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

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

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

#rockbox log for 2010-10-17

00:00:06G4OblivionSorry if I was wrong, just checking.
00:00:29TheSevenyou mean r28290?
00:00:31mc2739n1s: FS #10809 is no longer occuring with my sansa, but I no longer have the win 7 system to test on
00:00:51G4OblivionYes, my bad.
00:01:18TheSevenG4Oblivion: that comma doesn't make a difference
00:01:32G4OblivionAlright, just making sure.
00:01:36n1smc2739: i was just asking since noone had commented in almost a year and i would have though such a problem would be noticed quite often
00:01:53n1smc2739: so do you thik it should be closed or remain open?
00:02:29 Quit krazykit (Ping timeout: 264 seconds)
00:02:55mc2739I would say close it - a new task can be opened if iy occurs in the future
00:03:02TheSevenG4Oblivion: you can optionally place a comma after the last element of a list, probably to allow for easier auto-generation of such "code", and to allow moving lines around without bothering about commas. one of those got removed here.
00:03:23n1smc2739: ok
00:04:08TheSevengevaerts: could you enhance your tool to also print which function accounts for how much of the stack space?
00:04:15 Join krazykit [0] (
00:04:24linuxstbTheSeven: Hi. Do you know if anyone has done anything on the newer ipods over the past few months?
00:04:31TheSeveni'd guess it's ~500-600 bytes inside the storage driver
00:04:56TheSevenlinuxstb: there has been some poking around, but nobody really tried to disassemble the bootloader
00:05:39gevaertsTheSeven: I'm still concentrating on the database building side now. Analysis will definitely have that
00:05:44TheSevencode can be executed on the nano 2g/3g/4g and classic 1g/2g/3g, we have bootrom dumps and decrypted flash dumps of at least some of them
00:05:47gevaerts372 in storage_write_sectors
00:06:27TheSevengevaerts: hex or decimal?
00:07:18TheSeventhat isn't the worst-case path then
00:07:26TheSevenhm, might be the missing pushes
00:07:32TheSeventhe path from my paste is 424 bytes
00:07:33grawityBy the way, I noticed that occasionally, when skipping to next song, Rockbox waits for it to be completely buffered (up to 4-5 secs) and only then starts playing. Is this something that's hard to fix?
00:07:48gevaertspushes can account for that
00:07:55gevaertsI'll add those reasonably soon
00:08:30TheSevenlinuxstb: the problem is that they're using efi on every platform besides the nano2g, which doesn't make disassembling easier
00:08:56TheSevenit surely can be done, but it's an awful hell of inter-module calls through runtime-generated call tables
00:09:00n1sgrawity: how do you know rockbox waits for it to be completely buffered?
00:10:40n1sYeah, i had hoped someone would have gotten some useful info from the newer ipod firmwares by now, too
00:13:11 Quit G4Oblivion (Quit: ChatZilla 0.9.86 [Firefox 3.6.12pre/20101016051926])
00:13:23n1si'm just too terribly bad at reading asm to do anything useful myself at this stage :(
00:14:14grawityn1s: what else would it be doing? :) I actually don't know for sure, I just tried skipping songs in the "buffering thread" screen and seems like the new song only starts playing after the real/useful bars reach 1/5 or so
00:15:38grawityn1s: btw, doesn't happen in r27586 which I currently have (though it still lags by half a second or so)
00:16:02grawitystill occasionally*
00:16:12TheSevenhm, if a function isn't reentrant anyway, I could just make every local variable static, right?
00:16:41TheSeventhat might shave off a lot of stack space in the storage driver
00:19:42*grawity polishes his rockupgrade script.
00:25:21 Quit bertrik (Quit: :tiuQ)
00:28:18TheSevenis there a way to tell the compiler that function aren't reentrant, and that it can make things static if that saves something, but may keep them on the stack if that is better?
00:29:23 Join skfunnyboy_ [0] (
00:29:55gevaertsWouldn't that require a compiler with a good idea of what is better? :)
00:30:41grawityTry −−dwim
00:31:11 Quit grawity (Quit: Good night.)
00:33:08 Join Judas_PhD [0] (
00:36:49 Quit _s1gma (Quit: Leaving)
00:40:13*TheSeven just got rid of 70-80 bytes in the storage driver by making some variables static and caching some information
00:40:41TheSevenno need to recalculate things like the number of system-reserved blocks each time i need it
00:47:58 Quit leavittx_ (Ping timeout: 240 seconds)
00:52:32 Part smartboyathome
00:52:58CIA-7New commit by moos (r28292): Fix minor typos in description fields.
00:55:34 Join nerdy_kid [0] (
00:55:56CIA-7r28292 build result: All green
00:58:22 Quit earcar (Quit: bye)
01:03:00TheSevenis there any particular reason why screen_dump() needs to have a MAX_PATH-sized file name buffer?
01:03:08TheSevendon't we know the name will be way shorter?
01:04:06n1syeah, seems way overkill
01:05:17TheSevenit finally survives with (DEFAULT_STACK_SIZE + SECTOR_SIZE + BMP_LINESIZE) at 97% stack usage
01:07:10***Saving seen data "./dancer.seen"
01:08:17nerdy_kidhow often are rockbox stable releases made?
01:08:39n1sthe create_datetime_filename filename will be a max of 22-23 chars and the numbered one is shorter
01:08:47gevaertsnerdy_kid: every three to six months I'd say
01:09:00TheSevenn1s: I'll just go for 32
01:09:06nerdy_kidgevaerts ok thanks :)
01:09:14n1sTheSeven: nice and round :)
01:09:22 Quit nerdy_kid (Remote host closed the connection)
01:10:11TheSevenscreendump fix:
01:12:24TheSevengevaerts: I shaved off a total of about 512 bytes of stack usage
01:12:40CIA-7New commit by moos (r28293): Remove a lot of untranslated strings in the turkce translation since it was marked as "good translation" at, while it ...
01:14:29CIA-7r28293 build result: All green
01:17:51 Part domonoky
01:18:25gevaertsTheSeven: nice!
01:18:46TheSevengevaerts: what do you think about the open_internal hunk?
01:19:28TheSeventhis saves a lot of space in the average case, but makes things less predictable => might hide away potentioal stkovs
01:20:04TheSevenworst case: 4 bytes more stack space, a few bytes higher binsize/ramsize, and probably some neglegible performance loss
01:21:04gevaertsTheSeven: good question
01:21:27gevaertsI don't have a serious opinion
01:22:19gevaertsTheSeven: meanwhile, my tool outputs things like
01:23:46 Join binaryhermit_ [0] (~binaryher@
01:23:48gevaertsIt should take push/pop into account (but it counts them all, so if you do pop;call();push, the maximum stack usage gets counted for call()), but not variable length arrays or alloca() and friends
01:24:57gevaertsOh, and it doesn't do function pointers either
01:25:33TheSevenit should probably warn if it hits an instruction that messes with SP which it can't track (with the name of the function where it was found)
01:25:53 Quit binaryhermit (Ping timeout: 252 seconds)
01:25:54gevaertsAnd it requires handholding for recursion...
