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

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

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

Notice: Only Gecko based browsers prior to FF4 support the multipart/mixed "server push" method used by this log reader to auto-update. Since you do not appear to use such a browser, this page will simply show the current log, and not automatically update.

#rockbox log for 2011-09-27

00:02:41 Part Zagor
00:15:56 Join liar [0] (
00:21:34 Quit mikroflops (Ping timeout: 248 seconds)
00:30:07 Join Scromple [0] (
00:33:50 Quit mudd1 (Ping timeout: 248 seconds)
00:38:58 Quit Strife89 (Ping timeout: 258 seconds)
00:44:22 Join Strife89 [0] (
00:46:40 Quit mamarley (Remote host closed the connection)
00:51:03 Quit matze` (Read error: Connection reset by peer)
00:52:20 Join mamarley [0] (
00:52:39 Join [Saint] [0] (~st.lasciv@
01:03:36 Quit bluefoxx (Ping timeout: 276 seconds)
01:06:18 Join bluefoxx [0] (
01:22:08***Saving seen data "./dancer.seen"
01:24:24 Join mikroflops [0] (
01:35:59 Join mystica555_ [0] (
02:11:16 Quit liar (Remote host closed the connection)
02:18:31 Quit Strife89 (Quit: Goodnight!)
02:33:50 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
02:54:00 Quit Keripo (Read error: Connection reset by peer)
02:55:25 Join Keripo [0] (~Keripo@
03:00:36 Quit Keripo (Ping timeout: 276 seconds)
03:02:37CIA-14New commit by 03fredwbauer (r30610): Do not move NULL pointers in buflibmove_callback(). Fixes some skin crashes when changing themes.
03:04:33CIA-14r30610 build result: All green
03:22:11***Saving seen data "./dancer.seen"
03:39:17JdGordon[Saint]: are you not shipping your 24px imageset with the theme?
03:40:26[Saint]I can't remember if I put it in there or not off the top of my head.
03:40:43[Saint]I want to re-do the icons. I'll probably do so tonight.
03:40:52[Saint]24px is too small for 480x800
03:41:04JdGordonlooking at God_Eater's screenshot, the icons make it look shit :)
03:42:44[Saint]I'll probably do: 240x320 == 18x18, 320x480 == 28x28, and 480x800 == 38x38
03:43:06[Saint]Iconsets are a royal PITA to do, though... ;)
03:43:07JdGordondont they all use the same font size?
03:43:40[Saint]What? No...shit no.
03:43:52JdGordonDAMN NO! FUCK NO!
03:43:56JdGordonJUST FUCKING NO!
03:44:22[Saint]46pt font on the 240x320 port would be.....interesting ;)
03:44:37[Saint]You'd get ~5 items in the list or so. :P
03:46:17[Saint]JdGordon: I'm unsure what to do about making the titlebar use .lang strings...
03:46:35JdGordonI don't want those string in the lang file
03:46:48JdGordonI know for sure others dont either
03:47:01[Saint]I need at least one additional string. But I'm not sure about how "ethical" it would be to add it.
03:47:08JdGordonthe only good way to do it is have a skin translation system external to the core
03:47:21[Saint]There's only one (I think) I need that isn't there.
03:47:28[Saint]"Screen Locked"
03:47:40JdGordonyou sure that isnt there alreayd?
03:47:49[Saint]I'm pretty sure it'd be ok to add this, with the view that this theme will *eventually* be the default.
03:48:23JdGordon"Buttons Locked" is there
03:49:04[Saint]Hmmmm....that *might* do. Its not quite right, though.
03:49:17[Saint]It would imply the HW buttons were locked.
03:49:20[Saint]they're not.
03:49:35[Saint]Oh...I also have a skin issue for you.
03:49:49JdGordonsee a dermatologist :)
03:50:11[Saint]the touch area "none" doesn't act like the other touch areas. It doesn't wait to see if its a short or long press.
03:50:22[Saint]And, it can't be used as a long press.
03:50:27[Saint]It fires immediately.
03:50:53[Saint]This makes "Hotkey" interesting in the theme, as the popup fires as well.
03:52:27 Join fyrestorm [0] (
03:52:35 Quit evilnick (Ping timeout: 240 seconds)
03:54:04JdGordonerr... hmm...
03:54:59[Saint]I had a look at the source, but I couldn't see why it behaves differently.
03:55:19[Saint]Admittedly, I didn't really understand how it was implemented ;)
03:55:37JdGordonthe none action is quite a hack
03:56:53[Saint]Its not /crucial/, but, ideally (with a mind of eventually getting this into SVN) it'd be nice if it a: acted like the other touch areas, and b: could be used with "long_press"
03:57:03[Saint]But, not crucial.
03:57:12JdGordonwhat was none used for?
03:57:58[Saint]I use it to fire the pop-ups
03:58:16JdGordonuse skin vars instead
03:58:20[Saint]well, I use it to set the skin variable, which fires the pop-ups.
03:58:45JdGordoncant they be set directly from %T?
03:59:05JdGordonok, no
03:59:11[Saint]Nah, they can't.
03:59:17[Saint]Ah, too late.
04:00:04[Saint]Even if there /were/ a different way to do this (there isn't ;)), it'd still be nice for "none" to act as every other touch area does.
04:00:25 Quit MethoS- (Read error: Connection reset by peer)
04:00:39JdGordonok, gimme a sec
04:01:00[Saint]There's no rush. I justy wanted to make you aware of it.
04:01:13[Saint]I can open a tracker task if you'd prefer.
04:01:24[Saint]But, there's certainly no rush, none at all.
04:02:55JdGordonthe issue was none was returning ACTION_TOUCHSCREEN which is the same action that gets returned if no region is pressed
04:03:35JdGordonthough that doesnt completly explain the problem
04:07:09 Quit ender| (Ping timeout: 244 seconds)
04:19:20 Join ender| [0] (
04:29:28 Quit [fred] (Ping timeout: 260 seconds)
04:32:43 Join tmzt_ [0] (
04:34:44 Join [fred] [0] (
04:36:10 Quit tmzt (Ping timeout: 252 seconds)
04:38:46 Quit amiconn (Disconnected by services)
04:38:46 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:38:47 Quit pixelma (Disconnected by services)
04:38:49 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:38:51 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:39:08 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:43:22 Quit TheSeven (Disconnected by services)
04:43:34 Join [7] [0] (~TheSeven@rockbox/developer/TheSeven)
05:18:20JdGordon[Saint]: does silence mean you've exploded from the awesomeness that is that patch?
05:22:13***Saving seen data "./dancer.seen"
05:35:00 Quit ps-auxw (Ping timeout: 240 seconds)
05:39:18 Join ps-auxw [0] (
05:44:25 Quit Horschti (Quit: Verlassend)
05:54:09 Join Rob2223 [0] (
05:58:17 Quit Rob2222 (Ping timeout: 255 seconds)
06:00:40 Quit jhMikeS (Read error: Connection reset by peer)
06:03:54 Join jhMikeS [0] (
06:03:55 Quit jhMikeS (Changing host)
06:03:55 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
06:05:38 Quit ps-auxw (Quit: leaving)
06:06:01 Join ps-auxw [0] (~arneb@2001:470:c807:0:1532:4e5f:2ad3:4123)
06:12:40 Join ReimuHakurei [0] (
06:15:56 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
06:15:56 Quit amiconn (Disconnected by services)
06:16:18 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
06:17:44 Quit ReimuHakurei (Quit: Linkinus -
06:18:21 Join ReimuHakurei [0] (
06:40:59 Quit tmzt_ (Read error: Connection reset by peer)
06:41:23 Join tmzt [0] (
07:00:37 Join evilnick [0] (
07:00:38 Quit evilnick (Changing host)
07:00:38 Join evilnick [0] (~evilnick@rockbox/staff/evilnick)
07:19:50 Quit BHSPitMonkey (Remote host closed the connection)
07:22:15***Saving seen data "./dancer.seen"
07:25:03 Join tmzt_ [0] (
07:28:08 Quit tmzt (Ping timeout: 256 seconds)
07:35:11 Join tmzt [0] (
07:38:43 Quit tmzt_ (Ping timeout: 260 seconds)
07:39:26 Join bluefoxx_ [0] (
07:40:07 Quit bluefoxx (Ping timeout: 276 seconds)
07:50:30JdGordon[Saint]: did you try the patch?
07:55:00 Quit [Saint] (Ping timeout: 252 seconds)
08:00:32 Join [Saint] [0] (~st.lasciv@
08:10:26 Join Jak_o_Shadows [0] (
08:32:17 Quit powell14ski_ (Quit: powell14ski_)
08:37:08 Join B4gder [241] (~daniel@rockbox/developer/bagder)
08:41:07 Join ender` [0] (
08:50:51 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
08:54:04 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
08:54:10 Part Zagor
08:59:21 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
09:07:14kugel[Saint]: I was very happy with 24x24. don't you think 28x28 is too big?
