the fine manual we have now :)
15:42:42mueslii hate manuals ;)
15:42:55CIA-61New commit by mcuelenaere (r22379): Utils/Analysis/ fix wrongly recognizing addresses as being in codec or plugin space when their addresses were invalid
15:52:38 Join jgarvey [0] (
16:03:52 Part Ubuntuxer
16:06:40 Quit moos ("Rockbox rules the DAP world")
16:06:54 Join z35 [0] (
16:10:01 Join froggyman [0] (
16:11:30 Nick alexbobp_ is now known as alexbobp (
16:18:16 Join fml [0] (n=4fd3c7d3@gateway/web/cgi-irc/
16:19:03mueslifor all who care: yes, the green light turns off when its done
16:19:15fmlmuesli: IIRC the green LED is just an indication that the player is connected to the charger. It's not an indication that the battery is really being charged, i.e. not full yet.
16:19:50fmlmuesli: Ah, I read your question and just responded, but wrongly!
16:20:24fmlI didn't know that the greed LED is that "intelligent"!
16:22:12 Join polobricolo [0] (n=polobric@
16:23:20mueslilets c if it is ;)
16:24:21amiconnThe green led *is* a charging indicator (actually *the* charging indicator - you cannot read charging status from the CPU, only voltage)
16:25:41fml kugel: does this mean that in order to always see e.g. the current song's title (i.e. even when in menu etc.) we'd have to stuff all the information into the status bar?
16:26:21fmlamiconn: but it's not the "not full" indicator, right? Just the "connected" one.
16:34:05 Quit Kopfgeldjaeger (Read error: 60 (Operation timed out))
16:37:03 Quit mt (Read error: 113 (No route to host))
16:39:05 Quit AndyIL ()
16:40:12 Join Polo [0] (n=polobric@
16:41:05 Quit polobricolo (Read error: 113 (No route to host))
16:41:38 Nick Polo is now known as polobricolo (n=polobric@
16:43:21 Join toffe82 [0] (n=chatzill@
16:47:02 Join AndyI [0] (i=AndyI@
16:56:43 Quit polobricolo (Read error: 104 (Connection reset by peer))
17:00:40 Quit muesli ()
17:01:14 Quit thegeek ("( :: NoNameScript 4.2 :: )")
17:04:15 Join Lss [0] (
17:06:05 Join mt [0] (n=MTee@rockbox/developer/mt)
17:09:00amiconnfml: It *is* the "not full" indicator
17:14:46 Join polobricolo [0] (n=polobric@
17:16:07***Saving seen data "./dancer.seen"
17:17:24 Quit polobricolo (Read error: 104 (Connection reset by peer))
17:17:42 Join froggyman_ [0] (
17:18:08 Quit froggyman_ (Client Quit)
17:18:28 Join webguest83 [0] (n=47ba09e0@gateway/web/cgi-irc/
17:19:10 Quit webguest83 (Client Quit)
17:19:26 Join froggyman_ [0] (
17:20:34 Quit froggyman_ (Client Quit)
17:20:53 Join froggyman_ [0] (
17:21:46 Quit froggyman (Nick collision from services.)
17:21:49 Nick froggyman_ is now known as froggyman (
17:28:01 Quit Grahack (Read error: 104 (Connection reset by peer))
17:30:01 Join Grahack [0] (
17:30:06CIA-61New commit by wincent (r22380): PDBox: Made GUI use default back- and foreground colors.
17:36:02 Quit KBH (Read error: 104 (Connection reset by peer))
17:36:06 Join HBK- [0] (
17:39:11CIA-61New commit by wincent (r22381): PDBox: Simplified some float math functions.
17:45:14 Quit ch4os (Read error: 104 (Connection reset by peer))
18:00:30 Quit bertrik (Remote closed the connection)
18:05:17 Join jyu [0] (i=2669ecc2@gateway/web/freenode/x-bqizykhzhtxifbjt)
18:05:36 Join FOAD_ [0] (n=dok@
18:06:57fmlamiconn: nice to know!
18:07:58 Quit petur ("work->home")
18:09:03 Join faemir [0] (n=faemir@
18:09:37 Quit avacore^ (Read error: 110 (Connection timed out))
18:10:28 Join bertrik [0] (
18:10:50 Quit Xerion (" ")
18:14:44 Join polobricolo [0] (n=polobric@
18:18:47mtamiconn: This has ~0.5% speedup on PP502x, Could you test for CF ?
18:18:55mtSorry, This :
18:19:33 Nick jyu is now known as CaptainKwel (i=2669ecc2@gateway/web/freenode/x-bqizykhzhtxifbjt)
18:22:09 Quit FOAD (Read error: 110 (Connection timed out))
18:22:09 Nick FOAD_ is now known as FOAD (n=dok@
18:22:37 Quit polobricolo ("Good bye")
18:29:44 Quit bertrik (Read error: 104 (Connection reset by peer))
18:30:28 Quit Marvin_ ("CGI:IRC (EOF)")
18:30:37 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
18:31:44 Quit robin0800 (Remote closed the connection)
18:33:02 Join Xerion [0] (
18:45:03 Quit fml ("CGI:IRC (EOF)")
18:57:59BdN3504commit this please.
18:59:20linuxstbBdN3504: Why not simply link to ?
18:59:46 Join ch4os [0] (n=ch4os@gentoo/user/ch4os)
18:59:54linuxstba) it won't need updating with new targets; b) it won't need updating if the theme site changes structure
19:04:24BdN3504linuxstb: wtf. it was like that in the old manual, i simply updated it and took an awful lot of time because a) no one helped when i literally yelled, and b) because i kept the order the manual download page lists the targets.
19:04:35BdN3504don't tell me this is no improvement!!!
19:05:31BdN3504and as the new targets won't get comitted all at the same time, i don't think it's much of a fuzz to add a line which i have prepared and replace the names.
19:05:40BdN3504links repectively
19:07:28 Quit froggyman ("ChatZilla 0.9.85 [Firefox 3.5.2/20090729225027]")
19:08:18 Quit AndyI ()
19:09:58linuxstbBdN3504: I think the fact that that list was so out of date (e.g. no links to the Sansa e200 or Gigabeat wps galleries) shows that things like this get forgotten. Hence my suggestion to just have a generic link that will never get out of date.
19:13:17 Join kugel [0] (n=kugel@rockbox/developer/kugel)
19:14:31 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
19:14:58BdN3504me worked much on this. find suggestion less convenient because must click each target respectively while my solution tak to gallery respictve target directly.
19:15:56BdN3504say sentence befroe last sentence, if you read: target adding not much happening, so list no much get update
19:16:11***Saving seen data "./dancer.seen"
19:17:54kugelfml: yes
19:20:23kugelBdN3504: I also think linuxstb's suggestion is more useful
19:21:53BdN3504aaaah whole rb dev against me. can assign task to me. will keep up to date until end of time.
19:24:42linuxstbBdN3504: Please don't take it personally - I'm not against you, I'm just in favour of simple solutions.
19:24:50 Join JdGordon| [0] (n=Miranda@
19:25:52kugelJdGordon|: hey, I hacked a skin'ified statusbar together. Works well, except that images aren't loaded (find_image fails)
19:26:21JdGordon|why dont the images load?
19:26:29JdGordon|probably because the folder is wrong
19:26:49JdGordon|the bmp folder i mean
19:27:08kugelJdGordon|: no they are parsed fine
19:27:10 Quit BdN3504 ("CGI:IRC (Ping timeout)")
19:27:21JdGordon|loadeding happens after parsing
19:27:28 Join kachna [0] (
19:27:30kugelI didn't change the folder, my .sb is in .rockbox/wps
19:27:41JdGordon|did you fix token_value() or whatver than damn funciotn is called so it doesnt return if id3==NULL?
19:28:06kugelno, the statusbar doesn't show up before music was played as of now :)
19:28:28linuxstbBdN3504: What are you referring to when you say " no one helped when i literally yelled" ?
19:29:05kugelbut it really can't manage to find a single image (it seems the id is off by one actually), which is frustrating because it works so properly besides that (and the id3 == NULL thing)
19:29:29JdGordon|kugel: yeah, thats something which needs to be fixed... not being able to work if no music is playing is just bad :)
19:30:01JdGordon|I assume its a pretty big ram hit adding another set of gui_wps sturcts?
19:30:08 Quit einhirn ("Miranda IM! Smaller, Faster, Easier.")
19:30:27kugelyea, probably, that's why you're supposed to convert the arrays into the skin_buffer quickly ;)
19:30:48JdGordon|I'll go at the same pace as custom ui :)
19:30:54JdGordon|except no, i want to get this done
19:31:00JdGordon|viewports tonight hopefully
19:31:00kugelahh no, don't
19:32:03JdGordon|the viewport one is the last "easy" one I think... the lines, sublines and tokens will need some thinking about
19:32:45JdGordon|I'm tihnking that for tokens at least they will be allocated on init to some value, large for wps, small(er) for sb
19:32:58JdGordon|like, 200 tokens for sb sounds reasonable
19:34:03kugelvery, but why the limiting?
19:35:13pixelmalinuxstb: are you sure the database's track guessing was removed?