01:27:03*TheSeven wants a way to do post-mortem memory dumps, without corrupting parts of the dump in the process or relying on cold boot attacks
01:27:16gevaertsright. write_long_name() does "sub sp, sp, r3", which I could detect...
01:29:16TheSevendo we want to sacrifice ~2KB of binsize/~3KB of ramsize for such a feature?
01:31:55gevaertsmight be nice
01:32:47 Quit elcan (Read error: Connection reset by peer)
01:33:52 Join elcan [0] (
01:42:33 Quit DerPapst (Read error: Connection timed out)
01:44:08 Join fenugrec [0] (
01:44:09 Join DerPapst [0] (
01:45:32 Quit DerPapst (Client Quit)
01:46:29fenugrecHi, on my Fuzev1, bug FS #11608 is giving me a lot of grief; is there some place I can download a build between 3.6 stable and ~28000 ?
01:47:12CIA-7New commit by moos (r28294): Restore translated strings removed by accident in the previous revision.
01:47:20fenugrecI just can't get my build environment straight anymore, and I'd like to avoid re-building it completely...
01:47:50 Quit S_a_i_n_t (Disconnected by services)
01:47:52 Join S_a_i_n_t [0] (S_a_i_n_t@
01:49:15CIA-7r28294 build result: All green
01:50:48gevaertsfenugrec: not that I know of
01:51:16gevaertsWe'll probably branch for 3.7 in a few days though
01:51:46fenugrecgevaerts: from what I gather the main change between 27996 and the current daily is the system clock settings... is there a way to hex-edit the firmware to change it back ?
01:52:00fenugrecprobably not worth the effort
01:53:06 Join jepler [0] (~jepler@emc/developer/pdpc.professional.jepler)
01:53:29gevaertsI wouldn't try that
01:54:34fenugrecheh. Are you aware if there's going to be a fuze-specific regression for this stability issue ?
01:54:54gevaertsCould you rephrase that?
01:55:59fenugrecSorry, the sentence is missing a piece : is 3.7 going to undo the changes done circa ~28000 to the fuzev1 clock settings
01:56:26gevaertsYes, unless a real fix is found earlier (which doesn't seem likely)
01:56:29fenugrecWhich, as I understand, is causing problems to only a few users
01:56:49jeplerHi. I have a FUZEv2 that I'd like to use in "car mode", but the bootloader I get when I use RockboxUtility-v1.2.8-64bit seems to boot OF when USB is plugged in on boot. It looks like r27075 changed this (to require select button too), but how can I easily get a bootloader built after r27075? (commandline is OK, but I'd rather avoid having to install a toolchain to build rockbox)
01:57:22fenugrecI'm really puzzled at that problem. I guess reverse-engineering the original firmware didn't give good enough clues as to CPU frequency + voltage management ?
01:57:55jepleralso, is there any easy way to determine which revision a bootloader file such as is built from?
01:58:33fenugrecjepler : have you tried mkamsboot ?
01:59:50 Quit ender` (Quit: It's tough to make predictions, especially about the future.)
02:01:06jeplerfenugrec: I'm sure I could, so maybe my question is just how to get a bootloader-fuzev2.sansa that is r27075+..
02:03:56fenugrechmm I thought the daily builds included the BL but apparently not
02:05:53jeplerwell the date of the bootloader in the directory listing is after r27075 so maybe it's what I need
02:06:56jeplerhere goes nothing
02:08:20 Quit moos (Quit: ChatZilla 0.9.86 [Firefox 3.6.10/20100914125854])
02:08:22 Quit elcan (Ping timeout: 240 seconds)
02:08:26jeplerdrat, still booting to OF when plugged in
02:09:09 Join elcan [0] (
02:09:41fenugrecyou merged the .sansa file with an original fuzeA.bin right
02:10:29jeplera file called fuzpa.bin from, copied to fuzpa.bin on the device
02:10:44jeplerI did get the "updating firmware" message
02:10:53fenugrecsounds right
02:12:06jeplerI wonder if I'm mistaken about what r27075 is about, or whether the bootloader is older than it, or .. something else I haven't considered
02:12:10jeplerI guess I'll build it myself next
02:14:01fenugrechave you tried with the latest daily ?
02:14:58jepleryes. I'd installed with rbutil a few months ago and then tossed the device in the drawer
02:15:17jeplertoday I installed today's build using the current version of rbutil
02:16:21n1sthe bootloaders aren't available as daily builds, only specific, tested versions are releases and only after some major change that someone feels is important enough
02:17:01jeplerI am not surprised, as the impact of a bad bootloader is pretty high
02:17:14fenugrechmm anyone know how to force the boost_counter to 1 and up the clock to 248MHz ? I'm in the debug menu but can't get the freq to change...
02:18:00n1sfenugrec: go into CPU frequency and spin the wheel
02:18:08jeplerbbl. thanks for the help and info. I'll be back once I brick my device.
02:19:06fenugrecnls: yeah, it doesn't change...
02:20:20n1sit does for me
02:20:31fenugrecjust turn the wheel one click or 10 turns ?
02:20:49fenugrechmm weird it jumped to 248 for a split second !
02:21:01n1syou only need to spin it a little
02:21:12n1sone "step" should do it
02:21:20fenugrecI feel dense
02:22:04fenugrechuh. It works now, go figure...
02:28:04 Quit n1s (Quit: Lämnar)
02:33:46 Nick binaryhermit_ is now known as binaryhermit (~binaryher@
02:39:46 Join drizztbsd_ [0] (~quassel@unaffiliated/drizztbsd)
02:41:16 Quit noamsml (Read error: Connection reset by peer)
02:41:18 Join noamsml_ [0] (~noamsml@
02:41:43 Quit drizztbsd (Ping timeout: 260 seconds)
02:42:38 Nick drizztbsd_ is now known as drizztbsd (~quassel@unaffiliated/drizztbsd)
03:07:14***Saving seen data "./dancer.seen"
03:18:03 Join guest__ [0] (411dfb31@gateway/web/freenode/ip.
03:19:33guest__I know the clip plus and clip v2 are still unstable, but the plugin buffer size shoudln't be adversely affected to make it stable, correct?
03:21:04gevaertsAre you asking if the plugin buffer is likely to get smaller? If so, I don't think so
03:21:23guest__yes that was what I was asking. thank you
03:22:27 Quit guest__ (Client Quit)
03:33:26 Join robin0800 [0] (
03:38:28 Quit robin0800 (Remote host closed the connection)
03:38:55 Join robin0800 [0] (
03:43:55 Quit robin0800 (Remote host closed the connection)
03:44:22 Join robin0800 [0] (
03:44:31 Join digitxp [0] (4426d217@gateway/web/freenode/ip.
03:44:52digitxpQuestion: is noise cancellation still a bad idea?