09:07:49kugelI hope you don't just scale down from 480x800 as thatcompletely ignores the dpi of the screen
09:10:48 Quit Scromple (Quit: Leaving)
09:15:04 Quit Jak_o_Shadows (Ping timeout: 252 seconds)
09:22:17***Saving seen data "./dancer.seen"
09:24:29 Join n1s [0] (~quassel@rockbox/developer/n1s)
09:33:08 Quit bertrik (Read error: Operation timed out)
09:39:51 Quit fyrestorm (Read error: Connection reset by peer)
09:49:03 Quit n1s (Remote host closed the connection)
09:59:33 Join LinusN [0] (
10:01:18 Join fyrestorm [0] (
10:03:23 Quit jhMikeS (Ping timeout: 260 seconds)
10:18:08 Quit user829385 (Ping timeout: 256 seconds)
10:21:38 Join dfkt|n [0] (
10:21:39 Quit dfkt|n (Changing host)
10:21:39 Join dfkt|n [0] (dfktn@unaffiliated/dfkt)
10:22:09*[Saint] has no idea what kugel is on about...
10:23:26 Join user829385 [0] (~aoeu@
10:25:09[Saint]JdGordon: Sorry, actually I was out doing my grocery shopping ;)
10:25:56[Saint]I can't test it presently. No way to build. But if you can test it, and it works, sweet.
10:27:08[Saint]If it stops "none" from firing immediately, instead of not waiting to see if its a long or short press then its "fixed".
10:27:46kugel[Saint]: the icons
10:29:06[Saint]It basically just needs to act like the other touch actions which allow for a short and long press touch action to overlap each other and not fire fire them both in the long press case.
10:29:10[Saint]JdGordon: ^
10:29:33kugeltouchscreen presses give results immediately because the coordinates change
10:29:45kugelthat's why none is timing out so early
10:31:11kugelinternally, anyway
10:31:29[Saint]kugel: I'm still not quite sure what you mean, regarding the icons. I make each set from the Tango! SVG sources individually, making one .bmp set and scaling it just. doesn't. work ;)
10:31:45kugelI mean scaling the sizes
10:31:50[Saint]If you scale a .bmp image with our "magic colour", the result is hillarious.
10:32:11God_EaterI would imagine that depends on the scaling algorithm
10:32:17God_Eatercolour bleeding usually hurts that a lot
10:32:39kugellike if 480x800 is X then 320x4800 must be Y (where Y is simply scaled down from X and the resolution)
10:32:54kugelI'm not asking whether you're actually scaling the images with a program
10:33:14[Saint]But yeah...regarding the *sizing*, I made a mockup of the UI viewport and the font height and thought that those sizes just looked right.
10:34:05kugelmockup on your desktop lcd?
10:34:35pixelmaI found the icons looking a bit "clamped" in the lists and thought there could be a bit more spacing (maybe "included" in the image source)
10:37:01[Saint]pixelma: kugel has some configurable list height patch thing but hasn't yet committed it. That would help for individual prefernce with list spacing.
10:37:03kugel[Saint]: I suggest trying an android emulator from the sdk. it has the option to scale the window to match the real screen. that way you can judge better
10:37:44pixelma[Saint]: sorry I wasn't clear - I eant the horizontal spacing between scrollbar, icons and menu items
10:37:50pixelmameant too
10:37:55 Join nick-p [0] (
10:38:03pixelmanot the vertical spacing at all...
10:39:04pixelmaI know about the list spacing - I once tried kugel's old patch and found the result looking way to big on my phone's screen although the number of lines was the same as in android's native lists, e.g. the app manager. My guess was that the separator lines and more info per item makes quite a difference
10:39:26pixelmain appearance
10:39:41[Saint]pixelma: Ah...I actually have a one line patch that spaces the list text one character away from the icons. I wouldn't mind that in svn actually. I think that it looks better with a slight space between the list and the icons too.
10:40:33pixelmaI'd include the spacing in the icon (just some blank/magic magenta transparent pixels)
10:41:41[Saint]Hmmm, I guess that's the better option.
10:41:46pixelmano need for a code solution
10:49:31 Join swilde [0] (
10:51:42pixelmaand another comment (IIRC) - I'd prefer the dB information when changing volume be in place of the icon during that, not on the info line - in the yelloish colour and maybe even turning red over 0dB as the icon basically does. That's something I'd want for all cabbies btw.
11:05:13[Saint]RaaA doesn't go above 0dB
11:05:30[Saint]I could make it red *on* 0dB, though.
11:05:34 Join pamaury [0] (
11:05:34 Quit pamaury (Changing host)
11:05:34 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
11:05:40LloreanIt probably doesn't need that. Red typically implies "warning"
11:06:12[Saint]I was just thinking that, and 0dB doesn't == "warning".
11:06:30[Saint]So, yeah, perhaps thats not such a great idea... ;)
11:06:33LloreanDoes RaaA adjust android's media volume, or does it adjust Rockbox's volume relative to android's media volume?
11:06:40kugelalso it's not a dB scale in RaaA :)
11:06:43pixelmawell I meant that in general, and as I said - for all cabbiev2s which means targets that can go above 0dB
11:16:20 Join JdGord [0] (~AndChat@
11:20:05 Quit user829385 (Quit: Leaving.)
11:22:21***Saving seen data "./dancer.seen"
11:30:07 Join user829385 [0] (~aoeu@
11:33:07 Quit JdGord (Quit: Bye)
11:34:44 Quit bluebrother (Disconnected by services)
11:34:45 Join bluebrother^ [0] (~dom@rockbox/developer/bluebrother)
11:35:16 Join pamaury_ [0] (
11:36:08 Quit pamaury (Ping timeout: 260 seconds)
11:36:20 Quit fs-bluebot (Ping timeout: 256 seconds)
11:37:31 Join fs-bluebot [0] (
11:49:44 Join casainho [0] (
12:18:47 Quit nick-p (Quit: Leaving)
12:59:48 Quit casainho (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110905174115])
13:22:22***Saving seen data "./dancer.seen"
13:30:32 Join T44 [0] (
13:34:33 Quit Topy44 (Ping timeout: 260 seconds)
13:37:35 Join rf [0] (5a92c48b@gateway/web/freenode/ip.
13:39:19rfdo you know, if there is a datasheet for the Ipod classic screen or the ipod video screen
13:40:03[7]well, there are datasheets of similar devices at least, but as I told you, all of these have 6800/8080 interfaces, so they're not suitable for your project
13:40:38rfhope i do not annoy you with that question
13:41:31[7]so any recommendations on a device with lots of storage and an excellent DAC or a platform on which one could build such a thing?
13:42:00*[7] doesn't know the non-apple targets well enough to answer that question himself
13:43:27[7]the ipod classic has a rather crappy dac and stability problems, so that one has been ruled out already
13:46:12linuxstbrf: If you're wanting to build your own DAP, this may be of interest -
13:48:25[7]that's also where the mini2440 port i mentioned earlier came from
13:51:09 Quit ReimuHakurei (Remote host closed the connection)
13:52:27rfLyre is nice
13:52:52rfbut i already received the beagleboard...
13:52:53 Join pamaury [0] (
13:52:53 Quit pamaury (Changing host)
13:52:53 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
13:55:37 Part rf
13:55:51Tornedoesn't the beagleboard have atrocious sound?
13:56:22Tornei don't think anyone was paying very much attention when routing the audio signals
13:56:26 Quit pamaury_ (Remote host closed the connection)
13:56:30Tornegumstix certainly weren't when they did the overo
13:59:16[7]Torne: it does have I2S and I2C on the expansion header though, so you can add a sane DAC to it
14:05:06 Join wodz [0] (
14:06:24wodzThis look quite interesting and could allow us to take stack backtrace on ARM on exception Can arm experts look at this?
14:11:19pamaurywodz: we can also compile in debug mode and use the dwarf compiler output
14:11:31Tornepamaury: this is unwinding on device, though
14:11:43pamauryyes, same thing on device
14:11:59Tornethat would be massive :)
14:12:18pamauryone would need to check, most of the debug info useless and could be stripped
14:12:24Torneno, i mean the code to hadnle it
14:12:51pamauryI don't think so, we don't need to handle the full dwarf spec
14:13:35pamauryI'm not saying it's free of course :)
14:14:25 Join JdGord [0] (~AndChat@
14:14:51Zagorwe could have an address->function resolver on the web site
14:15:31pamauryZagor: that's pretty much useless without backtrace most of the time no ?