19:35:20kugelso, you don't have any idea why images work fine from the .wps, and not for the .sb (the .wps and .sb are exactly the same, except no backdrop for the .sb)
19:35:45JdGordon|kugel: because making a linked list of tokens is probably going to be more wasteful than just doing a boring array allocation
19:36:03 Quit HBK- (Read error: 104 (Connection reset by peer))
19:36:07 Join KBH [0] (
19:36:08JdGordon|kugel: they are named the same also exept the extension?
19:36:27JdGordon|whip out gdb :)
19:38:03JdGordon|anyone have any thoughts about allocating tokens/lines/sublines in blocks of 15 or so and have them in linked lists?
19:38:25linuxstbpixelma: Err, no... Are you saying it still does it?
19:38:31kugelthe id seems off by 1, for the wps e.g. n=26 manages to match, but for the .sb it doesn't because max id is 25. For the next image all numbers are +1
19:40:34JdGordon|the image id thing is very confusing..... feel like changing it to store the char instead of an int in svn?
19:40:50 Join Omlet [0] (
19:40:53 Join moos [0] (i=mustapha@rockbox/staff/moos)
19:41:04kugelhow's that useful?
19:41:43JdGordon|figuring out why image 41 didnt load is harder than 'E'
19:42:01JdGordon|also, longer term.. I want to change the image id's to strings and subimage numbers...
19:42:05pixelmalinuxstb: I just can't remember a "remove track number guessing" commit since the time I'm aware of it but could have missed something. In case it is still in, I'm in favour of removing it
19:42:13JdGordon|completly remove the limits
19:42:32pixelmalinuxstb: need to find something I could test with
19:43:01kugelI don't think chars for the *internal* representation are going to make it easier
19:43:33JdGordon|in gdb it will :)
19:43:41 Join bertrik [0] (
19:44:31*JdGordon| wonders if Jeff Goode would be interested in doing the software mixer idea from ages ago to fix the no-talking-while-paused problem
19:45:21kugelJdGordon|: certainly if you take his eyes out
19:45:38linuxstbpixelma: Hmm, looking at the changelog for tagcache.c, it looks like you're right. Unless it was done elsewhere.
19:46:22*linuxstb edits his post
19:46:30*pixelma pings Slasheri maybe he can tell (and is around somewhat)
19:47:18 Join Blue_Dude [0] (
19:47:35linuxstbpixelma: Yes, looking at the code, it still does tracknumber guessing...
19:48:01pixelmamight give weird results on Ipods :\
19:48:06Blue_DudeJdGordon|: you rang?
19:48:11JdGordon|oh hey
19:48:15pixelmaand somwhere else too
19:48:45bertrikJdGordon, wasn't this also the difference between the paused and stopped state (one allows talking, the other does not)?
19:48:50Blue_DudeJdGordon|: What's broke again? Voice won't play while paused?
19:49:03JdGordon|bertrik: yes
19:49:26kugelBlue_Dude: that was always broken
19:49:28JdGordon|Blue_Dude: yeah, because voice + audio is mixed far too early
19:49:56JdGordon|there has been an idea for ages to keep them seperate and only mix them right ebfore thye get sent to the DAC
19:50:27 Nick Omlet is now known as Omlet_Away (
19:50:36JdGordon|I tinhk jhmikes has a better idea of what was needed... he's not around thoguh :/
19:50:39Blue_DudeJdGordon|: sounds promising, but it means pcmbuf and playback would need major rewrites.
19:50:52JdGordon|doesnt that sound like fun? :)
19:51:01JdGordon|why playback?
19:51:10Blue_DudeSure, I'll knock it out this weekend. :)
19:51:38JdGordon|if its not too much trouble :)
19:51:41Blue_DudePlayback controls the low level pcm state. If it pauses, then nothing gets out.
19:51:41linuxstbpixelma: Unless I'm misunderstanding, the code seems strange. IIUC, it searches backwards from the "." part of the filename, and looks for two consecutive digits. If it finds them, they're used for the track number.
19:51:57linuxstb(exactly two digits)
19:52:10JdGordon|Blue_Dude: ok, although I would think that would be the easy bit of the whole thing to fix :)
19:52:21kugelBlue_Dude: playback wouldn't be involved if voice is seperate
19:53:14kugelI was thinking the voice wouldn't even reach the pcm buffer, but get a seperate one, voice and pcmbuffer then get mixed and sent to the hw
19:53:17Blue_Dudekugel: problem is that if playback forces the pcm state to pause, then nothing will get out.
19:53:27 Quit bertrik (Read error: 104 (Connection reset by peer))
19:54:02Blue_DudeIn which case playback will need a rewrite to only control the playback buffer and not the PCM output buffer.
19:54:23JdGordon|the very high level idea was to have a seperate channel for each audio source (voice, crossfeed, audio) which could be individually pasued or played, and mixed at the last moment
19:54:33pixelmalinuxstb: I remember reading a similar description in the forums where the track number guessing was discussed (someone wondered where a weird track number was coming from)
19:54:38 Join efyx_ [0] (
19:54:44Blue_DudeIf you want to mix that late in the signal chain then a lot of assumptions on hardware state get thrown out the window. that good or bad?
19:55:20Blue_DudeIt's good, but it means more than tweaking pcmbuf.
19:55:39martian67bad if you dislike bug squishing ;)
19:56:02kugelmoos: it works when you load it in the settings menu?
19:56:41moosnegative, it doesn't. It fails. Let me share a zip
19:57:09 Join bertrik [0] (
19:58:00linuxstbpixelma: From 4th April 2006 - "22.01.34 # <Slasheri> amiconn: hmm, true. that might work better. I will consider dropping that code" (I guess he never did...)
19:58:22JdGordon|damn irc logs wins again!
19:58:42 Join avacore [0] (
19:58:56pixelmalinuxstb: he just "considered" it ;)
19:59:29 Quit bertrik (Remote closed the connection)
20:01:08 Quit efyx_ (Remote closed the connection)
20:01:23mooshere it is:
20:02:14Blue_DudeWe'd need another layer of abstraction, one that is interposed between playback/pcmbuf and pcm.c. It could be called mixer.c and it would have sole authority over hardware states. It would also need to take over the PCM buffer, getting its input from the playback buffer (what is now called the PCM buffer) and the voice buffer.
20:02:24kugelmediafire is crap :(
20:02:59JdGordon|Blue_Dude: yay! more abstraction!
20:03:02JdGordon|sounds like fun
20:03:43Blue_DudeIt could actually be simpler this way. pcmbuf and playback are hybrids. Maybe they could act more like APIs.
20:03:59mooskugel: where do you ant it?
20:04:08JdGordon|Blue_Dude: have a tihnk about it :) I just figured it would be good to get that idea into your head... I dont tinhk anyone else is interested in dsp-ish sort of stuff which i guess this is close to
20:04:14kugelmoos: nevermind, worked now
20:04:24moossorry :)
20:04:34 Join efyx_ [0] (
20:04:53JdGordon|moos: unless kugel finds the issue, can you email that to me? I'm at work so cant really do anything about it for a few hours
20:05:09amiconnI'd just mix voice in the isr that handles pcm, just a few samples at a time
20:05:36moosJdGordon: sure. Just 2sec... (the .gmail one, right?)
20:06:06linuxstbamiconn: On targets using audio DMA, isn't that about 32KB? Or could/would that size be reduced?
20:07:22JdGordon|although would doing late mixing make it harder to shrink the PCM buffer on low ram targets like the clip?
20:07:29JdGordon|and flash targets
20:08:21moosJdGordon: done
20:09:32moosJdGordon: Thanks to look at it
20:09:49Blue_DudeJdGordon|: well, really what's the benefit of having a huge output buffer? Why not whittle it down on all targets to something small, and keep the rest for music playback?
20:10:37*JdGordon| tihnks he is getting confused
20:11:05kugelBlue_Dude: it saves boosting
20:11:59Blue_Dudekugel: yes, but do you need to boost to keep the buffer from underrunning, or to process input?
20:12:00kugelnot on avarage, but the process of boosting alone isn't free on many so few long boosts are better than lots of short ones
20:12:21moosBlue_Dude: on the DSP, no-do list, with have "low latency" too. It seems to not be easy task /me hides :)
20:12:48kugelthe pcmbuffer would underrun on the majority of targets and codecs without boosting at refilling yes
20:13:09 Quit efyx_ (Remote closed the connection)
20:13:37 Join bluebrother [0] (n=dom@rockbox/developer/bluebrother)
20:14:52Blue_Dudekugel: what I mean is, you can have two buffers here. One is a small one that is read by the hardware for output. Another one, much larger, is kept full with occasional boosting and represents playback content.
20:15:07Blue_DudeStill involves an extra memcpy though...
20:15:11 Join bertrik [0] (
20:16:07Blue_Dudemoos: low latency would be handled no differently than it is now.
20:16:42 Join saratoga [0] (n=9803c6dd@gateway/web/cgi-irc/
20:17:06pixelmaBlue_Dude: would your suggestion affect buffering the voice file?
20:17:09 Join Strife89 [0] (n=michael@
20:17:47Blue_Dudepixelma: it already is. I think it could just be mixed in later than it is now.