03:47:27 Quit skfunnyboy_ (Read error: Connection reset by peer)
03:50:57 Quit robin0800 (Remote host closed the connection)
04:02:33 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:02:33 Quit amiconn (Disconnected by services)
04:02:39 Join CaptainKewl [0] (
04:02:53 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:03:08 Quit pixelma (Disconnected by services)
04:03:10 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:03:12 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:03:25 Quit TheSeven (Disconnected by services)
04:03:27 Join The_Seven [0] (~TheSeven@rockbox/developer/TheSeven)
04:03:41 Nick The_Seven is now known as TheSeven (~TheSeven@rockbox/developer/TheSeven)
04:04:25 Quit digitxp (Quit: Page closed)
04:12:02 Quit fenugrec (Quit: Leaving)
04:24:27 Quit Barahir (Ping timeout: 265 seconds)
04:26:04 Join Barahir [0] (
04:42:23 Quit guymann (Quit: emerge -uDNav world)
04:43:25 Join dys` [0] (
04:45:55 Quit dys (Ping timeout: 276 seconds)
04:46:33 Quit bluebrother (Disconnected by services)
04:46:35 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
04:49:47 Quit Judas_PhD (Quit: This is a quitting message)
04:57:01TheSevenmy proof of concept post-mortem memory dumper has 1018 bytes binsize and 1120 bytes ramsize :)
04:58:52JdGordonhow useful is the dump?
05:00:55 Quit amiconn (Disconnected by services)
05:00:56 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
05:01:16 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
05:01:53 Quit pixelma (Disconnected by services)
05:01:55 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
05:01:57 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
05:05:12 Quit MethoS- (Remote host closed the connection)
05:07:04TheSeventhat depends on the type of problem you want to debug
05:07:15***Saving seen data "./dancer.seen"
05:07:43TheSevenyou can basically download memory contents if something panics
05:10:45 Quit advcomp2019 (Read error: Connection reset by peer)
05:19:32 Join advcomp2019 [0] (~advcomp20@unaffiliated/advcomp2019)
05:20:16 Quit dfkt (Quit: -= SysReset 2.53=- Sic gorgiamus allos subjectatos nunc.)
05:23:37 Quit advcomp2019 (Client Quit)
05:23:43 Quit binaryhermit (Quit: Leaving)
05:24:00 Join advcomp2019 [0] (
05:24:10 Quit advcomp2019 (Changing host)
05:24:10 Join advcomp2019 [0] (~advcomp20@unaffiliated/advcomp2019)
05:27:54 Quit krazykit (Ping timeout: 240 seconds)
05:29:06 Join krazykit [0] (
05:31:31jeplerhmph, even the fuzev2 bootloader built from tip of trunk boots into OF when plugged into usb
05:31:41jepleror else I'm updating it wrongly
05:31:44*jepler gives up for the night
05:33:52 Quit ps-auxw (Ping timeout: 245 seconds)
05:40:52 Join guymann [0] (
05:40:58 Quit elcan (Read error: Connection reset by peer)
05:45:40 Join ps-auxw [0] (
05:47:18 Join elcan [0] (
06:02:06 Join fyre^OS [0] (
06:04:50 Quit fyrestorm (Ping timeout: 264 seconds)
06:06:30 Join Hillshum [0] (4ba5e3c8@gateway/web/freenode/ip.
06:11:10 Join flan_suse [0] (61675f5a@gateway/web/freenode/ip.
06:11:14flan_suseThis is a strange one.
06:11:32flan_suseEven with the latest current build, no files are shown in the file browser. Only folders.
06:11:58flan_suseI can still access podcasts and music via the database browser, but when using the file browser, all directories are shown as empty.
06:12:06S_a_i_n_twhat's your "Show files" setting?
06:12:18flan_suseIf I boot into the OF, all files are visible under the file browser. This is a Sansa Clip+.
06:12:58flan_suseS_a_i_n_t: Where is this setting under?
06:13:14S_a_i_n_teasiest way to get to it is via the quickscreen.
06:13:19flan_suseS_a_i_n_t: I don't recall ever touching it, and I noticed this happening just today.
06:13:43flan_suseS_a_i_n_t: Quickscreen?
06:13:48S_a_i_n_tIt is easily messed up via inadvertantly going to the quickscreen...first thing I'd check, personally.
06:14:13S_a_i_n_tQuickscreen: ^
06:14:13flan_suseS_a_i_n_t: I never heard of the Quickscreen. Maybe I did enable it by accident?
06:14:24S_a_i_n_tperhaps...check the manual.
06:14:56S_a_i_n_tI don;t own your specific player, so I can't say for sure how to envoke the quickscreen...the manual can however.
06:15:26flan_suseS_a_i_n_t: Ahhhh, no wonder...
06:15:34soapscorche, could you Please look into a forum plugin which prevents editing of posts after they have been replied to?
06:15:42flan_suseS_a_i_n_t: To access my quickscreen (Sansa Clip+) I have to hold down the "Home" button.
06:15:53flan_suseS_a_i_n_t: Maybe when it was in my pocket, it did this?
06:16:16flan_suseS_a_i_n_t: Thank you so much.
06:16:18S_a_i_n_tThe problem is fixed now I assume?
06:16:22flan_suseS_a_i_n_t: Turned out to be an easy fix! :)
06:16:31flan_suseS_a_i_n_t: Indeed. And thanks for being patient with me.
06:16:40S_a_i_n_tNo worries.
06:17:10flan_suseS_a_i_n_t: And as an added bonus, I learned about the Quickscreen. Never knew my player could do that. (I always used the "Home" key for something else, didn't realize by holding it down, I could access a hidden screen.)
06:18:04S_a_i_n_tflan_suse: That's where reading the manual comes in handy ;)
06:18:50flan_suseS_a_i_n_t: I didn read the manual, but only the parts of interest for me.
06:18:53flan_suse*I did
06:19:17flan_suseS_a_i_n_t: It's how I learned Rockbox for my player. But some parts I skipped over that did not look like I needed to bother.
06:20:22flan_suseS_a_i_n_t: This problem, at first glance, seemed like a bug or a fluke. Never would have thought it would be associated with a menu. But I'm sure there's a law somewhere out there that says "It's always the easy fix that's overlooked."
06:20:56flan_suseS_a_i_n_t: On another topic, just wondering if you've ever used custom entries with the disktidy plugin?
06:21:07S_a_i_n_tI have just seen it enough times now from other users to know that that is the first thing to check.
06:21:28S_a_i_n_tAnd, just the same...most of them have never seen the Quickscreen so have no idea they did it themselves.
06:21:29flan_suseS_a_i_n_t: Nice, and thanks again I have to say.
06:21:38HillshumIf certain playlist insertions from the database cause crashes, how exhaustive should my list of actions that will cause said crash be before I post the issue to FlySpray?
06:22:03S_a_i_n_tflan_suse: What about the disktidy plugin in particular?
06:22:10S_a_i_n_twhat are you trying to achieve?
06:22:20flan_suseS_a_i_n_t: Yeah, the "infamous clip" of the Sansa Clip+ broke for me, surprise, surprise. So I have to wear it in my pocket. That's probably how the Quickscreen menu was accessed and fiddled around with unknowingly.