14:15:32TorneZagor: that's not very useful
14:15:38Torneunless you save entire stacks
14:15:48Torneand most people have nowhere to write those to
14:15:53Tornesince that pretty much requires serial
14:15:54ZagorI mean for multiple addresses
14:16:05TorneRight, but you don't know which things are addresses
14:16:13Torneso you end up having to display the entire stack on the screen
14:16:17Zagor...for a stack trace
14:16:42Zagoras opposed to needing full debug symbols in the builds
14:16:51Tornethis isn't about symbols
14:16:54Tornethis is about .debug_frame
14:17:00Tornethe cfi data to unwind the stack
14:17:50pamauryI'm trying a debug build to see the size
14:18:18Tornei think the more interesting part is "how big a binary is libunwind" :)
14:19:03Tornedwarf's actual encoding is weird and scary and i wouldn't want to implement it
14:19:31pamauryI had a look at it and it's not that bad I think; libunwind is probably massive otoh, handling lots of weird stuff
14:19:52pamauryhey, why the f+ debug build fails on undefined reference to `H_TO_BE32' ??
14:20:37Tornepamaury: it is pretty incomprehensible..
14:20:44Tornepamaury: the CFI format is only part of it..
14:20:53pamauryyes I know
14:21:45pamaurybut if you only want to stack trace, I *think* it's ok, I might be wrong though
14:22:04pamauryI think someone broke the debug builds
14:22:21pamauryclip+ build: rockbox/apps/recorder/pcm_record.c:1674: undefined reference to `H_TO_BE32'
14:22:44pamaurycan someone else try a debug build ?
14:23:04[7]smells like a missing include
14:23:25Torneyou don't realyl want a debug build, do you?
14:24:49pamauryTorne: does the compiler generate dwarf info in all cases ?
14:24:49pamauryindeed, H_TO_BE32 is compiled out in debug build
14:24:58Torneif not then we should do
14:25:02Tornebecause it's cheap and useful
14:25:56[7]what about implementing post-mortem memory dumps on some more targets?
14:26:09Tornedumping to where?
14:26:32[7]we have that on the nano2g, and it's <1K ramsize overhead, at least on that target
14:27:00[7]might be a bit more with less fancy OTGs
14:27:09Tornepamaury: it doesn't, no, but it should. plz fix :)
14:27:53Tornesensible debug builds change nothing except #defines
14:28:00Tornewe can jsut do -g always
14:28:56pamauryTorne: I'm willing to fix it but why the hell is it compiled out and nobody noticed ?
14:29:09Tornebecause people never build non-debug builds with -g
14:29:17Torneeven though you should :)
14:29:36pamauryI'm talking about H_TO_BE32 !
14:29:47 Quit JdGord (Read error: Connection reset by peer)
14:29:52 Join JdGord [0] (~AndChat@
14:29:56Tornenobody builds with DEBUG defined
14:30:10Tornei meant, please fix that you don't get symbols when you aren't building debug
14:30:17pamaurybut yeah, we should probably compile with -g, although I'm a bit scared by the makefiles we have so if someone else can do it I prefer :)
14:30:20[7]is that maybe just misspelled? don't we call it HTOBE32?
14:30:46pamaurythere is a define H_TO_BE32
14:30:59[7]well, then that file isn't included in the right place
14:31:04kugelTorne: why should we build with -g?
14:31:32 Quit JdGord (Read error: Connection reset by peer)
14:31:33 Join jdgord_ [0] (~AndChat@
14:31:49pamaury[7]: strange, I don't find any reference to it in apps/ :-/
14:32:00pamauryand only one in firmware/ : the define itself
14:32:20Tornekugel: so you don't have to go back and recompile everything when you want debug info?
14:32:41kugelbut isn't it a waste of space?
14:32:45[7]host to big endian conversion isn't exactly something that's done very commonly
14:32:58Torneno, debug info is not a loadable section
14:33:01Torneobjcopy will just ignore it
14:33:06pamauryoh I just understood, in debug build it seems to include more information and uses a ENC_CHUNK_MAGIC which relies on H_TO_BE32 which is not defined anywere
14:33:34Tornekugel: the point is to leave it all in the ELF binary
14:33:36pamaurybut we do have a htob32
14:33:38Tornekugel: not the actual firmware binary
14:34:19[7]a dwarfed elf in combination with a full post-mortem dump is about the best that you can possibly get
14:34:27pamaurysome info: fuze+ debug build:
14:34:28pamaury.debug_frame: 75Kib
14:34:54Torneso even if the code to decode it is free that's pretty large
14:35:03pamaury.debug_info: 1.2Mib
14:35:12Torneyou can implement a basic unwinder for ARM in under 4kb as wodz's link does :)
14:35:18TorneARM is really, really easy to unwind by hand
14:35:19pamaurywe need .debug_info but we wan't to strip most of it
14:35:35kugelTorne: I never compile with -g (explicitely) but the .elf output is just fine
14:35:43Tornethanks to the function prolog/epilog being completely predictable
14:35:52Tornekugel: ..huh?
14:35:52[7]just get the data off the device and analyze it on a pc
14:36:02Torne[7]: The point is for *end users*
14:36:09Tornewho currently report just a single address for a crash
14:36:11[7]the end users can send us a dump
14:36:13Torneso we can print, say, 5 addresses
14:36:25Torneor whatever fits on the screen
14:36:31Torneand know that they are probably the right ones
14:36:44Tornekugel: if you don't compile with -g you don't get debug info
14:37:28Tornethe .debug_* sections don't exist without that
14:37:47[7]if we want to make it easier for the end users to obtain a dump and sacrifice some more ramsize for it, we could emulate a UMS drive with a debuginfo.dat file on it :)
14:38:01pamauryHow does the wodz's link work ?
14:38:17Tornepamaury: it just disassembles a few interesting arm instructions and follows execution backwards
14:38:34TorneThe format of a functoin on ARM is so predictable that you don't need any actual unwinding data at all
14:38:35pamauryyou can't to that reliably can you ?
14:38:48TorneFor gcc-generated code you can do it with basically 100% accuracy
14:38:52Tornefor assembler it might miss a few
14:40:03Tornei ahven't looked at that specific implementation, but i've seen it done
14:41:07 Join n1s [0] (~quassel@rockbox/developer/n1s)
14:41:10*[7] likes the crashkernel (kexec on panic) approach of obtaining memory dumps
14:41:24Tornepamaury: you just take the value of lr and "jump" there, starting to "execute" instructions until you hit the function epilog
14:41:40Torneyou can ignore almost all instructions, so your emulator can be really stupid
14:41:59Tornewhen you get to the epilog you can deduce the form of the current stack frame and work out the previous return address by reading the stack
14:42:02Tornethen repeat
14:42:20[7]and a simple crashkernel fits in 1KiB of memory
14:42:25Tornesome minor caution for tail calls is about all you need :)
14:42:30pamaurythis assumes the stack is allocated/freed at once, is that an abi requirement ?
14:42:44Tornepamaury: Nope, but it always will be in compiler-generated code
14:43:09[7]you can also detect if part of it is freed prematurely while searching for the tail
14:43:29TorneThat only ever happens in handwritten assembler, though, srsly
14:43:36Tornegcc just plain doesn't do taht
14:43:43[7]premature freeing could be interpreted as an epilog with a fallthrough into the next function
14:44:08[7]so it's actually pretty much straightforward to handle it
14:45:20wodzso maybe this worth a try to extend our panic screen?
14:45:24Torneanyway. if that implementation works for our code (it's old, btu should be reasonable i think) then it's almost a no-brainer
14:45:42[7]hm, so would the crashkernel approach we're doing on the nano2g be feasible for a) other targets, b) end users?
14:45:42TorneIt gives you the ability to print a speculative return stack on the screen easily
14:46:12Torne[7]: I don't think *any* approach that involves doing something more trivial than typing in what they see ont he device's screen is going to have a lot of end user takeup
14:46:31TorneWe get a lot of people on the forum who type in what the panic screen says, including addresses/etc
14:46:39[7]well, dumping through mass storage would be more complex for us, but pretty much straightforward for the user
14:46:41TorneThey almost never respond to requests for any followup information
14:46:58Torneespecially if the way to get it involves following any instructions
14:47:11[7]print a text "please connect usb and attach dbginfo.dat to bug report" on the panic screen
14:47:32Torne[7]: if you implemented that it would be neat, and some people would use it
14:47:34n1sit would be awesome if we could write the dump to disk and just ask them for that
14:47:43Tornebut i suspect less people will bother than if it was just a couple more numbers on the screen
14:47:51pamauryof course this assumes 1) usb works :) 2) there is a computer not far away 3) user wants to do it
14:47:55[7]Torne: is it worth the ramsize overhead?
14:48:04[7]i'd estimate 16-32KB depending on the OTG
14:48:07Torne[7]: the ram size of the unwinding thing wodz found is <4kb
14:48:19[7]well, the ramsize of my trivial dumper is <1KB
14:48:31Tornewhat is that dumping to?