20:17:51saratogai've emailed Buschel asking him how the heck you do an inverse quadrature mirror filter in less then a 100 MHz
20:18:19saratogasince he wrote one subband codec and once helped me optimize another, he will hopefully know this
20:19:11Blue_DudeThe memory structure now is something like |−−−−−−−−−−−−−−−−PCM−−−−−−−−−−−−−−−−−−|CF|Vc|everything else|
20:19:43pixelmaBlue_Dude: I mean, would you have to spin up to load other voice clips? (not sure how it is handled on swcodec currently though)
20:19:49Blue_DudeWhy not |−−output−−|−−−−−−−−−−−−−−−−−−playback−−−−−−−−−−−−−−-|CF|Vc|everything|
20:20:02kugelyea, why not?
20:20:16JdGordon|because codecs arent all >100% realtime without boost
20:20:21Blue_Dudepixelma: dunno.
20:20:23JdGordon|so you have to have some PCM buffer
20:20:50Blue_DudeJdGordon|: yeah, but they'd boost to fill the playback buffer, not the hardware one.
20:21:16kugelman, just do it! :p
20:21:18JdGordon|playback is compressed audio or pcm?
20:21:36SalineHi all. I'm doing my best to rape the plugin api. I'm calling a plugin (i've made reentrant) from screens.c, and get the following error message when compiling: ".text_0xfoo): undefined reference to
20:21:41 Quit elcan (Read error: 110 (Connection timed out))
20:21:45JdGordon|ah ok, that would work on flash, but disk targets would be spinning up too often
20:21:54Blue_DudeSame as now, but instead of going right out, it's waiting to be mixed.
20:22:14JdGordon|alot more compressed sits in the buffer than pcm (obviously :) )
20:22:29kugelwhat else?
20:22:38Salinemy guess is that the structure isn't initialized somewhere, but i'd better ask for your help..
20:23:07kugelMany stuff is packed into the audio buffer, I thought the pcm is for pcm (>decoded audio) only?
20:23:29Blue_DudeWell, codecs get their own input buffer. That wouldn't be affected. I had in mind splitting the current PCM buffer instead of taking from somewhere else.
20:24:07kugelI think that sounds good
20:24:16amiconnlinuxstb: DMA chunks would need to be made smaller if that's the case
20:24:48Blue_DudeNo idea how it would work in practive though. There's some overhead involved in mixing, but maybe that would wash out since filling the playback PCM buffer would be simpler.
20:25:13amiconnWith DMA, the DMA end isr could do the mixing, directly into a small DMA buffer, e.g. the equivalent of 10ms (1 tick), i.e. 441 samples when playing at standard frequency
20:25:33amiconnOn ARM, the FIQ routine could do the mixing on the fly
20:26:14Blue_Dudeamiconn: that's kind of what I was thinking. Something really small that would be easy to keep filled without a lot of overhead.
20:26:52Blue_DudeIf all you need to do is mix two samples, and then copy the result to a memory location, that's about as easy as it gets.
20:27:22JdGordon|it could be more than 2 samples...
20:27:43Blue_DudeWhat else would you like?
20:27:51Blue_DudePlayback, voice...
20:28:09amiconnCrossfade is done way earlier
20:28:28Blue_DudeCrossfade would be handled in playback.
20:28:32amiconnIt doesn't need to be realtime, unlike voice
20:29:02amiconnThe two streams need to be decoded sequentially anyway, because only a single codec can be active at a time
20:29:22JdGordon|oh.. another thing which would be nice is DSP controlled volume fading instead of using the output volume the way its done now
20:29:37amiconnMore ifdefs?
20:30:08Blue_DudeOutput volume is handled in hardware, yes? What needs to be fixed there?
20:30:52JdGordon|fading currently doesnt work with line out
20:30:57amiconnFade on stop/pause is done using master volume. This has the advantage that it works for both hwcodec and swcodec
20:31:14amiconnJdGordon|: That's target dependent; you can't say that in general
20:31:33Blue_DudeThat's so late in the signal chain that it would affect voice, too, no matter where you mixed it in.
20:32:23 Join RyoS [0] (
20:32:31RyoShi all :)
20:32:55amiconnThat's true, but then this fade isn't very long
20:33:07RyoSim in search for an collection (zip/tar) of anti aliased fonts
20:33:34RyoSif there is something like that? i use (test) the patch kugel posted on flyspray :)
20:33:36JdGordon|you're apparently lost then
20:34:03Blue_Dudeamiconn: maybe fade could be inhibited if voice is currently being mixed in. That seems reasonable.
20:34:33amiconnThere are some advantages for doing the fading in pcm, but that at least requires splitting hwcodec and swcodec here
20:34:48JdGordon|with really late mixing you could do fun stuff like makign the audio quieter if voice is happenign at the same time yeah?
20:35:20amiconnAlso, this fade needs to be kinda realtime, like voice, and unlike crossfade.
20:35:23JdGordon|or is that done already?
20:35:31Blue_DudeJdGordon|: in fact, that would be required unless you wanted to clip the output.
20:35:36amiconnRockbox cannot know in advancce when the user presses pause or stop
20:37:22Blue_Dudeamiconn: please refresh my memory. All this affects only swcodec targets, correct? hwcodec targets already do their own thing?
20:37:49amiconnHwcodec cannot mix music and voice, since there's only one decoder
20:38:23amiconnSo voice is simply muted while music is playing. No crossfade as well, for the same reason
20:38:31Blue_Dudeamiconn: I have to experience with hwcodec targets, so I don't know how they're implemented.
20:38:39amiconnFade on stop/pause works on them, because it uses master volume
20:38:39 Quit bertrik (Read error: 104 (Connection reset by peer))
20:38:53Blue_DudeSo Rockbox is pretty much a front end on those targets?
20:39:14Blue_Dude−− I have *no* experience −−
20:39:46 Join Tristan53 [0] (i=tristan@
20:39:54Blue_DudeA front end, meaning mostly a new interface.
20:40:14amiconnRockbox is a complete replacement firmware on all targets it supports. On hwcodec just the actual audio decoding doesn't happen on the main cpu, but on a preprogrammed dsp
20:40:36pixelmaRockbox is a replacement firmware as it is on other targets too, just playback works differently due to different hardware systems
20:40:43*pixelma too slow
20:41:49pixelmaBlue_Dude: and Rockbox was first developed on (for) these targets
20:42:07Blue_DudeSure, Rockbox is a complete firmware replacement, but it can't do any signal manipulation because it doesn't have custody of the signal chain?
20:42:23*Blue_Dude needs to read up on his history.
20:42:38amiconnWhile the dsp is preprogrammed to decode mp2/mp3 (and on some of these targets, also encode mp3), it can also be reprogrammed
20:43:28amiconnThat's rather difficult because it's a custom dsp core with no public docs
20:43:58amiconnBut after several years, we got the pcm pass-through codec for it from archos+micronas with official permission to use it in rockbox
20:44:31amiconnThis is currently only implemented in plugins and not integrated into the main playback engine on hwcodec
20:44:43kugelmoos: the original free state works
20:44:45*amiconn has some ideas how this could be done
20:45:17Blue_DudeAh, that's where I'm going with it. Does the pcmbuf code need to address hwcodec targets?
20:46:04amiconnNo, that code is swcodec only, and will stay that way
20:46:26mooskugel: I slightly modified it, to fit my needs.
20:46:36kugelwhat did you change?
20:46:39amiconnHwcodec will get a different low-level module, but I'll try to reuse buffering (app level)
20:47:08 Quit flydutch ("/* empty */")
20:48:24mooskugel: I don't remenber well, but I think I just riped infos that I don't need
20:48:51Blue_Dudeamiconn: I guess that's the real question: should hwcodec targets be addressed in a pcm code rewrite to allow them to receive processed audio?
20:48:54 Quit Strife89 ("Goin' home momentarily.")
20:49:07Blue_Dudeok then
20:49:21kugelmoos: I think it lacks a 's' image
20:49:56kugelI see %xds, but no corresponding image
20:50:08Blue_DudeSo we're talking swcodec only. Which mean that files in apps/ are fair game, but not firmware/
20:50:19amiconnThe pcm code isn't used on hwcodec atm, and it won't be used in the future either. The MAS has several special needs which need to be catered for in the new low-level module (e.g. switching between pcm mode for waw/aiff playback and mpeg mode for mp2/mp3 playback)
20:50:59 Join bertrik [0] (
20:51:12amiconnDue to the CPU all data fed to the MAS needs to be bitswapped, that's another thing that will be handled there (and no longer in buffering)
20:51:41Blue_DudeJust trying to get an idea of the scope involved here.
20:51:55kugelmoos: it seems you've mostly taken out some images. You need to remove their references as well
20:54:42Blue_DudeMixer.c would have to do all the memory management too... Just thinking aloud...
20:55:29Blue_DudeOr maybe it would just be assigned a memory pool and it would divvy it up as it needs to.
20:55:59mooskugel: but why all was good before? I will anyway thanks
20:56:20kugelmaybe JdGordon| accidently fixed a bug :)
20:56:24JdGordon|moos: its been made slightly stricter apparently
20:56:30*JdGordon| slaps kugel
20:57:03moosaie :( then that will break a lot of wps then
20:57:33JdGordon|it will?