06:22:39flan_suseS_a_i_n_t: It does not delete custom files when I add and enable them.
06:22:46S_a_i_n_tIIUC the Clip(s) have a software keylock.
06:22:53S_a_i_n_tworth looking at perhaps.
06:23:32flan_suseS_a_i_n_t: I added a custom entry (*.bmark) and set it to disabled. (Per the instructions.) Then I enabled it in Rockbox, and started the cleaning. But it always says "0 files cleaned."
06:23:54flan_suseHere's a pastebin of what my disktidy_custom.config looks like:
06:23:59HillshumS_a_i_n_t: Where in the manual might I look for the keylock info?
06:24:00S_a_i_n_tare there any .bmark files present to clean?
06:24:16flan_suseS_a_i_n_t: Lots of them, yes.
06:24:21S_a_i_n_tHillshum: No idea sorry.
06:25:09S_a_i_n_tflan_suse: Your pastebin didn't come through.
06:25:49flan_suseS_a_i_n_t: Of course, I enable it when in Rockbox and it becomes "yes".
06:26:53flan_suseS_a_i_n_t: Someone else, a while ago, seems to have noticed the same thing:
06:27:40flan_suseUnless my syntax is wrong.
06:28:01 Quit anewuser ()
06:29:09S_a_i_n_tIt *seems* to be correct, however I must say I didn't have much luck trying to edit the "files to be deleted" config either, though my use case was slightly different...I was trying to do a partial match on a filename, so I'm not surprised that didn't work.
06:29:57flan_suseS_a_i_n_t: No problem.
06:30:34S_a_i_n_tfrom the looks of it, it seems as though you could mess up the wildcard and try ".*" and it should just delete everything ;)
06:30:40flan_suseS_a_i_n_t: For the time being, I put a simply bash script in the root of my media player, which searches for all .bmark files and deletes them. Quick double-click, and it's cleaned.
06:30:45*S_a_i_n_t hopes that is gaurded against ;)
06:30:58flan_suseI'm not that brave enough!
06:31:54 Quit Hillshum (Quit: Page closed)
06:32:19S_a_i_n_ttry searching the tracker, and see if it has already been reported. If not, add a new entry and a description of what you've tried/where its failing.
06:37:36flan_suseDo I add a new task?
06:37:53flan_suseOkay, used to bugzilla. I got it now.
06:38:12S_a_i_n_tIf searching the tracker doesn't turn up anything suggesting it has been previously reported, then, yes.
06:38:20scorchesoap: there is a "Maximum time after posting to allow edit " setting
06:39:26flan_suseS_a_i_n_t: Already searched before coming here.
06:43:50flan_suseOkay, done:
06:43:50 Quit factor (Read error: Connection reset by peer)
06:43:56 Quit saratoga (Ping timeout: 265 seconds)
06:48:55 Quit flan_suse (Quit: Page closed)
06:49:10S_a_i_n_tHmmmm, yes. Seems it is quite borked.
06:49:36S_a_i_n_tI just tried adding "AlbumArt_*.jpg" which I would expect to work, but...nope :/
06:50:20S_a_i_n_tHmmm, trying to match the files full name "AlbumArt_{3516AD66-F98B-45C6-A872-73A764E55371}_Large.jpg" doesn't pick it up either.
06:50:36*S_a_i_n_t wonders why the default files are cleaned properly.
06:50:55JdGordonmust be magic
06:51:08S_a_i_n_tIndeed, that's my answer for it ;)
06:52:06JdGordonhave you tried putting the custom ones in disktidy_custom.config ?
06:52:57S_a_i_n_tYeah, that's where I'm editing it.
06:53:34JdGordonok, what about disktidy.config?
06:53:49S_a_i_n_tI'll try that in a sec.
06:54:20JdGordonit does load both files.. so probblay wont make a difference
06:55:07*JdGordon is getting more and more annoyed with lazy port developers not enabling plugins
06:55:28JdGordondisktidy is going to be completly broken on RaaA because it doesnt get the user path for the files
06:56:22 Join antil33t [0] (
07:00:24 Join factor [0] (
07:07:19***Saving seen data "./dancer.seen"
07:08:27S_a_i_n_tG4Oblivion: If you happen to read the logs, re: your question about Nano2G bootloaders, all you have to do is move the .ipod bootloader file to your Nano2G and pass it through the plugin "crypt_firmware.rock" using the "Open With . ." context menu.
07:09:10S_a_i_n_tYou'll then have both an .ipod and (an excrypted) .ipodx bootloader, the encrypted bootloader can be loaded to the device using ipodpatcher
07:14:21 Nick dys` is now known as dys (
07:31:11 Join chrissavery [0] (~chris@
07:32:22S_a_i_n_tchrissavery: Do you have any insight on the "Pictureflow on Nano1 - 2G performs like a bag of ass while audio is playing" angle?
07:32:53S_a_i_n_tI have a sollution, but it doesn't seem anywhere near correct...and I can't really see *why* it works.
07:33:06chrissaveryNo idea at all.
07:34:20chrissaveryI have no access to any Nanos. I ahven't even looked at PictureFlow code since my changes a few months ago. Sorry.
07:35:02S_a_i_n_tadding "#if !defined(IPOD_NANO) && !defined(IPOD_NANO2G)" around:
07:35:12S_a_i_n_trb->yield(); // allow audio to play when fast scrolling
07:35:12S_a_i_n_tbmp = surface(slide->slide_index); // resync surface due to yield
07:35:12S_a_i_n_tptr = &src[column * bmp->height];
07:35:29S_a_i_n_tseems ti fix it, but I can't see *why* it fixes it, or why it should be broken.
07:36:15S_a_i_n_tThought you might have an idea, no worries.
07:38:59chrissaveryI would guess some difference in how memory is handled on Nano. I put that code in to fix a possible memory release during the yield. On Fuze it was causing some cover art to show as "blotched bitmaps". So this forces it to reload an image if it was released. I think. Maybe on Nano that causes further overhead in memory management that slows it way down.
07:40:01S_a_i_n_tNano1G..yeah, I could possible imagine that. Nano2G is pretty damn powerful...something is definitely choking it, but I wouldn;t expect it to.
07:40:03chrissaveryIt's all a bit fuzzy now. But if that ifdef works then it's probably a good thing to do.
07:40:58S_a_i_n_tIt *works*, but I'm sure finding the right solution instead of "cutting off a limb" is technically "more correct" ;)
07:41:03chrissaveryI added the yields to make the scrolling smoother. But only tested on Fuze.
07:41:32S_a_i_n_tAha...irony. That small section completely borks scrolling on the Nanos ;)
07:41:55chrissaveryYes. Just can't win... everywhere at once.
07:41:59S_a_i_n_tyou *can* scroll..but the lag is awful, and fps terribly low.
07:42:44chrissaveryHave you had the ifdef patch submitted for others? I haven't been around and have no idea of any status on that code.
07:43:16S_a_i_n_tThe patch is on the tracker, unless it was closed.
07:43:21S_a_i_n_tShould still be there.