14:48:34Viperfangstore the dump in memory, then used a reserved bit to indecate there there is a dump there. reboot without flushing memory, allow the fresh kernel to check the bit, then dump to disk when its accessible?
14:49:06TorneViperfang: yup
14:49:09Tornethat's the other one.
14:49:14[7]Viperfang: that requires some reserved memory space for the dumper kernel
14:49:17Tornebut that's not possible on all devices
14:49:21Torne[7]: No it doesn't
14:49:32[7]otherwise it will destroy some evidence
14:49:35wodzit may not be possible to reset without killing the content of the ram
14:49:41Tornewodz: yes, that's the problem
14:49:58TorneViperfang: lots of our devices we have no way to reset reliably without going through the OF's bootloader again
14:49:59[7]well, i've dumped lots of stuff through coldboot attacks on ipods :)
14:50:02Tornewhich often reinitialises ram
14:50:14Torneor just plain overwrites it :)
14:50:34[7]but i still don't see how rebooting gets you around the need for reserved crashkernel memory
14:50:43[7]and if you have that, you can just kexec it directly
14:50:51Torne[7]: you don't need another kernel
14:51:03Torne[7]: you just dump ram into somewhere boring
14:51:10Tornelike the end of the audio buffer
14:51:18Tornewrite a magic word at a fixed physical address
14:51:19[7]yeah, but the something that does that needs some place to live in
14:51:19Tornethen reset
14:51:35Torneno, you just do it in abort context or whatever, from the panic handler
14:51:43[7]well, yeah, this is the "destroying evidence" method
14:51:49Tornehow does that destroy anything?
14:51:55[7]the important bit might have been where you copied the unimportant bits to
14:52:11Tornebut it won't be
14:52:33[7]that's a question of probabilities but if you want to be sure, you need to reserve some small amount of memory
14:52:49Torneno, i don't think it is a question of probabilities
14:52:53 Quit bluefoxx_ (Ping timeout: 255 seconds)
14:53:06 Quit wodz (Quit: Leaving)
14:53:23pamauryif you use the audio buffer for example ? (the part which is used for audio data)
14:53:29[7]for the simple dumper crashkernel i'd go for the 1KB overhead. for a UMS-based one or dumping to disk it might be worth to go through a reset if at all possible
14:53:53Tornei don't see how you can do anything useful or interesting in 1KB or whatever :)
14:54:00[7]pamaury: corrupted audio data (or memory corruption in general) might have been the cause of the crash
14:54:07Torneif you want to save the complete state of RAM you need to reserve 50% of RAM
14:54:16[7]Torne: I have a full USB dumper crashkernel for the nano2g in <1KB
14:54:42Tornebut that doesn't do the same thing
14:54:51*Torne gives up
14:55:13[7]doesn't do which thing? dumping to disk or via UMS?
14:55:15pamaurybut that's not suitable for the end user
14:55:29[7]well yeah, as i said, a UMS version might be slightly bigger
14:56:12[7]UMSboot is like 16KB but contains some unneccessary things, OTOH it has a rather easy-to-handle OTG
14:56:36[7]removing the LCD driver from that one might shave off quite some amount of ramsize
14:56:56pamaurywhat is UMSboot ?
14:57:11[7]a bootstrapping tool we use for the newer ipods
14:57:34[7]emulates a USB mass storage drive with fat16, allows the user to copy a file to it, and executes that file from DRAM start
14:58:03[7]that's what our exploit binaries contain :)
14:58:14 Join JdGord [0] (
14:58:26pamauryit reimplements the usb driver ?
14:58:52[7]or rather copy&pasted from rockbox/emcore/whatever
14:59:19 Join WalkGood [0] (~4@unaffiliated/walkgood)
14:59:41pamauryAnyway for most controllers you can greatly simplify the code if you know the purpose of it
15:00:15[7]guess how i managed to shrink my dumper to 1KB :)
15:00:30pamaurybut still, the inconvenient is that you need a computer
15:00:51[7]writing to disk is not always a good solution though
15:01:04[7](file system corruption could be the crash cause)
15:01:14pamauryyes I know, but the same for usb
15:01:28 Quit jdgord_ (Ping timeout: 276 seconds)
15:01:29[7]why that?
15:01:39[7]USB will need a PC, yes, but that's it
15:01:46pamauryif your usb code is buggy or you managed to put the usb controller is a strange state, the dumper will not work
15:02:09[7]thats why a rather trivial and well-tested dumper will completely reinitialize the USB core
15:03:03[7]there can of course always be a bug in the exception handler, but that's something that can be tested comparatively well
15:06:20 Quit B4gder (Quit: Konversation terminated!)
15:07:20 Quit antil33t (Read error: Connection reset by peer)
15:07:41 Join antil33t [0] (
15:08:46pamauryalso regarding the size of the "dumper" (whatever shape it would have), I think there is no point in crushing its size as most as possible since most recent targets have >32Mb of memory and rockbox.bin is already 600Kib in size
15:09:15[7]I'm fairly certain that 16KB will be doable at least with some OTGs
15:09:30[7]with thumb code probably even less
15:10:19[7]remember we've even managed to fit a readonly NAND+FTL+FAT32 driver into <4K on nano2g
15:10:38[7]but yeah, that was with both thumb and UCL
15:10:58pamauryas usual it depends on the hardware
15:11:01[7]it was a very tight fit, but in the end it worked
15:11:17[7]I like that hack :)
15:11:30[7]it allows us to install rockbox on a factory nano2g by just copying a single file to it and unplugging it
15:11:54[7](bootstrapping through a buffer overflow exploit in the OF)
15:12:25[7]if that kind of thing can be done, i'm fairly sure that we can manage a mass storage memory dumper as well
15:12:45pamauryyeah such a hack is nice :) But if we are going to a general dumper, we can sacrifice size a bit to make it run on several targets
15:13:05[7]that highly depends on the target
15:15:02pamaurythere is the device part and the OTG part
15:15:27pamauryyou don't want to copy&paste the same OTG controller code for each device using it !
15:16:15[7]i'm more worried about implementing the mass storage part, which will probably need to be forked from our current driver
15:16:24[7]the OTG part can probably be used unchanged
15:16:51pamauryyes the ums part will need to be changed
15:17:04pamaurymost probably we want to emulate at the same time ums+fat
15:17:09[7]possibly ifdef out some unneeded stuff in the OTG drivers
15:17:56[7]fat is trivial, that's big mux based on the sector number being read
15:18:37[7]the first couple of sectors will be generated on the fly, containing the basic FS metadata, and all sector numbers above a certain threshold will just map to DRAM+IRAM linearly
15:19:37pamaurywe want to dump registers too
15:20:05[7]those would just be dumped into some reserved space somewhere (end of DRAM?)
15:20:23pamauryyou can reserve one sector for it in the fat
15:20:36[7]this would have to happen at the very start of the exception handler, which is still in the "old" kernel
15:20:51[7]otherwise rockbox's panic handler will have trashed the regs already
15:21:17pamauryAren't the registers pushed on the stack or something ?
15:21:26[7]not all of them
15:21:40pamauryok, then it needs to be done
15:21:44[7]probably none at all for the current non-recoverable panic handlers
15:22:13kugelTorne: I don't know. I never added -g and could use the .elf just fine (like passing them to and it showed which symbols got bigger)
15:22:15 Quit dfkt|n (Ping timeout: 245 seconds)
15:22:23***Saving seen data "./dancer.seen"
15:22:28[7]so you'll have to dump them to memory somewhere anyway, and a fixed address at the end of RAM seems to be perfect for that
15:22:47[7]if you throw them on the stack, you might overflow it and corrupt other (important?) data
15:22:48Tornekugel: symbols are not debugging information
15:22:49pamaurykugel: without -g so still get info about which functions are where because we generate a map
15:22:53Tornekugel: symbols are symbols
15:23:00[7]and nobody really cares about ~160 bytes at the end of RAM
15:23:02Tornean entirely different class of thing
15:23:24Tornekugel: debug info is stack unwinding data, source line info, etc..
15:23:37pamaury[7]: right, and that avoids reserving a sector
15:24:58kugelTorne: oh, I thought you're only asking for the symbols for the backtrace
15:25:17 Quit mamarley (Remote host closed the connection)
15:25:19Tornekugel: no, we want unwinding info to generate the *addresses* to show in teh backtrace
15:28:37 Quit mystica555_ (Read error: Operation timed out)
15:30:22 Quit mystica555 (Ping timeout: 248 seconds)
15:33:07pamauryso, who wants to implement that ? :D
15:34:02pamauryperhaps a discussion on the ML would be useful
15:35:34 Join mystica555_ [0] (
15:37:55 Join ReimuHakurei [0] (~reimu@
15:41:22 Join mystica555 [0] (
15:44:15 Quit JdGord (Read error: Connection reset by peer)
15:49:01 Quit God_Eater (Ping timeout: 252 seconds)
15:49:45 Quit saratoga (Ping timeout: 252 seconds)
16:07:02*Zagor spots WindowManager.LayoutParams.FLAG_FULLSCREEN in android/src/org/rockbox/
16:07:15TorneZagor: aha
16:07:48TorneZagor: we probably don't want that, no ;)
16:08:52Tornebut it will make all the screen sizes change :)
16:09:18Zagoryes, [Saint] will be extatic :)
16:09:54 Quit ReimuHakurei (Quit: Leaving...)