20:57:37pixelmawhy that?
20:58:10JdGordon|I do find it slightly humerous that the only people to complain about not working .wps are devs :)
20:58:22moosI think to people like me (not a wps author) that used a wps since a while and all worked and now no, then yes :)
20:59:21pixelmaI also didn't get all the "my WPS is broken" when the parser was made stricter wrt to escaping <|>
20:59:35pixelmathe rules were laid down before
20:59:36JdGordon|moos: would you prefer a hard crash (null pointer) or not loading wps?
21:00:20pixelmamoos: I don't think it will be many people who just delete the %xl tags
21:00:28pixelmaor some of them
21:00:51moossurely not. I don't complain, I just warn, that we could have few users encountering the same
21:01:11JdGordon|I wonder how hard a .wps verifier plugin would be on target
21:01:51moosBut sure that a more strict parser is nice at end...
21:07:33 Join elcan [0] (
21:08:54 Nick Omlet_Away is now known as Omlet (
21:09:50 Quit Omlet ()
21:09:55 Join TruthTaco [0] (
21:10:12kugelJdGordon|: you better think harder why the images don't work on my sb
21:10:28JdGordon|or else?
21:11:06JdGordon|get it going in gdb and find out where exactly its failing..
21:11:15JdGordon|is it the bmp_read_file()?
21:11:18JdGordon|or before that
21:11:48kugelI'm in skin_display already
21:12:24kugelevaluate_conditional() -> find_image
21:12:37RyoSkugel: do you have a collection of aff fonts? i am trying your patch on iaudio x5 :)
21:13:24kugelRyoS: nope. converting fonts is easy
21:13:29 Join merbanan [0] (
21:13:37RyoSmh, ok ;)
21:13:56RyoSfor; do; done, got it ;)
21:13:59kugelBlue_Dude: ...use the build system
21:14:54Blue_DudeThanks, Obi-Wan!
21:16:13***Saving seen data "./dancer.seen"
21:16:17Lloreankugel: Are you seriously encouraging someone not to bother testing, but just go ahead and break user builds?
21:16:37JdGordon|kugel: being at work i cant really help... have a look and make sure wps_data->images is getting populated correctly
21:16:48JdGordon|and yes, building all tagrtes locally is a huge waste of time
21:16:54JdGordon|the build system works well for that
21:17:09LloreanJdGordon|: Not all targets, but he *knows* it's likely to break targets with different size IRAM, so he should check a few of those at least
21:17:16Blue_DudeLlorean: I wasn't gonna commit without being reasonably sure it wasn't going to break something.
21:17:19LloreanLike the Gigabeat F
21:17:21kugelWhat's the point of doing it locally if there's 60 monsters just waiting for doing the job?
21:17:43pixelmaI don't like that attitude if you expect problems, do at least some test builds of a selection of different targets (like coldfire or arm here)
21:17:43JdGordon|moster is a bad word :) isnt that the 2nd slowest client :D
21:17:44kugelproducing a red build isn't a problem at all
21:17:47Lloreankugel: The point is if you know it's going to break something, you shouldn't push it out to users. The build system's other job is to make sure users have access to the latest binariers.
21:18:08JdGordon|if the build breaks the users cant get it..
21:18:21kugelLlorean: We're talking about red builds, not broken code (as in semantically broken), aren't we?
21:18:23LloreanJdGordon|: That's the point, it means we're distributing different versions to different users.
21:18:28kugelusers won't even get to download a red build
21:18:42Lloreankugel: Obviously not all the builds will be red, since it works in his local case.
21:19:09kugel"distributing different versions" that argument is just stupid
21:19:18LloreanAh yes, back to "discussion by insult"
21:20:20pixelmakugel: yours isn't even an argument as it doesn't give explanations just an opinion out of thin air
21:20:24kugelthe red is most likely fixed within 10 mins, what's the problem with that? That's so insignificant, why even bother with doing multiple builds
21:20:36Lloreankugel: Most likely based on what?
21:20:52*JdGordon| kicks kugel and Llorean to pm
21:20:52Blue_DudeLlorean: I will try a representative sample first. I just didn't want to sit on the patch forever for fear that one out of 60 builds would go red.
21:21:24JdGordon|the sample in the ml thread is a good one
21:21:27LloreanBlue_Dude: The obvious solution is to go to the -dev list with "On targets with small IRAM it probably won't build. Could someone help me with what these are?" or similar
21:21:41pixelmaJdGordon|: why that? Apparently you had to say something about it too
21:21:43LloreanBlue_Dude: You can ask for help in identifying where it might have problems.
21:21:56Blue_DudeLlorean: done that. Working on it ever as we speak.
21:21:58JdGordon|pixelma: because its a boring argument
21:22:20 Join Kopfgeldjaeger [0] (
21:22:45 Join funman [0] (n=fun@rockbox/developer/funman)
21:23:06LloreanBlue_Dude: But the build system makes builds immediately available to users, and if there are red builds that means a disparity between what some users download and other users download. We've gotten bug reports because a download didn't match what was supposed to be the current version before.
21:23:36 Join FOAD_ [0] (
21:24:12funmanthere is also a contest for the developer who'll make the most red points, but beating the high score isn't necessarily a good thing ;)
21:24:19Blue_DudeLlorean: No way I'm going to build 60 versions. I'll pick the ones most likely to have problems and build those. After that, I'll fix reds if they happen.
21:24:46JdGordon|Llorean: that arguemnt isnt really valid anymore... we except most users to use the releases
21:24:51pixelmaBlue_Dude: no-one expected you to test build everything
21:24:56LloreanBlue_Dude: I didn't say build 60 versions. But kugel's apparently been advocating "don't worry about it and let the build system show you"
21:24:57Blue_DudeOh good.
21:25:01LloreanWhich is very irresponsible toward the users.
21:25:04 Join Stephen__ [0] (
21:25:25kugelso what
21:25:29Blue_DudeWell, yes, but if you've been following the -dev list, I have been requesting input. And I got it, so don't worry about it.
21:25:37funmanbuilding 4 or 5 targets different enough is enough to see anything obvious
21:25:48kugelI really don't care if a user or two download a SVN-1 build
21:25:48LloreanBlue_Dude: Yeah, but nobody suggested low-IRAM targets
21:25:58LloreanThe list you got was the standard "each CPU and most disk types"
21:26:16kugelbut I do care on not wasting time on doing something that a whole farm of computers does in 3-4 minutes
21:26:57Lloreankugel: You could always get to work on a test build system that doesn't affect users, then.
21:27:01Blue_DudeLlorean: I already test built on the h140. It worked great. What else should I try?
21:27:22LloreanBlue_Dude: If I recall the Gigabeat F has no IRAM (or so little as we don't use it)
21:27:30LloreanSo you're almost certainly going to need an alternative solution there.
21:27:40LloreanOr maybe it's the other Gigabeat.
21:27:50funmanthe gigabeat F 'iram' sections are put into dram if they use the correct attribute (IBSS_ATTR and similar)
21:27:52Blue_DudeLlorean: then it will likely ignore the IRAM attributes and run slowly.
21:28:15Blue_Dudefunman: that too. :)
21:28:28markunnot that slowly I think
21:28:38markunthe cache makes up for the lack of iram
21:28:40funmanthere is no noticeable speed difference between iram and dram on sansa ams
21:28:49LloreanBlue_Dude: But my point is, why don't you check which ones have the least iram and try those?
21:28:58LloreanIsn't the amount defined somewhere anyway?
21:28:58funmaniram is perhaps 25% faster without caching
21:29:13funmanin the linker scripts
21:29:37kugelfunman: I only had 7% speed up from putting the lcd framebuffer into iram
21:29:44Blue_DudeOK, great, which are those targets and I'll do it. Really.
21:30:08funmankugel: it was before we enabled the caches, right?
21:30:22linuxstbBlue_Dude: Just write a little script and build them all. It would be quicker than arguing about it...
21:30:42Lloreanlinuxstb: All targets would take forever, he's right.
21:30:51Blue_Dudelinuxstb: it would take all day.
21:30:53kugelfunman: yea
21:30:56Blue_DudeWhat he said.
21:30:56*JdGordon| agrees with chopper read.. "harden the fuck up" and commit :D
21:31:04 Quit pamaury ("exit(*(int *)0 / 0);")
21:31:20linuxstbBlue_Dude: Or all night... You don't have to be there when they're running.
21:31:22LloreanBut he knows the circumstances problems are likely under. Low-IRAM targets.
21:32:01 Join AndyI [0] (i=AndyI@
21:32:36 Quit FOAD (Read error: 145 (Connection timed out))
21:32:37 Nick FOAD_ is now known as FOAD (
21:32:42Blue_DudeAnybody know which targets are low-IRAM?
21:32:53LloreanBlue_Dude: I think funman said it's in the linker script
21:33:23 Join stripwax [0] (
21:33:29kugelbuggers, just let him commit and fix any problems afterwards. I really fail to see the problem with *possibly* producing read (if he's having a plan for fixing that is)
21:34:33kugeldemanding him to not only do multiple builds, but also scan through the source, is absurt, IMO
21:34:35pixelmashows some maturity
21:34:56Blue_DudeOK, I can't commit this minute. I'll try a few test builds first. My lawn needs mowing anyway so it'll have to wait.