07:43:29chrissaveryI should actually update my Fuze because I'm using a version from many months ago now.
07:44:29chrissaveryI guess someone with authority will eventually "commit" it then.
07:46:26 Quit JdGordon (Ping timeout: 252 seconds)
07:46:57S_a_i_n_tactually, it seems it might need your input to get committed.
07:47:27 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
07:47:43S_a_i_n_tfunman suggested a comment explaining why yeilding is apparently a good thing on Fuse, but not for the Nanos and I can't explain that.
07:50:19S_a_i_n_tHmmm, actually, comments on FS #11387 seem to suggest that the yeilding was incorrect to begin with...gah...I dunno.
07:50:46*S_a_i_n_t just wants a decent-speed pictureflow for 3.7 ;)
07:56:53 Join Horschti [0] (~Horscht@xbmc/user/horscht)
08:00:05 Quit Horscht (Ping timeout: 265 seconds)
08:01:44chrissaveryI'm afraid I don't know enough about these things on the various platforms to be able to say why it works. When I was testing I noticed that audio dropped out a lot on Fuze when scrolling. I started looking for a fix and got the idea it was related to the audio buffer depleting while images were loaded. So I tried some yields to help the audio code to run while images were loading/rendering and that worked very well at fixing the dropouts and smoothing
08:02:26S_a_i_n_tdid you read kugels comments on FS #11387?
08:04:50chrissaveryYes. At the time I tried slow scrolling too and didn't see a problem. I don't know if that fix is the best way, only that it worked in my case. Maybe rather than an ifdef excluding it the correct solution is an ifdef that only includes it for Fuzev2. That may be safer for all other platforms.
08:05:56S_a_i_n_tI suggested that when it caught my eye, but noone was aware if there were greater numbers of targets misbehaving, or running as expected.
08:06:13S_a_i_n_tI could only speak for my cases at the time on the Nanos being totally borked.
08:06:17chrissaverykugel suggested a lower PRIORITY and if that works for Nano then it may be another way to fix it.
08:06:46S_a_i_n_tyeah, that's a little beyond me I suspect.
08:07:34chrissaveryYa, well me too, really. I don't know what code he changed for PRIORITY.
08:08:31S_a_i_n_tNor I, I may bug him when I see this bugs been around for a long time now and it would be nice to get it working again, but he's pretty tied up in Android I guess.
08:10:03chrissaveryI don't get in here much but I think if a post is made on Flyspray it does send an email to me. So if my input is needed there then that would be the way to get my attention.
08:12:50 Part chrissavery
08:17:16 Quit edboyer93 ()
08:39:04 Join n1s [0] (
08:39:10 Quit n1s (Changing host)
08:39:10 Join n1s [0] (~n1s@rockbox/developer/n1s)
08:56:05 Quit amiconn (Remote host closed the connection)
08:56:06 Quit pixelma (Remote host closed the connection)
08:59:15 Join pixelma [0] (quassel@rockbox/staff/pixelma)
08:59:32 Join amiconn [0] (quassel@rockbox/developer/amiconn)
09:03:37 Quit amiconn (Remote host closed the connection)
09:03:37 Quit pixelma (Read error: Connection reset by peer)
09:07:21***Saving seen data "./dancer.seen"
09:07:28 Join Rob2223 [0] (
09:10:43 Join edboyer93 [0] (
09:11:34 Quit Rob2222 (Ping timeout: 265 seconds)
09:13:31 Quit edboyer93 (Client Quit)
09:14:15 Join pixelma [0] (quassel@rockbox/staff/pixelma)
09:14:15 Join amiconn [0] (quassel@rockbox/developer/amiconn)
09:22:05 Join Kitr88 [0] (
09:25:17 Quit Kitar|st (Ping timeout: 276 seconds)
09:25:53 Join bmbl [0] (
09:25:53 Quit bmbl (Changing host)
09:25:53 Join bmbl [0] (~bmbl@unaffiliated/bmbl)
09:26:29 Quit Kitr88 (Ping timeout: 255 seconds)
09:27:17 Quit bmbl (Client Quit)
09:28:46 Join bmbl [0] (
09:28:46 Quit bmbl (Changing host)
09:28:46 Join bmbl [0] (~bmbl@unaffiliated/bmbl)
09:30:31 Join JdGord [0] (~jd@
09:32:11 Join Kitar|st [0] (
09:42:17 Join stoffel [0] (
09:44:26 Quit CaptainKewl (Ping timeout: 264 seconds)
09:45:28 Join slooopy [0] (
09:46:46 Quit slooopy (Client Quit)
09:56:35 Quit JdGord (Quit: Bye)
09:57:13 Quit xavieran (Ping timeout: 252 seconds)
09:59:01amiconn[23:35:09] <TheSeven> with incompatible plugins installed, entering credits shows an incompatibility message and a rockbox logo. when you press a key, it returns to the main menu, but with remnants of the rockbox logo still visible <== Probably a viewport/ screen clearing issue
09:59:02 Quit sasquatch (Quit: WeeChat 0.3.2)
09:59:28 Join sasquatch [0] (
10:00:53S_a_i_n_tpossibly areas left outside the UI viewport not being cleared?
10:10:43 Join xavieran [0] (
10:20:12 Join stoffel__ [0] (
10:21:11 Quit stoffel (Ping timeout: 276 seconds)
10:24:02 Quit S_a_i_n_t (Quit: I'm only going to Heaven if it tastes like caramel...)
10:29:06 Join S_a_i_n_t [0] (S_a_i_n_t@
10:29:15n1sJdGordon: have you seen fs#11572 ?
10:29:49JdGordonyeah, havnt looked into it yet thoguh
10:30:48 Quit factor (Read error: Operation timed out)
10:32:51S_a_i_n_tAha...I saw that in the sim, but the themeeditor said all was well so I didn;t know what to think.
10:33:13S_a_i_n_tDidn't bother me hugely as I haven't met a single person with FM hardware for Nano yet ;D
10:47:53 Join ender` [0] (
10:48:07CIA-7New commit by nls (r28295): Accept FS #11606 by Michael Gentry, making ACTION_TREE_STOP stop radio too.
10:49:28CIA-7New commit by nls (r28296): Oops #ifdef -> #if
10:49:42CIA-7r28295 build result: 4 errors, 0 warnings (nls committed)
10:51:22CIA-7r28296 build result: All green
10:56:59 Join bertrik [0] (~bertrik@rockbox/developer/bertrik)
11:07:24***Saving seen data "./dancer.seen"
11:14:38 Join fragilematter [0] (~fragilema@
11:28:05 Join AlexP [0] (~alex@rockbox/staff/AlexP)
11:28:07 Quit fragilematter (Ping timeout: 252 seconds)
11:46:35 Join hannesd [0] (
11:51:52 Quit hannesd ()
11:56:43 Join balintx [0] (
12:06:48S_a_i_n_tJdGordon: displays the FS #11572 behaviour in the simulator
12:07:04JdGordonoh goody :)
12:07:21S_a_i_n_tthe 2G doesn't have FM.