16:14:14JdGordonI think we wanted to hadnle that by either pushing the rockbox framebuffer down, or simply not drawing where the android statusbar is
16:14:18JdGordonor something like that
16:14:29JdGordonso the screen size rockbox sees doesnt change
16:16:25Tornei'm not sure why we want it not to change..
16:16:36Torneunles syou mean we want to support this being a setting
16:16:51Tornei would be happy with it never being fullscreen, at which point the themes can just all be slightly smaller..
16:17:38 Quit bieber (Remote host closed the connection)
16:19:19JdGordonjust to stop wierd screen sizes
16:19:33JdGordon480*754 or something
16:19:46JdGordonand yeah, to make it a settings
16:19:46Tornewhy is that particularly weird, compared to the random sizes we already have?
16:20:03JdGordonis the android bar always the same size?
16:20:17Torneprobably not :)
16:21:47Torneanyway, you will inevitably have weird sizes
16:21:48kugelit isn't
16:21:59Tornesee honeycomb which has soft buttons permanently displayed
16:22:07Torneand thus *no* app can use the full LCD size
16:22:21kugelI hoped to implement it as "we don't draw into this area". of course that conflicts with themes, thus I wanted to make it configrable
16:22:27Zagorisn't it a fixed size for each resolution/dpi?
16:22:39TorneZagor: the combination of resolution and dpi, yes
16:22:52kugelI even had it implemented (kugel-rb.git/shortlog/refs/heads/host_statusbar">, however there's a glitch when enabling the bar
16:23:37Torneanyway. just fyi, "full" fullscreen is not a thing in honeycomb
16:23:46Torneand may not be on all devices in future either
16:23:55Torneso weird sizes are guaranteed
16:24:20Tornethe part you can't ahve on honeycomb is substantially larger and also at the bottom
16:24:32Zagoryeah, the screen size is quite a hurde for the android app
16:24:54TorneYou might even have more than one edge you can't use :)
16:24:54kugelsizes* :)
16:28:28 Join powell14ski_ [0] (
16:37:08 Join scanf [0] (~x32@unaffiliated/scanf)
16:37:20scanfso i have rockbox on a fifth gen ipod video
16:37:28scanfbattery runs out
16:37:33scanfi plug it into my USB charger
16:37:47scanfit starts up, but its the default apple ipod crap
16:37:56scanfbut then when i reboot it it somehow is rockbox again
16:38:01scanfwhat is this sorcery
16:38:36marazpower on with hold on = boot to original firmware
16:38:55marazapplies to power on by plugging in as well
16:44:28 Join Strife89 [0] (~Strife89@
16:48:43 Join matze` [0] (
16:55:05 Quit user829385 (Max SendQ exceeded)
16:55:56Tornegit users: i hope you saw my mail to rockbox-dev. you need to change your remote url to keep using the current git mirror!
16:56:38 Quit evilnick (Quit: Leaving)
17:00:24LalufuTorne: that would be this mail?
17:00:43Torneer, no
17:00:51Tornehe one before :)
17:03:03LalufuAh, OK.
17:03:55Lalufuso, currently git:// refers to the old content, but that may change at any time, and the old content is permanently (kind of) available under git:// ?
17:04:21Tornewell, not any time. we will give people notice, i hope :)
17:05:42 Part Zagor
17:06:25Torneonce we get automatic mirroring using the new format then people should be able to move their work-in-progress over to that (via one of a variety of methods)
17:10:55 Quit n1s (Ping timeout: 244 seconds)
17:11:16*Torne tries to work out how to remove all the $Keyword%s
17:14:54Torne"carefully", would appear to be the order of the day :)
17:18:10 Join dfkt|n [0] (
17:18:10 Quit dfkt|n (Changing host)
17:18:10 Join dfkt|n [0] (~dfkt@unaffiliated/dfkt)
17:22:27***Saving seen data "./dancer.seen"
17:40:38Tornewe had svn:keywords set on more than a few bitmaps.
17:40:46Torneit's fun what i discover doing this stuff :)
17:43:48 Quit Farthen (Ping timeout: 240 seconds)
17:45:16 Join y4n [0] (y4n@unaffiliated/y4ndexx)
17:46:46preglow[Saint]: where can i get a piece of this android ui rework i'm hearing about?
17:46:54 Join Farthen [0] (~Farthen@2a01:4f8:101:2a4:0:bc28:b2e1:9)
17:47:28 Quit dfkt|n (Ping timeout: 258 seconds)
17:48:07Tornewell, 2506 keywords deleted..
17:48:12Tornethere are still a lot though
17:49:50Tornemany of them are in files that didn't have svn:keywords set :)
17:49:52Torneso weren't being expanded
17:51:18 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
17:58:51 Join n1s [0] (~quassel@rockbox/developer/n1s)
18:00:25n1sTorne: yeah, some people decided they were not going to follow the svn properties conventions and just ignored them because of broken tools
18:07:30 Quit matze` (Remote host closed the connection)
18:15:42Tornehow do you tell which
18:15:46TorneDo we care which $Id$ lines were from upstream sources and which are ours? :)
18:19:18 Join mortalis [0] (
18:20:55 Join Keripo [0] (
18:21:02 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
18:25:58 Quit Keripo (Read error: Connection reset by peer)
18:43:40 Quit swilde (Quit: ERC Version 5.3 (IRC client for Emacs))
18:45:01 Join Keripo [0] (
18:46:17 Quit Keripo (Read error: Connection reset by peer)
18:52:09 Join Keripo [0] (
19:03:46 Join jhMikeS [0] (
19:03:46 Quit jhMikeS (Changing host)
19:03:46 Join jhMikeS [0] (~jethead71@rockbox/developer/jhMikeS)
19:07:10 Quit mortalis (Quit: CGI:IRC)
19:09:59 Part WalkGood
19:11:49ukleinekdoes some has an idea how to restore the filesystem (for the OF) of my Archos player? I tried a few variations on partition layout and mkfs.vfat parameters, but when the filesystem fills the player crashes
19:22:03 Join Horscht [0] (
19:22:03 Quit Horscht (Changing host)
19:22:03 Join Horscht [0] (~Horscht@xbmc/user/horscht)
19:22:30***Saving seen data "./dancer.seen"
19:27:05 Quit Keripo (Quit: Leaving.)
19:27:52bertrikIs there any work going on to fix the regressions after r30589? I'm seeing bug reports even at the anythingbutipod forums now
19:29:39[Saint]preglow: Sorry, I was asleep...bein' on the other side of the world 'n all... ;)
19:29:47 Join Horschti [0] (~Horscht@xbmc/user/horscht)
19:29:50[Saint]preglow: Here 'ya go -
19:30:00kugelhow does that look like?
19:30:54[Saint]Not bad.
19:32:01[Saint]Good, in fact.
19:32:26[Saint](See, I can do compliments too...I'm not /always/ an asshole ;))
19:33:00 Quit Horscht (Ping timeout: 244 seconds)
19:36:41 Join ReimuHakurei [0] (~reimu@
19:37:49kugelthat's list spacing with slightly smaller lines + line separators + 25px ubuntu font + the icons, the rest is your cabbie
19:37:58 Join Zagor [242] (~bjst@rockbox/developer/Zagor)
19:38:18kugelZagor: :)
19:38:37ukleinekkugel: looks German
19:39:24kugelpressing ctrl+- thrice in firefox gives about the real size on my phoen
19:41:36Zagorkugel: is this the line space patch?
19:42:02kugelthat too yes
19:43:46 Join TheLemonMan [0] (
19:44:12Zagorare the separating lines the correct color? they look a little bright.
19:44:47kugelwell, what's correct is to be decided :)
19:45:30kugelit's what I did in the code. font color but darker (rgb each scaled down by 2)
19:46:20 Quit ReimuHakurei (Ping timeout: 252 seconds)
19:46:55Zagor"correct" defined as "same as native list"
19:47:17kugelimo that looks far batter than, and is usable too
19:47:58kugelI didnt check the color but I don't think it needs to be the _same_
19:48:22ZagorI agree, same is not vital. but a darker shade looks thinner, which is more appealing in my eyes
19:49:46kugeldarker is probably completely invisible at the top lines due to cabbies background. but I'm not at all opposed to making it darker
19:49:50ZagorI wouldn't say it's *far* better. it is a bit more native-like, which is a good thing.