21:35:20Lloreankugel: Yes, but you've seen what happens when a red *doesn't* take a short time to fix, and then the person who can fix it has to run off for some reason
21:35:32LloreanIf you know where a problem is *likely* you should try to fix it in advance. It's just sound to do so.
21:35:34rasherI don't think using the build system as a test build is a terrible offence, as long as you're alert to it and planning to either fix or revert within say, an hour
21:36:11 Quit KBH (Read error: 54 (Connection reset by peer))
21:36:15 Join HBK- [0] (
21:36:24*Blue_Dude fires up the mower.
21:36:46Blue_DudeI'll check back this evening and go from there. Thanks, guys.
21:37:21Lloreanrasher: There's a difference between "using it to check for unknown problems" and "I think I know what will be a problem already but haven't tested it for some reason" though
21:38:16linuxstbrasher: But that just creates a mess in the svn history... There's no excuse for not doing multiple builds yourself as a test - that's NOT the job of the build system IMO.
21:38:25 Quit pixelma (Nick collision from services.)
21:38:26 Join pixelma_ [0] (i=quassel@rockbox/staff/pixelma)
21:38:29 Quit amiconn (Nick collision from services.)
21:38:31 Join amiconn_ [0] (i=quassel@rockbox/developer/amiconn)
21:38:43 Nick pixelma_ is now known as pixelma (i=quassel@rockbox/staff/pixelma)
21:38:47*markun agrees with linuxstb and Llorean on this one
21:38:51 Nick amiconn_ is now known as amiconn (i=quassel@rockbox/developer/amiconn)
21:39:04LloreanIf the build system is supposed to be used for "now let's see what breaks so I can continue working" we should create an alternate one just for that
21:39:08RyoSkugel: cant compile convttf.. ->
21:39:15LloreanIt's idle often enough that the current network could probably deal with the cases where that's needed.
21:39:21RyoSusing the latest posted source file
21:39:27rasherMeh, if it's reverted it won't show up in the history unless you're looking for it.
21:39:50markunrasher: it will in my git history :)
21:40:02kugelRyoS: seems you need to link some lib, try to add -lm
21:40:35*JdGordon| points out that really.. its more liely to cause issues on low IRAM targets than just build breaks :)
21:40:38RyoSkugel: works ;-)
21:40:46linuxstbrasher: It also means non-atomic commits - main commit, then "fix red", then "fix red", then "oops, forgot something, fix more red"... I just think SVN needs to be treated with respect, and not as someone's personal workspace.
21:41:26JdGordon|maybe we shold be using git then!
21:41:32LloreanJdGordon| used to have a build system up somewhere that you could upload a patch to, then compile, right?
21:41:33kugelnice idea
21:41:47JdGordon|kugel: make sure you send me a patch or put it on FS before you give up for the night....
21:42:08JdGordon|Llorean: yes, but only on a chosen target
21:42:55LloreanJdGordon|: But it's the basic idea we need. Upload a diff of your changes, and the build system builds and discards a build from it and clean SVN without an SVN update, and gives you breakage stats.
21:42:57kugelJdGordon|: of course..
21:43:09kugelI don't think I'm heading to bed before you're going home anyway :p
21:43:23pixelmaI find the discussion a bit weird, my reason for testcompiling is that I want it to look as perfect as possible...
21:43:25JdGordon|Llorean: yeah, that would be nice
21:43:57JdGordon|what is the saying "perfection is the enemy of <something?>"
21:44:23Llorean"getting enough sleep" I imagine.
21:45:07pixelmaJdGordon|: if you have something against striving for perfection... it would explain things ;P
21:45:18RyoSkugel: sorry to bother you.. i either get an error ("Could not reset device") or segfault
21:45:30JdGordon|pixelma: perfection is going backwards for me :D
21:45:42kugelRyoS: no idea, I'm not really into that patch anymore
21:45:52RyoSoh, i see
21:48:42 Join Zagor [242] (
21:49:39saratogaa way to use the build system without doing an SVN commit would be nice
21:49:47 Quit funman ("free(random());")
21:50:01UnhelpfulJdGordon: "perfect is the enemy of good" iirc
21:50:19*Unhelpful agrees with saratoga on that
21:50:55Zagorsaratoga: rebuilding an old revision?
21:50:59saratogaone way to do it would be to have a branch of the main tree thats updated concurrently with the main tree but doesn't update the download serer
21:51:17Unhelpfuli try to test a sim for each of mono/greyscale/color, plus my two devices, and a build for a target of each processor family, but that's about where i stop, generally.
21:51:22JdGordon|it will have problems where a patch adds a file
21:51:26saratogajust someone way to test builds on all targets without doing a commit or waiting hours for all of them to run
21:51:32saratoga"some way"
21:52:02UnhelpfulJdGordon: i think that problem i had with your patch was only because the "add" was a "move and edit"
21:52:14Zagorsaratoga: testing uncommitted changes?
21:52:37saratogaZagor: yes, above in the logs it was commented that we too often test builds by commiting them, making a mess of teh SVN tree
21:52:38JdGordon|Unhelpful: no, adds (unless they are deleted after) will cause problems when you try to apply the patch again
21:53:00saratogathe obvious solution to this is to have a way to use the build system without making a build
21:53:05UnhelpfulJdGordon: trying to apply *any* patch twice is a fail
21:53:19saratogasince most of us are not easily able to run a hundred test builds on our local machines in a reasonable amount of time
21:53:20JdGordon|well yes
21:53:22Zagorsaratoga: I'm not sure I agree we have a problem
21:53:47JdGordon|I also think its not an issue... this isnt something which comes up very often
21:53:58JdGordon|and the red builds are when we dont expect them anyway
21:54:42Unhelpfulif we had the ability to remove commits from the history entirely we could do the build first and then delete the revision if it's red ;)
21:54:43saratogaZagor: well I don't really mind adding major changes in two commits, but it was suggested that its bad
21:55:49pixelma hmm? Who suggested that?
21:56:02ZagorI don't think "fix red" commits happen nearly often enough to consider it a problem
21:56:25pixelmathe disussion was about "just commit it and see what happens" if someone already expects problems
21:56:33JdGordon|Zagor: except in short bursts :)
21:56:36saratogaBlue_Dude: regarding IRAM, coldfire device all have it [H100, h300, iaudio X/M], all PP [too many to name], all AMS [newer Sansas] and some of the other new targets [TCC]
21:57:15pixelmasaratoga: the question was for low IRAM targets
21:57:23JdGordon|kugel: fixed it yet?! i'm cursios to see screenshots
21:57:24Zagorpixelma: I agree that's a too lazy approach. but that is fixed by yelling at the lazy bum, not by me spending 50 more hours on the build system ;)
21:58:00saratogapixelma: I don't think theres really a distinction
21:58:18saratogaall low IRAM targets don't have the IRAM defines enabled
21:59:07 Quit LambdaCalculus37 ()
21:59:31saratogaunless you count 96KB as low, since those are the smallest that have the IBSS defines enabled AFAIK
21:59:55 Join raphi [0] (
22:00:17pixelmasaratoga: is there a place you can look this up?
22:00:25kugelJdGordon|: I'm having weird problems with a debug field I added to wps_data, I set it to true in init, but it's false while parsing
22:00:40saratogapixelma: probably just the source code
22:01:04Unhelpfulwhat kind of threading tools do we have for targets with COP? i understand that we do rather a lot of the multi-core stuff just by using cacheline-sized-and-aligned data that is unique to each core, or uncached data, but we obviously have some things for communicating between them?
22:01:25saratogaactually archos should also have IRAM enabled, but its not SWCODEC so that shouldn't matter
22:01:43pixelmayes, but is there (a) certain file(s) where you can find which target has what?
22:01:59*Unhelpful is still stuck on making buffer compactible :)
22:02:05saratogaUnhelpful: theres a bunch of functions for synchronisation, stuff like queues, semiphores, etc
22:02:05pixelmaor maybe which arch
22:02:28 Join efyx_ [0] (
22:02:42saratogai think you'd have to look at the config-xxx.h files for your target and see what they define, or else look at the linker script section pertaining to your player
22:02:57saratogabut i am not an expert on that part of firmware so I could be wrong
22:03:01Unhelpfulsaratoga: and all of those work just as well between cores as they do between processes on one core? certain assumptions can obviously only be made in the latter case.
22:03:21saratogaUnhelpful: at least on PP they are meant to be used for multicore stuff
22:03:35saratogai wrote versions of the mp3 on cop patch that used a few different structures for IPC
22:04:30kugel"data in init: 0x817cc00, .debug: 1" vs "data in parser: 0x817cc00, .debug: 0" −− I don't understand this :(
22:05:40saratogaUnhelpful: see the v6 patch and onward
22:06:07JdGordon|kugel: the data struct is zerod in init...