12:07:42*S_a_i_n_t turns into 'State the Obvious Man!'
12:07:51JdGordonbugger.. thats what autoconf selected for "nano-sim" :/
12:08:10S_a_i_n_tdidn't you implement that?
12:08:19*S_a_i_n_t lols
12:08:30JdGordonI dont know why it chose nano2g... order in must be wrong
12:09:30S_a_i_n_tjust to doublecheck, expected behavious is a missing FM backdrop, yes?
12:09:39S_a_i_n_t*behaviour, even.
12:09:51S_a_i_n_tYeah, that's what I was seeing...drove me mad.
12:10:30S_a_i_n_tI thought I was hitting the skin buffer limit, but I should have a reasonable ammount spare with that theme.
12:11:02S_a_i_n_tlook, we've been through this ;)
12:11:09S_a_i_n_tit's perfectly readable ;P
12:13:28JdGordonI tihnk it is actually out of skin buffer
12:14:01JdGordon176*220*16/8 ~= 77K
12:14:16JdGordonskin buffer is showing 71K free....
12:14:51JdGordonthis is where the bmp swapping needs to be implemented
12:15:24S_a_i_n_tshould be fine.
12:16:10S_a_i_n_ta 176x132 bitmap is 68Kb
12:16:28JdGordonah wrong dimensions?
12:16:43JdGordon176x220 is the e200... woops
12:17:08JdGordonhmm... it isnt loading a buffer... umm
12:17:21n1sshouldn't we add some kind of splash for skin buffer full?
12:17:25S_a_i_n_tso yeah, it *shouldn't* be skin buffer...and the fs reports he *can* get the backdrop to display
12:17:40 Quit krazykit (Ping timeout: 265 seconds)
12:17:59S_a_i_n_tif he goes to the preset menu, and back to FMS, the backdrop appears...I *think*
12:18:23 Join krazykit [0] (
12:18:26S_a_i_n_tpretty sure I read that...hope I'm not crazy.
12:19:01 Join leavittx_ [0] (
12:19:08JdGordonn1s: yes, better on target debugging is required.. but boring.. if you want to add it please feel free :)
12:19:12S_a_i_n_tn1s: Now that an exceeded skin buffer doesn't just fail to load...then, yes.
12:19:36JdGordonnot loading a backdrop wont fail it, not loading a image or font will though
12:19:48n1shmm, ok
12:20:03JdGordonnot loading a backdrop *shoulndt* be failing it anyway
12:20:10n1si'll put that on the end of my todo list then, if i can just find the list
12:20:22JdGordonits the big papery thing :p
12:21:59JdGordonp skin_buffer_freespace()
12:22:00JdGordon$6 = 45584
12:22:06JdGordoni.e it is running out of buffer
12:22:24JdGordonnice number that :p is it running out of buffer?
12:23:13JdGordonsize > skin_buffer_freespace()
12:25:03S_a_i_n_tperhaps I can lessen the size of things by combining common elements into the .sbs, but, that'll be annoying.
12:25:28 Quit Dreamxtreme (Read error: Connection reset by peer)
12:25:35 Join Dreamxtreme [0] (~Dre@
12:25:39 Join JdGordon1 [0] (
12:25:40S_a_i_n_tSo it's pretty safe to say that 11572 isn't a "bug" but a "things could be done better" thing.
12:26:14JdGordon1just closed saying that
12:26:40JdGordon1I'll have a quick play at unloading backdrops now, depends on the changes i'll commit it when its done
12:27:25JdGordon1if it has to swap the sbs and fm backdrop images out then it will suck going between those screens
12:27:35S_a_i_n_tif that's possible, that'll give me a bunch of room to play with...if not, I'll have to look into using .sbs in my .wps and .fms
12:28:51 Quit JdGordon (Ping timeout: 264 seconds)
12:31:11JdGordon1grr.. who coded this crap?! skin_backdrops() doesnt have a nice way to figure out which screen uses this image
12:38:52 Join kugel [0] (~kugel@rockbox/developer/kugel)
12:47:59 Quit markun (Read error: Connection reset by peer)
12:48:27JdGordon1arg.. this isnt as trivial as I'd hoped
12:49:27 Join markun [0] (
12:49:27 Quit markun (Changing host)
12:49:27 Join markun [0] (~markun@rockbox/developer/markun)
12:49:58JdGordon1do we actually have a way to stop the buffer, do a bufalloc() call and resume safely?
12:52:33JdGordon1because skins are initally loaded before audio is restarted (playback is initialised though) there is no real penalty in loading the backdrops into the audio buffer *on boot*. If a later theme uses more backdrops than the one loaded at boot it would use the skin buffer (which would more likely have room now)
12:55:44n1sif you touch the audio buffer i think playback needs to be restarted so a total rebuffer
12:56:03JdGordon1yes, but is there a correct way to do that yet?
12:56:38n1syes, i think so, if you load a voice file when playback is running it does that
12:57:13 Join DerPapst [0] (
13:03:30S_a_i_n_twtf is pdbox?
13:03:48bertrikThe rbutilqt progress bar goes only to 1/3rd when querying the TTS engine (festival) on my system in the TTS configuration dialog
13:04:00S_a_i_n_tIt build for Nano2G (and takes its sweet time doing so) and I have never actually seen it on the device before.
13:04:07bertrikAlso I can't abort voice creation (abort button does not seem to work)
13:04:59S_a_i_n_tbertrik: Have you tried using VoiceBox?
13:05:15bertrikno, never heard of it
13:05:49S_a_i_n_tAnother voice file make/builder app thingy
13:06:23 Quit Topy44 (Ping timeout: 250 seconds)
13:07:11S_a_i_n_tlinked from
13:07:25***Saving seen data "./dancer.seen"
13:08:27JdGordon1S_a_i_n_t: its a sound generator isnt it?
13:08:43S_a_i_n_t'talk clip generator
13:09:01S_a_i_n_tI had a feeling it could do .lang files also...but I seem to be incorrect
13:10:16n1spdbox is the "pure data" plugin, some kind of sound/music generator thingy
13:10:53*S_a_i_n_t suspects something incorrect in the build process then
13:11:05S_a_i_n_tas it builds for Nano2G, but doesn't seem to be included
13:11:27pixelmahuh, VoiceBox is no target plugin AFAIK
13:11:42S_a_i_n_tpixelma: Noone said it is.
13:12:06S_a_i_n_t*No one, even
13:12:33pixelmaS_a_i_n_t: that's how I understood your last statements but just now realised you might have been referring to pdbox
13:12:51 Join teru [0] (
13:13:26JdGordon1n1s: after checking audio_status() == 0 and calling audio_get_buffer() (and then the buffer alloc), how do I get playback to reinit? audio_restore_playback(AUDIO_WANT_PLAYBACK) ?
13:14:45 Quit FOAD (Ping timeout: 276 seconds)
13:15:46S_a_i_n_tHmmm, plugindex shows that pdbox isn't available on the Nano (which I assume is 1G as it only says nano) and I distinctly remember it building for my Nano1G also, yet it never got included on device then either.