19:50:04Zagorah right, the background is faded..
19:50:58ZagorI think the separators should extend further to the left. at least as far as the icons, possibly all the way. (native is all the way)
19:51:11kugeli agree, it's just a first version
19:51:18Zagorbut how does this coexist with themes?
19:51:36CIA-14New commit by 03bertrik (r30611): FS #12282 - basque language improved by Asier Arsuaga
19:51:46kugelI think they should extend to the most left (aligned to the title). in this case it'll still be offset a bit because the UI viewport starts at 12
19:52:38Zagornow I'm tempted to make a "native android" theme :)
19:53:24CIA-14r30611 build result: 0 errors, 5 warnings (bertrik committed)
19:53:40[Saint]Zagor: Then, it would be best to not draw the seperators with the core, and do it with skinned lists.
19:53:58[Saint]kugel's approach here is already entirely possible without additional code.
19:54:50[Saint]If its implemented kugel's way, themes and/or the user need to be able to switch the seperators off.
19:55:03[Saint]themes definitely do.
19:55:12kugelthe core can provide a nicely usable default for *all* screens with being dependant on the theme
19:55:30kugelskinned lists still need someone to make a theme for that very resolution
19:55:41[Saint]Right, but it needs to be able to be switched off...that's all I'm saying.
19:55:47Zagorkugel: yes, but we still *have* to ackomodate themes
19:55:49GodEaterkugel: I like ;)
19:56:09kugelit can be turned of in my current patch
19:56:24Zagorby the theme?
19:56:38[Saint]Theme's need to be able to override it, it'd to /good/ if users can too...not, necessary.
19:56:49[Saint]if that's possible, great, I've no complaints.
19:58:50kugel is the current patch
19:58:55 Quit Horschti (Quit: Verlassend)
19:59:00kugelZagor: via a setting so yes
20:02:04[Saint]there's no "setting_get"
20:02:15[Saint]that'd be sueful...I might look into that.
20:02:23[Saint]*useful too
20:02:54kugel[Saint]: ?
20:03:15[Saint]setting_inc/dec/set/get (the latter not implemented yet) would make for some interesting uses in themes.
20:04:00[Saint]kugel: "%?if(setting_get,setting,value)<foo|bar>"
20:04:13[Saint]for example.
20:04:37 Quit Xerion (Ping timeout: 256 seconds)
20:06:43kugel[Saint]: %St is sure implemented
20:07:41[Saint] it is.
20:07:57kugelthat's much older than the other settings stuff actually :)
20:07:59 Join Xerion [0] (
20:08:45[Saint]setting_* foo is touch specific.
20:08:57[Saint]setting_get could still possible be usefyl.
20:09:08[Saint]*useful too
20:12:40 Join benedikt93 [0] (~benedikt9@unaffiliated/benedikt93)
20:14:25Zagorcommenting out FLAG_FULLSCREEN looks nice
20:14:28 Join Horscht [0] (
20:14:28 Quit Horscht (Changing host)
20:14:28 Join Horscht [0] (~Horscht@xbmc/user/horscht)
20:14:59*[Saint] shakes a fist at zagor! ;)
20:15:14Zagorheh, yeah
20:15:32[Saint]*thankfully*, its not too much of an issue, just a minor PITA
20:15:46 Join Biont [0] (
20:15:51BiontHello :)
20:18:33BiontI'm trying to wrap my head around the "skin breaking change" in the wps syntax. Sadly, the theme I made back in my D2 days did not survive the automatic conversion. Does anyone know common problems that break the converted themes?
20:18:52Zagorstatus bars are 38, 25 and 19 pixels high for high, medium and low density screens
20:19:04[Saint]touch area changes, biont.
20:19:11[Saint]the * and & system has changed.
20:19:23[Saint]whoops... Biont: ^
20:20:13BiontSo it would be wise to start there when trying to fix the converted wps/sbs?
20:21:00BiontThank you. Very nice interesting by the way. I had a look at them yesterday
20:21:06[Saint]specifically, take a look at "pepeat_press, long_press, and the new allow_while_locked"
20:21:10Bionterr "Very interesting changes"
20:21:28[Saint]*repeat_press", too
20:22:01[Saint]repeat_press and long_press replace * and &
20:22:25BiontI don't have any of that in my wps. Or did these replace something I might have in there?
20:23:05[Saint]You said D2, right...touchscreen?
20:23:43[Saint]the touchscreen syntax changed with regard to long press and repeat press. That _should_ be all that changed specific to touchscreen.
20:24:46Biontbtw, does the sbs support touch areas by now?
20:25:07[Saint]Biont: AFAIK, it always has.
20:26:14[Saint]An example of the changes for you: "%T(4,4,120,120,*hotkey)" is now "%T(4,4,120,120,hotkey,long_press)"
20:26:36[Saint]"%T(4,4,120,120,&hotkey)" is now "%T(4,4,120,120,hotkey,repeat_press)"
20:28:01[Saint]and if you used !volume its now going to be volume,reverse_bar
20:28:05Biontjust to make sure we're on the same page: we're talking about this, are we? Your example syntax looks like you're talking about another change I'm not even aware of :D
20:28:33[Saint]Oh....hell, you theme is *that* old :-S
20:28:50BiontIt apparently is :)
20:29:06[Saint]And yes, the changes I mention here also apply, and aren't listed there.
20:29:25[Saint]So you'll need to take them into account also.
20:31:18BiontWell, I sold my D2 a year ago and now I'm trying to make my theme work on my htc wildfire again..the skin updater sure saved me a big headache, but I still don't know why it doesn't work. It's been a terribly long time since I *knew* all that stuff as well :/
20:31:51[Saint]the skin updater script isn't up to date with current changes to the skin syntax.
20:32:01[Saint]namely the touch areas (and possible other tags)
20:32:16Biontalthough the buttons were always there), so the problem must be something different. I'll report back when I have an actual question to ask, okay?
20:32:28[Saint]Sure ;)
20:32:56Zagordon't we have screen width/height tag?
20:33:08BiontThis web client likes to eat parts of my messages oO
20:33:18[Saint]Zagor: A skin tag?
20:33:37[Saint]If yes, no.
20:33:47Zagoryes, couldn't that be used to make skins semi-adaptive?
20:34:16 Join Jerom [0] (~jerome@2a02:8420:215:f000:f66d:4ff:fe45:790f)
20:35:30[Saint]I can't think of how...the binary itself should know what size notification bar to expect by checking its screen size.
20:35:54[Saint]the theme shouldn't need to do this checkign.
20:36:37 Join low_light [0] (48db2090@gateway/web/freenode/ip.
20:36:39 Join bluefoxx [0] (
20:37:16Zagorbut that means having umpteen different themes for all odd sizes. if the theme could adapt to minor differences we could perhaps cut down that number.
20:37:50bertrikso, have some kind of layout manager? :P
20:37:54[Saint]but the theme can't adapt to all odd sizes.
20:38:06[Saint]you casn't load/unload a backdrop conditionally
20:38:47gevaertsBut you can scale it or crop it
20:39:39Zagor[Saint]: no, not *all* sizes. but at least for example the various variations of each size. such as the 800 height that comes in 762, 775, 781, 816, 829 and 835 pixel heights
20:39:48gevaertsI can't remember now, can viewports get negative positions?
20:40:01[Saint]gevaerts: Sure.
20:40:55Zagorare the current crop of 10-inch tablets configured as high, low or medium-dpi screens?
20:41:15[Saint]There's *presently* no way for a theme to check the screen height/width.
20:41:23[Saint]this magic could probably be added.
20:41:31[Saint]that's Jd's domain, though.
20:41:34bertrikmaybe you could add a layout-policy to a viewport to make it expand to the screen and distribute its elements for example
20:41:53*gevaerts will help think about this more when he's back at a real keuboard
20:42:09Zagorbertrik: yes, that could be done with simple tag arithmetic.
20:42:13kugelI think we should have percentage or android's density-independant-pixel units in skins
20:43:16low_lightZagor: fails with the gnu mirror because the gcc releases are located in "gcc/releases/"
20:43:23[Saint]Why make it Android specific? Presently all "Application" themes should "just work".
20:43:31Zagoryes we need to add *some* dynamic handling
20:43:34[Saint]Assuming the other ports ever make something of themselves.
20:43:54Zagorlow_light: right. *grumbles* I'll fix it. thanks.
20:44:05[Saint]Maemo, Pandora, etc.
20:45:35kugel[Saint]: why application specific. all themes just should work :P
20:46:34Zagorhaving themes scale from 1280x720 to 96x64 can be a bit challenging :)
20:48:25 Join webguest20 [0] (
20:49:02preglow[Saint]: no worries mate, looking forward to giving it a whirl
20:49:03 Quit webguest20 (Client Quit)
20:49:53Zagormy thinking was that with simple width/4 arithmetic, current tags could basically take percent without having to be rewritten
20:50:30Zagorthough I haven't written a wps in *years* so I don't remember how we stand on such things
20:51:33kugel[Saint]: did you see I posted the patch?