22:06:38kugelwhere? I set it to true at the end of the init function
22:07:09Unhelpfulthe basic plan i think *might* work is to have the buffering thread set a flag (via some thread-safe method) that indicates a compaction is needed, and to have the codec check the flag, stop everything including its COP threads, and wait for an event signalling compaction has completed
22:09:04Unhelpfultelling buffering that it's OK to start compaction, and telling codec that compaction is complete, is probably easy to do via an event queue... i'm not sure about signalling need for compaction, that needs to be very close to free when the flag is unset
22:09:46 Part Grahack
22:09:48JdGordon|kugel: you'll have to search yourself... skin_data_load maybe
22:11:01Unhelpfulkugel: did you figure out whatever the trouble was in font antialiasing?
22:11:20bertrikmarkun, the m6 seems to use a different magic FTL signature than both the openiboot code and the samsung yp-s3
22:11:47bertrikstrings indicate it's Whimory v2.1.2, the Samsung yp-s3 uses Whimory v2.2.0
22:11:49kugelUnhelpful: I just know that mpegplayer doesn't do it on LCD_PORTRAIT, and adapting is non-trivial (for me...)
22:12:20saratogaUnhelpful: which codecs need this?
22:12:24Unhelpfulkugel: ah, mpegplayer wants rotated text drawing on LCD_PORTRAIT
22:12:28saratogait seems like it'd be very complicated to make work
22:12:44pixelmaUnhelpful: that everything looks blurry ;)
22:13:42Unhelpfulsaratoga: it's not for codecs themselves, but as a solution for several other things... the main goal is to be able to move any items in the buffer during playback.
22:14:07kugels/is to be able to move any items in the buffer during playback./add malloc()/
22:14:37Unhelpfulkugel: hey, now. also to be able to move the buffer ends during playback
22:15:08Unhelpfulbut a more malloc-y buffer would stop us having to rebuffer if the AA size changes
22:15:24JdGordon|I'm pretty sure moving buffers around is not part of the malloc spec..
22:15:35JdGordon|realloc is different
22:16:54markunbertrik: and are these all completely different?
22:17:29bertrikthe names of the Whimory functions and even parts of code look very similar
22:19:23bertrikthe openiboot code seemed to refer to some strings that should be present in the flash (e.g. "BBT" for the bad block table), I can't find those in the meizu or samsung firmwares, but otherwise the meizu and samsung FTLs look quite similar
22:19:34 Join _lifeless [0] (n=lifeless@
22:19:46UnhelpfulJdGordon: the virtual addresses that the caller sees in an OS with VM can't change, since malloc gives a direct pointer to the allocated space. but moving the physical address and changing the mapping is something that can happen.
22:20:50linuxstbbertrik: The Nano2G code contains strings such as "c:\bwa\N36Firmware-38\srcroot\Firmware\Shared\Services\hwapi\soc\samsung\nand\Whimory2_1\core\VFL\VFLBuffer.c" which would suggest 2.1
22:21:56UnhelpfulVM is not always an option for us... and is probably more than we really need most of the time
22:24:08 Quit bertrik ("boot to linux")
22:24:17Unhelpfulactually... is the position saved when we stop/resume playback sample-accurate?
22:24:38kugelJdGordon|: ok, parse_image_load() is fine (the image gets his id). But later in displaying, it can't find that id in find_image()
22:24:42 Join FOAD_ [0] (
22:25:40kugelas soon as I start playback it seems to be lost (and I can't show the statusbar before)
22:25:48Unhelpfulit may be that we could simply add a "don't-really-stop" operation that works just like a stop, but doesn't do a fadeout or flush buffers
22:26:00 Join bertrik [0] (
22:26:19 Quit merbanan (Read error: 110 (Connection timed out))
22:26:23Stephen__babe ?
22:26:37kugelI'm wondering if the wps is overwriting our stuff
22:27:26kugelwps_data seems to be healthy, but the linked list not
22:28:57Unhelpfulkugel: add some canary data before/after the LL so that you can see when it gets trashed?
22:29:25bertrik"canary data"? :)
22:29:45bertrikIs that the same as "cookies"?
22:30:17kugelit's data made in canada I think ;)
22:31:14Unhelpfulbertrik: maybe? unimportant memory segments filled with known values that you can use to detect overwrites. like taking a canary in a cage into a mine with you, in the hope that if you get into anything unsafe to breath, it will die while you still have time to get out?
22:31:45bertrikI'd love to see something like that in rockbox
22:32:04bertrikwe do that for the stacks already I think
22:33:07Unhelpfulwe can actually do better on our MMU targets, by putting no-write mappings on either end instead of padding and testing
22:34:27saratogaUnhelpful: how hard would that be to try?
22:34:43bertrikthe AMS targets currently have a rather simple MMU setup, with properties defined with a granularity of 1 MB
22:34:46saratogai'd love to know if buffers get overwritten on AMS
22:34:50Unhelpfulsaratoga: i really have no idea, i don't know how exactly the MMUs we have work
22:35:26bertrikthe MMU is very well documented (although I don't know exactly which document)
22:35:59kugelthey're all the same, basically. only the beast's one uses different code
22:36:23Unhelpfulas i understand things on x86 you can only set these flags with page granularity. if we can do mapping with smaller sizes and alignments, and one-to-many mappings, we could potentially reuse one non-writable, non-readable page to trap buffer overruns anywhere we want...
22:37:00Unhelpfulhrm, if you're not allowed to read or write, does the mapping need to be to a valid physical address?
22:37:14bertrikI'd start with simpler non-MMU protection canaries/cookies first
22:37:24 Join efyx__ [0] (
22:38:22 Quit FOAD (Read error: 110 (Connection timed out))
22:38:23 Nick FOAD_ is now known as FOAD (
22:39:28Unhelpfulbertrik: those aren't necessarily easier, you have to test them often for them to be useful, and to know what's happened since the last test, or else you only know *that* the data has been trashed... which you know anyway, when important data was being corrupted before you added the canary
22:40:10bertrikand I'd like to add checks on uninitialised structs (e.g. mutex and wakeup) and checks on the proper context (e.g. detect when a kernel function is called in interrupt context when not allowed)
22:40:24 Quit efyx__ (Client Quit)
22:40:52bertrik*I* think it's easier: no integration with an MMU needed
22:41:45JdGordon|kugel: you *are* giving the statusbar gui_wps->datas new wps_data structs riught?
22:41:49bertrikNot all targets have an MMU anyway (although I don't know exactly which ones do and which ones don't)
22:41:53saratogawe could probably just align the buffer to the edge of a couple dozen MMU pages and then just wait to see if it goes out of them though
22:42:00kugelJdGordon|: yea, I only share wps_state
22:42:37Unhelpfulbertrik: easier to implement the canary itself, but not very easy to get useful debugging data from. you'll need to be able to get a trace of all threads executed since the last time the canary data was checked... or else you still don't know what went wrong.
22:42:58Unhelpfuldoing it with a mapping that forbids access means that you get a fault when the bad code executes
22:43:04bertrikUnhelpful, we can check the canaries on thread switch, this should give a pretty good idea of the offender
22:43:22bertrikUnhelpful, but what about targets without MMU?
22:44:03Unhelpfulbertrik: that's pretty good, actually, and i assume we can tell where the yield or sleep that caused the switch happened?
22:44:51kugelhrm, I don't even need to start playback.
22:45:02HorschtI am slightly confused...
22:45:36HorschtI installed the latest current build on my ipod video 80GB, but I noticed it buffers every 5 - 6 songs instead of 10 - 12 how it used to be
22:45:37saratogawhat about just doing it on x86? most of the code we'd want to debug is run on the sim as well
22:45:49saratogai doubt anyone is going to want to debug something like a driver anyway
22:46:04Horschtchecking the "rockbox info" in the debug menu shows only 28Mb of buffer, i do have 64Mb, though
22:46:21JdGordon|you installed the 32MB build then
22:46:41bertrikUnhelpful, I guess we can figure that out (not 100% sure)
22:47:03HorschtjDGordon no
22:47:39ZagorHorscht: which revision was it?
22:48:05kugelskin_data_load for wps seems to destory the linked list for the sb
22:48:36JdGordon|dodgey pointers somewhere?
22:48:37ZagorHorscht: that is built for 32MB. fixing...
22:49:09JdGordon|kugel: is this all a copy+paste error? :D
22:49:22Horschtah. I was just firing up my ubuntu VM to try to compile myself
22:49:29kugelI don't think so
22:50:16Zagorrasher: can the log entries be checked for updates in your graph.php? sometimes I re-run rounds and I want to see the new result.
22:51:07Horschtbtw, Zagor. I noticed this this morning. I installed the latest build (i have removed it now), went to work and noticed this. Back then i thought it was my fault, that I accidentaly downloaded the 32 MB build. Now i just redownloaded the most current version and it's 32 MB as well. Is there an issue with this in general? something wrong with the build system perhaps?
22:51:36ZagorHorscht: yes, the new build system erroneously built ipodvideo64mb for 32MB
22:51:59kugelJdGordon|: data->images is good at the beginning, but NULL after the wps is loaded
22:52:05*Horscht shakes fist at buildsystem:p
22:52:12JdGordon|only that one?
22:52:13CIA-61New commit by zagor (r22382): Build ipodvideo64mb for 64 MB.
22:52:17JdGordon|or all of the lists?
22:52:25kugelJdGordon|: progressbar and strings work
22:52:35JdGordon|whiskey tango fuck!