13:16:09S_a_i_n_tWhy does it build it if it's not going to get included?
13:16:12n1sdunno, actually the talk code seems to only call audio_get_buffer and then buffer alloc, that ends up with playback stopped and the user has to resume manually
13:16:33n1s(this is what happens when loading new langs/voicefiles
13:16:52pixelmaIIRC it started off as an H300 plugin only, UI being the critical part. This year it got an e200 port
13:17:02pixelmait = pdbox
13:17:17S_a_i_n_tpixelma: So, it *shouldn't* be building it?
13:17:23 Join MethoS- [0] (~clemens@
13:17:56 Join FOAD [0] (~dok@
13:17:59S_a_i_n_tMy compile time would certainly be a fair ammount shorter for a "make from make clean"
13:18:00pixelmaI don't think it should but I'm really not sure
13:18:06S_a_i_n_t+if it didn't build it.
13:18:31pixelmaand I can't remember seeing it being built on any of my targets
13:18:54pixelmawhat's plugins/SUBDIRS saying?
13:19:16S_a_i_n_tapparently (according to pluginindex) it's only for H300 and 1~4G iPods
13:21:24pixelmaas if the plugin's wiki page are updated regularly... ;)
13:21:35JdGordon1bloody hell! actually I tihnk there was a bug in the fm screen!
13:21:41S_a_i_n_tit builds because it's defined with the same CONFIG_KEYPAD that the NAnos use
13:22:09S_a_i_n_tSo, this seems to need some work, and I'm not great with ifdefs
13:23:23pixelmamaybe it works on the 2nd gen Nanos?
13:23:44S_a_i_n_tthat's what I'm actually bitching about here.
13:23:55n1syeah, why shouldn't it work on nano2g?
13:24:10S_a_i_n_tI just remembered it built for, but wasn't included in, the Nano1G builds also.
13:24:13n1sif the ui works on other ipods with the same physical controls and smaller displays
13:24:36S_a_i_n_tI have no idea if it should/shouldn't work...but it doesn't get included from what I can see.
13:24:37 Join Topy44 [0] (
13:24:52S_a_i_n_tI haven't managed to slap eyes on it in the player once.
13:24:53CIA-7New commit by jdgordon (r28297): (Partially) Fix FS #11572 - no backdrop in the FM screen..\n the preset question would clear it and not reset it, so reoder the code so it works
13:25:35JdGordon1bah, \n was supposed to be a new line...
13:26:26n1si think you need to use an actual text ecditor to get newlines
13:26:32CIA-7r28297 build result: All green
13:27:02JdGordon1how does look? specifically around lines 69+ ?
13:27:23JdGordon1if audio is stopped it will load backdrop buffers from there instead of the skin buffer
13:27:30JdGordon1if it needs to of course
13:27:35 Nick JdGordon1 is now known as JdGordon (
13:27:54 Join hebz0rl [0] (
13:28:15n1sS_a_i_n_t: from a quick test build the pdbox plugin is included for nano2g in the zip, it's a viewer though so you wouldn't see it in the plugin list
13:28:34S_a_i_n_taha! well that explains a lot.
13:28:35 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
13:40:42CIA-7New commit by teru (r28298): add plugin_crt0.c to OTHER_SRC so that dependencies will be genereted.
13:42:25CIA-7r28298 build result: All green
13:44:54JdGordondoes anyone have general objections to allocing from the audio buffer if audio is stopped for the (rather large) backdrop images?
13:46:07gevaertsIf more than one thing starts to do this, this could get "interesting"
13:47:19JdGordonas long as rules are followed it shuold be ok... but yeah..
13:49:37gevaertsMaybe it's worth it to spend some time on an API that can be implemented by the advanced nice memory management systems that have been talked about, and use that, so when people actually get around to do any of this fancy stuff the code that uses it doesn't need (m)any changes
13:51:05JdGordonyeah, maybe
13:51:16JdGordondon't worry, this is going on FS first
13:52:13gevaertsI'd certainly hope it wouldn't go in during a freeze :)
13:52:15JdGordonalso, it looks like loading skins causes playback to be stop/started already anyway (because of AA chaning dimensions).. which kinda sucks
13:52:42gevaertswell, s/causes/may cause/, but yes
14:06:25 Quit antil33t (Read error: Connection reset by peer)
14:06:35 Join antil33t [0] (
14:17:12kugelS_a_i_n_t: is what I found in one of my branches
14:17:32S_a_i_n_tkugel: works for you?
14:17:35 Join petur [0] (~petur@rockbox/developer/petur)
14:18:26kugelI believe yes but I can't remember exactly
14:23:44S_a_i_n_tkugel: If line 77 is commented (in the pastie you linked), then is 78 necessary?
14:28:12kugelmaybe not
14:28:48S_a_i_n_tlines 77, 78, and 79 were added as the "fix" for the dropouts, but only tested on the Fuse
14:29:06S_a_i_n_t(lines from your pastebin, not the .c file)
14:30:09 Join b0hoon [0] (
14:33:04amiconnHmm, seems we have a plugin dependency problem
14:36:42 Join fragilematter [0] (~fragilema@
14:37:38 Join anewuser [0] (kvirc@unaffiliated/anewuser)
14:37:43CIA-7New commit by peter (r28299): update Dutch language file
14:38:07 Quit soap (Quit: soap)
14:39:18CIA-7r28299 build result: All green
14:40:42 Quit dys (Ping timeout: 276 seconds)
14:43:27CIA-7New commit by b0hoon (r28300): HDD6330: Configure touchpad (maximum touch sensivity, reduce transmission ...
14:44:30 Join soap [0] (
14:44:30 Quit soap (Changing host)
14:44:30 Join soap [0] (~soap@rockbox/staff/soap)
14:44:48 Quit teru (Quit: Quit)
14:44:50 Join dys [0] (
14:44:51 Join wodz [0] (
14:45:54wodzamiconn: I noticed recently that doing make rocks on fresh svn checkout fails. Doing simply make works
14:51:22JdGordonoh this is no good! r28000 apparently isnt only hitting fuzev1's
14:51:26JdGordonFS #11684
14:52:15fragilematterhey guys, I just posted FS #11684
14:53:13fragilematterif you need any additional testing/help, I'll be happy to provide it
14:53:41b0hoonhmm, what's wrong with my commit?
14:54:10 Join robin0800 [0] (
14:54:41*b0hoon is impatient
14:58:28wodzAlexP: ping
15:03:19wodzAlexP: please test FS #11641 - I am interested if 1) it works 2) you see speedup from previous version on coldfire
15:03:32AlexPOK, can do in about 20 mins
15:07:28***Saving seen data "./dancer.seen"
15:18:29 Quit Topy44 (Ping timeout: 265 seconds)
15:20:14 Join Topy44 [0] (
15:21:12TheSevendoes anyone know how to trigger a panic on the nano 2g?