20:51:35[Saint]I think Jd looked into it and it seemed simple in theory and bloody hard in practice ;)
20:51:49kugelwould be nice if you gave it a try to see if the reduced line height is better
20:52:10Zagor[Saint]: hmm :(
20:52:12[Saint]kugel: I did, yeah...but there's not a lot I can do with it apart from look at it and say "Oh, that's nice".
20:52:17kugelit's a bit too small but I could live with it
20:52:31kugel+for me
20:52:43[Saint]I'm not doing my own builds at the present.
20:56:39 Quit Horscht (Quit: Verlassend)
21:00:16Zagorlow_light: unfortunately no mirror will work right now because screwed up the binutils dir.
21:00:42kugel[Saint]: fwiw, I don't think the default theme should override this list spacing
21:00:48kugelif it gets in
21:03:44 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
21:09:42low_lightZagor: Downloading binutils from the mirror via worked fine earlier. It seems sunet doesn't "mirror" the gnu/gcc directory, they added an additional "releases" subdir.
21:10:50Zagoroh right, binutils is only broken for sh. I'll commit a new mirror.
21:11:26 Join ReimuHakurei [0] (~reimu@
21:14:43bertrikhm, maybe we should host that stuff ourselves
21:15:21bertrikor would that get too heavy on the data usage?
21:19:29CIA-14New commit by 03zagor (r30612): sunet mirror was bad. let's try funet.
21:19:59Zagorno I don't think it would be too bad. but it just feels wrong. there's a huge number of mirrors already.
21:20:25 Nick kugelp is now known as kugel (~kugel@rockbox/developer/kugel)
21:20:57CIA-14r30612 build result: 0 errors, 5 warnings (zagor committed)
21:22:33***Saving seen data "./dancer.seen"
21:23:13 Quit ReimuHakurei (Quit: Linkinus -
21:24:51Biontdo language string tags still exist? Has anything changed about them?
21:25:15BiontI have some lines like this " %al%s%?ia<%ia|%Sx(Unknown)> "
21:25:28Biontlooks like they got converted badly
21:25:43bertrikBTW, 86 seconds!
21:25:49Zagorhey bertrik, those are your warnings
21:25:54Biontwell, not exactly, but these lines break the wps nonetheless
21:26:46*bertrik denies everything
21:34:09 Join ChickeNES [0] (
21:36:39 Part low_light
21:39:28Zagor[Saint]: we need the volume popup in the menus too. is that doable?
21:40:09 Join liar [0] (
21:44:18Zagoranother oddity: scrolling the settings lists changes the setting. highly non-obvious.
21:45:14[Saint]that _shouldn't_ happen.
21:45:34[Saint]Perhaps you're hitting the setting prior to scrolling?
21:45:35Zagornot to mention the select bar being visible at all times is quite non-android-style
21:46:02[Saint]ANd no, I can't do a volume popup in the menus.
21:46:09Zagor[Saint]: yes, I go into volume (for example) and scroll up and down to see the values. the volume changes as I scroll, with no visual feedback at all
21:46:34ZagorI mean no feedback that the volume is actually changed due to just me scrolling the list
21:46:34[Saint]Hmmmmm, I haven't tried with volume.
21:46:45ZagorI also tried with bass and treble. same thing.
21:47:14[Saint]That is rather non-obvious. And, wrong.
21:47:25[Saint]I'm pretty confident other lists don't do this.
21:47:33[Saint]Max playlist, for one.
21:48:48Zagorwe also need to get started on a manual for android
21:49:32[Saint]thankfully its pretty much already there. (all the contents, rather) just need to murder one of the touchscreen manuals.
21:50:08 Quit liar (Remote host closed the connection)
21:50:09*[Saint] is *not* a TeX guru, though.
21:50:14Zagoryes. though I'd like to add some "being an app"-specific bits
21:52:19 Join bluefoxx_ [0] (
21:52:51 Quit bluefoxx_ (Read error: Connection reset by peer)
21:55:29 Join bluefoxx_ [0] (
21:55:31 Quit bluefoxx (Ping timeout: 276 seconds)
21:55:36 Join liar [0] (
22:05:29bertrikStrife89, I think I was probably wrong thinking that the PMU settings affected USB working or not on AMSv2
22:06:01Strife89bertrik: Ah, so the most recent patch doesn't really work?
22:06:21Zagorhow do I make a screenshot in raaa? the usb trick doesn't seem to work.
22:06:55bertrikStrife89, I'm not certain, I thought it had some effect, but at other times USB just worked also without the patch
22:07:31bertrikpart of the patch has been committed already by the way, just not the part of it that increases some of the voltages inside the clip+
22:08:25Zagor[Saint]: I like this look of the sbs: (only play controls + adjusted for android status bar)
22:08:38bertrikOne of the changes I can imagine having an effect, is lowering the USB detection threshold voltage from 4.5V to 3.18V (IIRC)
22:09:01Zagoroh, that diff became bigger than necessary
22:09:13Strife89bertrik: Assuming a clean checkout, then, is it possible to apply any of those patches and get USB operational on my Clip+?
22:10:45bertrikYou need some changes to enable USB on AMSv2 anyway, and to do all of the PMU init the OF does you need to change an #if 0 into an #if 1 in system-as3525.c
22:11:20 Quit bluefoxx_ (Ping timeout: 255 seconds)
22:11:34 Join GermanMushroom [0] (
22:11:59Zagorwps tags can be nested, right? so can't we just add simple plus, minus, div, mul tags?
22:12:36gevaertsZagor: they can't in general. A tag has to be coded specially to take the output value of another tag
22:12:51gevaertsI'd like to see that changed too :)
22:13:14Strife89bertrik: Sorry, but where is the file?
22:13:20 Quit TheLemonMan (Quit: WeeChat 0.3.5)
22:13:50*[Saint] hands Strife89 grep
22:14:06bertrikline 284 in firmware/target/arm/as3525/system-as3525.c
22:14:15*Strife89 keeps forgetting how useful grep is
22:15:21Strife89Anyway, any other changes needed?
22:19:05 Join mamarley [0] (
22:19:12 Join mshathlonxp [0] (msh@
22:23:14 Quit y4n (Quit: only amiga makes it possible)
22:23:19bertrikYou need to enable rockbox USB in firmware/export/config/sansaclipplus.h but that's already in the FS task I think
22:23:33bertrikYou can experiment a bit with the #if 0/1 in system-as3525.c
22:23:45Zagorhum, how do I transfer files to virtual android machines?
22:25:43 Join wodz [0] (
22:25:44Strife89bertrik: I haven't applied anything from the FS task yet.
22:26:50wodzAm I correct that if we want to do stackbacktrace from exception handler we need first switch to supervisor mode?
22:26:56wodzon ARM I mean
22:26:58 Quit benedikt93 (Quit: Bye ;))
22:28:58bertrikStrife89, this is what I use to enable USB on clip+ :
22:29:30ukleinekwodz: you mean because the registers are shadowed?
22:30:09*ukleinek opens ARMARM
22:30:11wodzall our stack setups I saw so far uses IRQ stack in exception
22:30:12Strife89bertrik: So you only change two lines?
22:30:20bertrikoh, great, some abort on the clip+ and the display goes dark after the bootloader
22:30:40mshathlonxpwtf is Playlist Catalogue?
22:31:01*Zagor learns about adb push
22:31:48*Strife89 compiles with fingers crossed.
22:32:05bertriksame symptom as the guy on abi after r30589
22:32:31bertrikI'm really starting to think of reverting that, and r30590 and r30591
22:32:41gevaertsmshathlonxp: it's something that's been described in the manual for ages
22:32:55bertrikdamnit, can't JdGordon just ever get a commit right in 1 try?
22:33:08*mshathlonxp is adding missing lines to latvian translation
22:33:19mshathlonxpwhy it is at the end of the file then?
22:33:55 Join gbl08ma [0] (~gbl08ma@
22:35:10gbl08mahi. I'm conneting to this channel just to inform that the problems that I reported some days ago regarding some themes' WPS not loading on nano2g, is solved after the fixes on SVN build
22:36:05gbl08mathanks. now iLike and photoSkins work properly, and I didn't even need to change the amount of RAM allocated to the themes (by editing the new macgic file under .rockbox)
22:36:08ukleinekwodz: I think you need to walk the stack in exception mode first and when that ends switch to svc and walk there, too
22:36:42 Quit GermanMushroom (Quit: Ik ga weg)
22:37:07ukleineks/svc/mode in SPSR/
22:37:23gevaertsmshathlonxp: at the end of what file?