22:52:58kugelZagor: can you also update the source 7z? (or remove it) someone repeatedly comes and yells for it ;)
22:53:18rasherZagor: &update should do it
22:53:26Zagorrasher: ah, excellent
22:53:52rasherI figure it's rare enough that I shouldn't bother adding a link anywhere?
22:54:03amiconnkugel, Unhelpful: I'm not sure what needs to be tested for accesses going wrong, but both SH1 and coldfire have a memory access monitoring facility which can be used for this
22:54:14Zagorrasher: no, this is fine
22:54:25rasherZagor: Does it actually work? :)
22:54:28Zagoryes :)
22:54:58Unhelpfulamiconn: i was thinking along the lines of a general method that we could apply to many buffers, and which could be enabled as a debug feature
22:55:03stripwaxZagor - hm, as of since when did the 64mb build get built with 32mb settings?
22:55:04amiconnIt is used for the "Catch mem accesses" debugging aid in the debug menu, but can be reprogrammed to monitore almost arbitrary memory areas
22:55:23*stripwax noticed reduced battery life on recent builds and couldn't see when/why, so could be that
22:55:32Zagorstripwax: since the new system went live
22:55:44ZagorI assume, anyway
22:55:58Unhelpfulamiconn: so it wouldn't be hard on CF/SH1 targets to drop a "no-access" region at either end of the audio buffer, or the plugin buffer, or pretty much anywhere else?
22:56:34kugelJdGordon: how does that look? (that's debug output, not code so SFW I guess )
22:56:38amiconnYes. The CPU will throw an exception if that area is accesed. You can monitor read and write accesses separately
22:56:51Zagorstripwax: yeah, it's been 32MB all the time
22:57:32stripwaxHorscht - good find!!
22:57:38amiconnA problem on coldfire is that this exception puts the core into "emulator mode", from which there seems to be no reliable way out apart from a reset (which the debugging aid triggers by enabling the watchdog and then deliberately not servicing it)
22:57:59Horschtwhen did it go live?
22:58:27Unhelpfulamiconn: watchdog resets are actually a bit of a problem on beast... i don't always manage to get the fault address down before it restarts
22:59:25amiconnThis just means that such a capture is essentially dead end, only reset possible afterwards. It displays the faulting address just fine
22:59:32bertrikmaybe you can save it somewhere else and detect the watchdog reset on the next startup
22:59:54JdGordon|kugel: that wps_data_load is when the .wps is loaded?
22:59:54 Quit Blue_Dude ("ChatZilla 0.9.85 [Firefox 3.5.2/20090729225027]")
22:59:59ZagorHorscht: july 15
23:00:06kugelJdGordon: yes
23:00:20amiconnThe watchdog is usually disabled; uie() enables it on purpose to trigger a reset
23:00:30Horschtso... basicaly it has been built for a month now as 32Mb? funny
23:00:51ZagorHorscht: yes...
23:00:52amiconnLater I refitted this method to sh1 because it's a simple and reliable way to reset
23:01:49bertrikyeah, a watchdog reset is usually way better than jumping to the reset vector or something similar
23:02:07stripwaxthat's about when I noticed lower battery life too (but could still be a coincidence)
23:02:15amiconn(easier than taking care to reset all those hardware regs to default and calling the correct function)
23:03:01Unhelpfulamiconn: it being a dead end isn't a problem, i think... if it's 1) only in debug builds and 2) happens in a condition which probably has, or will, lead to corruption and crash
23:03:20Unhelpfulthe *sooner* you know a buffer has been overrun, the better for debugging it
23:03:53 Join GeekShado_ [0] (
23:03:57amiconnThe feature is already there, just the setup function needs to be extended to allow monitoring arbitrary memory areas
23:04:02JdGordon|kugel: how are you loading the .wps
23:04:15JdGordon|I'm going to laugh so hard if this is your "fix" commit from yesterday
23:04:15kugeluhm, it's working now
23:04:33JdGordon|it is isnt it?!
23:04:39Horschtstripwax, no. That's actualy the first thing i noticed. I assume this has several reasons: less buffer = more disk spinup = shortened runtime. Also, the 30GB models might have a battery with a lower capacity
23:05:07kugelJdGordon|: ah nevermind, just band-aided by putting sb loading after wps loading :/
23:05:07stripwaxthey do
23:05:12amiconnRight now it has just two settings: monitoring NULL pointer accesses (anything around address 0x0, so accesses involving offsets are caught as well), and monitoring write accesses to the rom area
23:05:17 Part RyoS
23:05:21kugelsettings_apply() all the time
23:05:28 Quit bertrik (Remote closed the connection)
23:05:32saratogaZagor: have the PP bootloaders been moved?
23:05:42JdGordon|kugel: close enough then... order matters...
23:05:59JdGordon|which is why the whole reset/buffering thing needs to be handled bettererer
23:06:04Unhelpfulit would not be hard to do on sim either, at least on platforms with working memprotect...
23:06:18stripwaxbut the battery capacity in the build shouldn't affect the actual runtime (but the smaller buffer will)
23:06:19kugelthe order matters? uhhh
23:06:41Zagorsaratoga: no. can you or someone else who knows them make a new zip with them in the right directories?
23:06:51JdGordon|it looks like you did load sb, reset buffer, load wps
23:06:56JdGordon|spot the booboo! :D
23:07:11 Quit GeekShadow (Read error: 145 (Connection timed out))
23:07:17saratogaZagor: c an't they just be copied?
23:07:23saratogaor even symlinked?
23:07:24 Join kkurbjunW [0] (n=karlk@
23:07:39Horschtstripwax, but it would affect the displayd estimated runtime, correct?
23:08:31 Join bertrik [0] (
23:08:48 Quit _lifeless (Remote closed the connection)
23:09:06 Join _lifeless [0] (n=lifeless@
23:10:22kugelJdGordon|: no, reset buffer, load sb, load wps is what I did
23:10:45JdGordon|pretty :p
23:10:56JdGordon|needs sometghing though... cant put my finger on what!
23:11:09Zagorsaratoga: sure they can. but that's 15 files I have to manually move around, and they're not even all called the same thing as before. I don't know what is correct, and hence would prefer someone who knows does it.
23:11:29kugelit's the wps backdrop, I think with another backdrop it would look more like a "widget"
23:11:43saratogaZagor: ah ok, I'm not actually sure either
23:11:45JdGordon|damn subtletly.... USE CUSTOMVP
23:11:51kugelI did
23:11:53 Join Horschti [0] (n=Horscht2@xbmc/user/horscht)
23:12:01JdGordon|you did? its brokn then
23:12:03saratogai suppose we should ask one of the rbutil people what they need to be called
23:12:22kkurbjunWZagor: Are you able to update the playerpics from SVN?
23:12:24kugel ;)
23:12:31Horschtioh well...good night guys and girls
23:12:40ZagorkkurbjunW: done
23:12:53bertrikI think funman once gave me a little tool called clipsplit to split an AMS firmware into smaller modules, does anyone have this?
23:13:23kkurbjunWZagor: great, thanks!
23:13:24kugelworks pretty well, updates itself pretty much every second
23:14:00JdGordon|we probably want it to update more often... so progress indicators work
23:14:07 Quit kkurbjunW (Remote closed the connection)
23:14:13kugelthat should be possible
23:14:19bluebrotherZagor: have you thought about using the build target name for the build folder for the build client? I.e. something like build-h120 instead of build-1234?
23:14:24 Join kkurbjunW [0] (n=karlk@
23:14:28JdGordon|its goig to be a decent batt hit though
23:14:39 Quit _lifeless (Remote closed the connection)
23:14:54 Quit Horschti (Client Quit)
23:14:55 Join _lifeless [0] (n=lifeless@
23:14:59kugelJdGordon|: it's free if unused on the other hand, just the code and ramsize is questionable yet ;)
23:15:32 Quit kkurbjunW (Remote closed the connection)
23:15:35JdGordon|this is now the 3rd proof of concept wps/skined statusbar :)
23:15:49 Join kkurbjunW [0] (n=karlk@
23:15:52Zagorbluebrother: hah, no I haven't! I was about to remove the build dirs altogether, but that is a much better solution.
23:16:16***Saving seen data "./dancer.seen"
23:16:18JdGordon|I think skinfm will be a nicer addition than statusbar
23:16:18kugelJdGordon|: yea, but the first one that isn't a baaaad hack :)
23:16:40JdGordon|I imaine yours is almost he same as mine against customvp no?
23:16:59kugelwhat do you mean?
23:17:01 Quit Zarggg ()
23:17:18JdGordon|mine was customvp + hack
23:17:19 Quit _lifeless (Remote closed the connection)
23:17:38kugelwell, I just call skin_update() instead gui_syncstatusbar_draw() in that event, that's all.
23:18:09kkurbjunWZagor: what determines what players are shown on the download page?
23:18:21 Join Zarggg [0] (
23:18:33 Quit Saline ()
23:18:37kugelbut my version is much cleaner! (beginning to take the credit) ... because of your skin engine rework (this time you also get some :D )
23:18:52 Quit bertrik (Success)
23:19:01kkurbjunWFor example this page exists, but it is not linked when you click on the Current build link:
23:19:02ZagorkkurbjunW: the script that creates that page. it's not connected to the actual builds list yet.