22:37:44mshathlonxplang file
22:38:17 Quit gbl08ma (Client Quit)
22:38:27wodzukleinek: why?
22:39:12 Join fml [0] (
22:39:39*gevaerts can't find anything that matches whatever mshathlonxp is trying to say
22:40:16mshathlonxpLANG_SET_AS_PLAYLISTCAT_DIR for example - wasn't it added recently?
22:40:28fmlspeaking about the magic file: will the commit be reverted? What it does is done in a very unusual way.
22:40:46 Quit mamarley (Remote host closed the connection)
22:40:58 Nick kugel is now known as kugelp (~kugel@rockbox/developer/kugel)
22:41:18gevaertsmshathlonxp: Ah, right. That particular setting was added recently, yes.
22:42:22 Join low_light [0] (48db2090@gateway/web/freenode/ip.
22:42:26 Join ReimuHakurei [0] (~reimu@
22:42:39 Join mamarley [0] (
22:42:50 Join mamarley_ [0] (
22:42:54 Quit mamarley_ (Remote host closed the connection)
22:44:37Strife89bertrik: ... Well, fuck. It blacked out on me, too.
22:45:05bertrikhold power for 15s or so, then boot while holding LEFT to return to the OF
22:45:23Strife89Way ahead of you there. :)
22:46:24 Quit liar (Quit: hallowed are the ori!)
22:46:41 Join saratoga [0] (9803ec71@gateway/web/freenode/ip.
22:46:48mshathlonxpplaylist catalog was renamed to playlist catalogue
22:47:07mshathlonxpand I was wondering why I haven't noticed that name yet... :D
22:47:07saratogaJdGordon: maybe its a good idea to revert the font commit until we have a better idea whats going on with some players?
22:49:13Strife89bertrik: Using #if 0 and rebuilding.
22:49:22 Join bluefoxx [0] (
22:49:53 Join liar [0] (
22:49:54fmlsaratoga: I meant the commit that introduced a magic file to init the skin buffer
22:50:42fmlsaratoga: (if if were following up on me)
22:51:13saratogafml: no haven't read the logs
22:51:14ukleinekwodz: what's your theory?
22:51:36Zagor[Saint]: here's a screenshot too:
22:51:52 Quit n1s (Remote host closed the connection)
22:52:12wodzukleinek: about what?
22:53:08Strife89Holy CRAP that's a high-res Cabbie
22:53:11 Quit ReimuHakurei (Quit: Leaving...)
22:53:28Zagorheh yeah I don't know how to make the emulator do 1:1 pixel mapping
22:53:52fmlZagor: my first impression is that icons are too close to the text. (I do not use android.)
22:53:54Zagorit's not terribly far off though
22:54:14Zagorfml: yes. my change is the playback buttons at the bottom.
22:54:55Zagorsaint's theme has volume, shuffle and repeat buttons squeezed in there too which I think is cramped an not so pretty
22:55:34Zagorplus, of course, this screenshot shows the android status bar..
22:55:34ukleinekwodz: about how to do the backtrace
22:55:45fmlZagor: OK. Just so that the other know :-) Also, I think that the icons are of different sizes. E.g. the "Resume playback" icon looks smaller than "Playlist catalogue"
22:56:34fmlWell, I understant that technically they are of the same size, but the look as if they were different
22:57:01LloreanZagor: I'd agree with not putting volume, shuffle, and repeat down there. Shuffle and repeat at least, both seem like something you wouldn't want to change often or suddenly
22:57:04Zagorfml: yes
22:57:09wodzThe stack unwinder I provided link before should be straight forward to use in rockbox. Basically we would need to implement __current_sp() and change __return_address() to __builtin_return_address(0)
22:57:23LloreanVolume is a little more of a gray area, I don't know how many Android phones have dedicated volume buttons.
22:57:24wodzThe question is if we are interested
22:57:40ZagorI don't think I've seen any without
22:58:26LloreanZagor: If you're in Rockbox, and you use the phone's volume buttons, do you get "rockbox sized" volume steps, or "phone sized" volume steps?
22:58:33Tornei think they all do so far
22:58:37LloreanBasically, is there a reason why the onscreen volume controls would b preferable in general?
22:58:39wodzukleinek: read back irclogs - I provided link to some (maybe suitable) solution for ARM
22:58:40Zagorin rockbox, you get "rockbox steps"
22:58:57LloreanAlright, then I'm perfectly okay with dropping the onscreen volume controls there too.
23:00:01ZagorI'd like us to get a volume popup in the menus too, like you get in the wps (and like you get in android)
23:00:57LloreanIs Rockbox volume its own separate thing, or is changing Rockbox volume just changing the "media" volume in Android?
23:01:01Zagorotherwise our small volume steps can easily be interpreted as non-functioning buttons
23:01:42ZagorLlorean: we coexist. basically we use our pcm volume to create many small steps between the bigger android steps
23:02:05ZagorI'm quite proud of it :)
23:02:13LloreanZagor: Interesting. I mean, for practical purpose, as long as I can get the volume as quiet as I'd like, I'm mostly happy
23:02:47LloreanBut there are times where I listen to media *and* play games. Most games have their own volume slider, but it helps if both my media and my game can be adjusted relative to the overall volume.
23:03:02LloreanStill, sounds "good enough" for me, at least
23:03:06bertrikBah, can't manage to revert r30589,r30590,r30591 anymore
23:03:38 Join ReimuHakurei [0] (~reimu@
23:03:39Zagorwe had a debate about that and the general consensus was that fully independent volume control would be too "technical" and confusing
23:05:15ukleinekwodz: yeah saw, but didn't read it.
23:05:30LloreanOdd. An awful lot of apps (not media players, though) seem to have fully independent volume control
23:06:13CIA-14New commit by 03fredwbauer (r30613): Delay settings_reset() until after font_init(). Fixes boot crash on Fuze(v1) and probably others.
23:06:24Strife89svn -r .
23:07:53CIA-14r30613 build result: All green
23:07:54ZagorLlorean: I can't say I've seen many, actually
23:09:26 Join freddyb [0] (
23:10:47freddybbertrik: did that fix your boot problem?
23:10:52 Quit fml (Quit: ChatZilla 0.9.87 [Firefox 6.0.2/20110902133214])
23:11:02bertrikfreddyb, yes it seems so
23:11:27bertrikoh wait no
23:11:38Strife89bertrik: Related to the patch or something else?
23:18:31freddybI was hoping I had fixed FS #12292.
23:18:32fs-bluebot 3Sansa Clip+: Cannot boot since r30589 with real target (bugs, unconfirmed)
23:18:42 Quit domonoky (Read error: Connection reset by peer)
23:22:35***Saving seen data "./dancer.seen"
23:22:59Strife89Well, shit. I didn't back up my old build.
23:23:18freddybbertrik: can you tell if it at least got farther into boot?
23:24:24 Part Zagor
23:25:40 Quit low_light ()
23:26:34 Quit liar (Remote host closed the connection)
23:27:04bertrikfreddyb, actually it does seem to fix trunk, but trunk + usb patch appears to boot only once, next boot is shows a dark display after the bootloader
23:27:05 Join liar [0] (
23:32:15bertrikfreddyb, thanks for the fix so far, I'll go get my clip+ back in usable state ...
23:34:45bertrikStrife89, well, so much for trying USB on AMSv2 ... :( goodnight!
23:34:52 Quit bertrik (Quit: And That, My Liege, Is How We Know the Earth to Be Banana Shaped)
23:35:13 Quit liar (Quit: hallowed are the ori!)
23:35:32 Join liar [0] (
23:38:35 Join blind [0] (
23:39:06blindhey im not sure what generation it is.. the short square ipod nano - I'm assuming that's not 2G and not supported?
23:42:13saratogawe do support the Nano 2G, but its not remotely square :)
23:43:43wodzso what's the plan with audio library GSOC outcome?
23:44:38saratogai want to get it merged into rockbox
23:48:13ViperfangGetting screen corruption in r30610 on ipod6g
23:49:38blindah, it's 3G - just making sure because i'd hate to be wrong, but 3g nano isn't supported?
23:50:10ViperfangScratch that, cant reproduce
23:50:49ViperfangIs there a way to add a reboot option to the system menu?
23:51:09 Quit ReimuHakurei (Ping timeout: 260 seconds)
23:51:41 Nick scanf is now known as doesntcare (~x32@unaffiliated/scanf)
23:51:45 Nick doesntcare is now known as scanf (~x32@unaffiliated/scanf)
23:53:11 Join FlynDice [0] (
23:53:37 Quit Jerom (Quit: Leaving.)
23:54:59FlynDicefreddyb: You fix works just fine for my clip+ now. Thanx...
23:55:19freddybYou're welcome!
23:57:23freddybViperfang: you can "play" the rockbox binary and it will ROLO.
23:58:36 Quit wodz (Ping timeout: 260 seconds)

Previous day | Next day