23:19:39JdGordon|kugel: a few things though... we cant kill the old statusbar.. and umm...
23:19:49kugelI think we can
23:20:02JdGordon|rec statusbar is giong to cause problems probably
23:20:31kugelwe can hardcode a skin-version of the current statusbar and use that as fallback
23:20:34JdGordon|we cant quite do everything in the current sb if its skinned
23:21:10domonokysaratoga: rbutilqt.ini lists all bootloader names and folders for the different targets :-)
23:21:15ZagorkkurbjunW: the daily builds are still done with Bagders' scripts. I don't know all the details of them.
23:22:48kkurbjunWok thanks, I'll talk to him about what needs to be done to get a target listed on the current builds page... I'm looking through the www and there are these t files that seem to be related to the builds shown, but they don't list all the targets
23:22:49 Join ehntoo [0] (
23:23:25 Join bertrik [0] (
23:23:36 Join dmb [0] (n=Dmb@unaffiliated/dmb)
23:23:55 Quit ender` (" “That’s right, babe,” Cochrane chortled. “We’re not possessors, we’re just like dimensionally disadvantaged.” -- Peter F. H")
23:24:11JdGordon|kugel: actually... hmm... no, everything can be done in a skin
23:24:16 Quit bmbl ("Bye!")
23:24:30JdGordon|we should remove the option to have graphical or number displays though to make t easier
23:24:40JdGordon|actually... haha no need
23:24:51JdGordon|it can all be done in the skin code
23:25:18JdGordon|rec sb is going to be the problem
23:25:56kugelman, this statusbar thing came into my mind last night when I went bed, I imagined so great themes, using tiny cover arts next to the list <3
23:26:10*kugel couldn't sleep for a while then
23:26:31JdGordon|I already asked Unhelpful about different sized AA for this...
23:26:49JdGordon|bassically... its going to be rediculously sweet when its all finished :D
23:27:58*domonoky wants 8x6 AA in the status bar... on a greyscale target :-)
23:28:34JdGordon|hmm... I wonder if we could disable all prettyness in the rec screen and use the old statusbar code (shrunk abit though) just for that
23:28:43bertrikanti-aliased fonts too of course
23:29:15JdGordon|hopefully all this motivates someone to do multifont properly :)
23:29:30*Bagder runs
23:29:55 Quit Horscht (Read error: 110 (Connection timed out))
23:30:00*stripwax is still a little sad that the main menu isn't in 3D
23:30:06 Join robin0800 [0] (
23:30:20pixelmamultifont would be very nice for remote targets at least
23:30:24JdGordon|sif 3d... didnt you know 5D is the bestest
23:30:40JdGordon|pixelma: have you seen the mr500 and its remote?
23:31:13 Quit moos ("reboot")
23:31:15domonokymultifont also would be nice to have a differnt font (size) in the wps then in the lists.
23:31:17pixelmano, but H300 and its remote is already quite "nice"
23:31:21kugelI think they make the remotes out of mr500's where a corner of the display broke away, and then sell both in a package :)
23:31:55kugelyou wouldn't notice the missing corner
23:32:01JdGordon|except one is mono and one is colour :)
23:32:12JdGordon|but yeah...
23:33:46ZagorBagder: can you look at why the source 7z isn't updated right by the daily build?
23:34:38 Quit robin0800 (Remote closed the connection)
23:34:41Bagderit isn't?
23:34:43 Quit raphi ("leaving...")
23:35:21saratogashould we even have a daily source? encouraging people to use SVN would be better
23:35:41Bagderyeah, but some people can't reach svn due to silly politics
23:35:44 Quit HBK- ()
23:35:45 Join moos [0] (i=mustapha@rockbox/staff/moos)
23:36:17ZagorBagder: no. has 30 copies of the same file (since july 18)
23:36:23Bagderyeah I see
23:36:39Bagderalthough I thought I already fixed this... :-)
23:37:52kkurbjunWBagder: Could you point me in the direction of what needs to be changed so that he M:Robe 500 starts showing on the current builds page?
23:38:31 Join robin0800 [0] (
23:38:41Bagderyou can't, that particular file isn't in svn... :-/
23:39:15BagderI added it now
23:42:31Zagorwhoa, ccache is a bit faster than non-ccache :-)
23:42:47ZagorCompleted: build fuze client lillebror1-zagor seconds 26.0 uplink 97 speed 792
23:43:05 Quit robin0800 (Read error: 60 (Operation timed out))
23:43:26JdGordon|Zagor: you mean it actually works if used properly?!
23:43:38ZagorJdGordon|: amazing, isn't it?
23:44:51kkurbjunWBagder: thanks for that - I have a small player picture in playerpics for the M:Robe 500
23:45:16Bagderthat needs to be added in
23:45:42 Join petur [0] (n=peter@rockbox/developer/petur)
23:45:48Bagderit has a lookup table for build => small pic name
23:45:57kkurbjunWgotcha, I'm looking at that now
23:46:04domonokyZagor: what do you think of my updated genlang.cgi patch ? :-)
23:47:46CIA-61New commit by kkurbjun (r22383): Add smallpic to table for M:Robe 500
23:48:02bertrikdomonoky, I found some of the m200v4 ascodec init sequences again
23:48:36bertrikregister 0x21 is written with value 0x16 through mask 0x3F
23:48:49kkurbjunWBagder: could you also add the MR500 to the daily.shtml page?
23:48:53bluebrotherZagor: is that with target-specific builddirs?
23:48:58bertrikregister 0x15 is written with value 0x3F through mask 0x3F
23:49:01Zagorbluebrother: yes
23:49:09bluebrothernice :)
23:49:15bertrikregister 0x16 is written with value 0x07 through mask 0x07
23:49:39bertrikregister 0x20 is written with value 0x08 through mask 0x08
23:49:43CIA-61New commit by zagor (r22384): More ccache-friendly directory names.
23:50:06Bagderkkurbjun: that's the script in the www root, it has a table at the top for all players it shows
23:50:16Bagdermrobe500 already gets built daily
23:50:37kkurbjunWyep, I saw that there's a daily page, but it's not linked in the daily table
23:51:03Bagderexactly, because it's not added in that table
23:51:05domonokybertrik: thanks. i will note it and try todo something with that info tomorrow :-)
23:51:14bertrikdomonoky, those 4 settings are done quite early in the boot process, there's lot of other settings but I suspect these may be the most important
23:51:15 Quit moos ("Rockbox rules the DAP world")
23:51:18gevaertsZagor: so now build times will geteven less predictable :)
23:51:43bluebrotherwho needs predictable build times? I want _fast_ build times. That's predictable enough! ;-)
23:52:02saratogawhen are GSOC evaluations due?
23:52:37linuxstbNext monday (24th)
23:53:10CIA-61New commit by kkurbjun (r22385): M:Robe 500: Add to dailymod table
23:53:12gevaertsbluebrother: we may just want fast builds, but Zagor is on the record as saying that he only cares about *efficient* builds :)
23:53:27kkurbjunWBagder: I see, thanks - I think I added what was needed
23:53:34JdGordon|ccache doesnt increase efficiency
23:53:47saratogalinuxstb: when can we fill them out?
23:53:49kugelZagor: did that also fix the parallel builds?
23:53:56JdGordon|efficient builds means no killed builds, ccache doesnt change anything there
23:54:01domonokybertrik: at least reg21 is interessting. they not only lower the charge pump to 3.1V but also lower CVDD to 1.1V so that may help..
23:54:04BagderZagor: care to update the web dir from svn so that kkurbjun's update appears?
23:54:17Zagorkugel: no, I don't know what causes that yet. I've added some more logging though to help me find out.
23:54:18linuxstbsaratoga: Starting now I think.
23:54:36kugelZagor: I thought because you removed a GIMMEMORE message
23:54:38 Quit petur (Read error: 104 (Connection reset by peer))
23:54:47ZagorBagder: done
23:54:50 Quit dmb (Read error: 113 (No route to host))
23:55:05kkurbjunWThank you
23:55:11Bagderfancy mrobe500 pics!
23:55:22Zagorkugel: yeah it was unnecessary, but I don't see it causing that bug.
23:55:56JdGordon|so the mr500 is now supported? :)
23:56:03Bagderdoes the mrobe500 manual build?
23:56:46kkurbjunWI've never tried a manual build
23:57:03Bagderhehe, then I bet it doesn't ;-)
23:57:23domonokybertrik: reg20 lowers IOVDD to 2.9V still searching what reg15 and reg16 is..
23:57:38kugeleither the mr500 is supported, or we also add the beast on the current build table
23:58:01JdGordon|just to make Zagor's best fitting more painful :)
23:58:16bluebrotherthe beast still has major installation issues. Plus this some-devices-don't-accept-single-boot.
23:58:26kkurbjunWWhat do I need to do to build a manual?
23:58:34JdGordon|mr500 has a out right terrible install experince
23:58:39Zagorrasher: &update doesn't seem to work now
23:58:46rasherZagor: Hm
23:58:56kkurbjunWJdGordon: I don't think the mr500 install is that bad

Previous day | Next day