15:21:22TheSeveni need one to test my crash dumper :)
15:21:36bertrikwe could write a plugin for that :)
15:21:44TheSeveni decreased the main thread stack size to half of what it was, and it's running fine :/
15:21:47gevaertsTheSeven: revert your local changes to the screen dump thing :)
15:22:39TheSevenok, I found a stkov
15:22:44gevaertsAh, good
15:22:51TheSevenwhen adding a track to a new playlist
15:23:34JdGordonnot so theoretical after all! :p
15:23:36TheSevenhm, no USB activity
15:24:11JdGordonAlexP: you dont happen to have a link to the irc discussion re the release suggestion do you?
15:24:18JdGordonI'm writing an email now :)
15:26:02AlexPJdGordon: Good work :)
15:26:06AlexPAnd yeah, one mo
15:27:32AlexPJdGordon: ta, I've been spectacularly lazy
15:27:43JdGordonno worries :)
15:29:23 Join b0hoon_ [0] (
15:30:06 Quit b0hoon (Ping timeout: 276 seconds)
15:35:59 Join SIGSEGV2 [0] (~user@
15:37:40 Quit pjm0616 (Read error: Connection reset by peer)
15:42:19 Quit robin0800 (Read error: Connection reset by peer)
15:44:05AlexPwodz: It works just fine, but isn't quicker, in fact it may even be slightly slower
15:44:13 Join domonoky1 [0] (
15:45:51TheSeven936 bytes binsize, 1040 bytes ramsize, and it isn't even using thumb...
15:46:07wodzAlexP: interesting
15:46:14TheSevenbut it still doesn't work when taking over control from rockbox :/
15:46:29 Quit domonoky (Ping timeout: 252 seconds)
15:48:27jeplerah, the code that boots fuzev2 OF when plugged into USB seems to be in mkamsboot dualboot.S. the code I was looking at in bootloader/ is way later when deciding whether to enable a rockbox usb mode..
15:50:12gevaertsjepler: be careful when modifying that. If that goes wrong, you're well into advanced unbricking techniques
15:53:38 Join lxsys [0] (
15:54:34 Quit wodz (Quit: Leaving)
15:55:38lxsysHi. I've just fired up rockbox utility 1.2.8(SVN 28154) and it's telling me that there's a new version available, which appears to be the same version.
15:58:45S_a_i_n_tlxsys: Yeah, you can disregard that.
15:58:53lxsysOk cool.
15:59:11 Quit fragilematter (Quit: Leaving.)
16:00:20 Nick b0hoon_ is now known as b0hoon (
16:00:43gevaertsYes, there's a bug with the version checking
16:01:34lxsysI think little bugs like that are cute, even if they are slightly annoying.
16:02:35b0hoongevaerts: Are you able to check what is wrong with the build system after my commit?
16:02:57gevaertsOnly the Swedes have access to that
16:02:59b0hoonok thanks
16:06:11b0hoonBagder: ping
16:07:31jeplergevaerts: yes, I know
16:07:39jeplerin fact, I think I'll throw in the towel at this point
16:07:58jeplerthe learning experience so far has been a good one, but one misstep here and I think I am the proud owner of a brick
16:12:47 Quit simonrvn (Read error: No route to host)
16:22:13 Join Staphylo [0] (
16:24:01 Quit Staphylo (Client Quit)
16:24:20 Join Staphylo [0] (
16:25:36 Quit chattr (Quit: gone)
16:26:09 Quit krazykit (Ping timeout: 250 seconds)
16:28:02 Join krazykit [0] (
16:28:25 Join CaptainKewl [0] (
16:32:17 Quit JdGordon (Ping timeout: 252 seconds)
16:36:25lxsyshmm.... this is interesting... I've installed the latest svn build on my H10 5GB, and set the theme to icatcher. Now I can play music, and the theme appears normal. When I push back to go to the menus, there's bits of theme left everywhere.
16:39:36 Join timccc [0] (~tim@
16:40:27S_a_i_n_twhat are the screen dimensions of the H10?
16:41:10 Quit krazykit (Ping timeout: 272 seconds)
16:41:31 Quit FOAD (Quit: I'll be back)
16:42:59 Join krazykit [0] (
16:43:30bertriklxsys, do you ever get the problem where volume starts to increase by itself when the battery gets low?
16:43:44S_a_i_n_tiCatcher isn't really the most "well coded" theme there is on offer, but, that shouldn't be happening.
16:43:50lxsysS_a_i_n_t: The H10 5/6GB models have a 128x128 screen
16:44:43lxsysbertrik: I havent seen it yet as Ive only just put a more recent build on, but I'll keep an eye out.
16:44:49S_a_i_n_tconverting iCatcher to viewports would help this I suspect.
16:45:11S_a_i_n_tWe really need to get these themes (everything except cabbieV2) out of the main build.
16:45:26 Join FOAD [0] (~dok@
16:45:34S_a_i_n_tHas anyone got an idea of the themes/targets that are stopping this from happening?
16:47:30S_a_i_n_tThe fact that they are in the build makes me want to code them properly, but, that's just not going to happen anytime soon...if at all. preperations were made to shift all the "built in" themes to the themesite, but, I have no idea if a list exists of what actually got moved and what didn't.
16:47:43 Quit bmbl (Quit: Verlassend)
16:48:24soapWould converting it to viewports "help" the problem or hide the problem?
16:48:47S_a_i_n_thelp, I would suspect...there's no way to hide it.
16:49:48S_a_i_n_tand, fix other problems.
16:50:03S_a_i_n_teven if it didn't fix the problem in question
16:51:38S_a_i_n_tThe main problem I see with the "pre-viewport" way of doing things is that selecting a larger text can push text over images/animations and make it look really weird.
16:52:22S_a_i_n_tthat, and using spaces and newlines for alignment when you don't have to is just plain wrong.
16:53:00lxsysS_a_i_n_t: I have a video of the weird stuff left over the screen.
16:53:30S_a_i_n_ta screenshot would be nice, if you can manage to get one.
16:54:12soapI though the issue was graphics from one screen persisting into another one?
16:55:06soapWhich, IIUC, should not be happening even if one doesn't use viewports. Isn't the use of viewports _hiding_ the problem here by forcing a redraw over the entire viewport?
16:55:35soapI don't recall this problem existing before viewports, so how can the (relatively) modern invention of viewports be the cure?
16:55:43S_a_i_n_tHmmm, looking at it like that...quite possibly.
16:56:32 Quit krazykit (Read error: Operation timed out)
16:57:33 Quit lxsys (Disconnected by services)
16:57:52 Join lxsys [0] (
16:58:00lxsys[resent] S_a_i_n_t: is there a way to take screenshots in rockbox?
16:58:16S_a_i_n_tEnable screendump
16:58:24S_a_i_n_t(debug menu IIRC)
16:59:04 Quit fyre^OS (Quit: Ur skills' fireproof like a wooden panel -- U got feds talking leet on your IRC channel!)
17:00:06 Join krazykit [0] (
17:06:25 Quit Dreamxtreme (Quit: Never look down on someone unless you're helping them up.)
17:07:30***Saving seen data "./dancer.seen"
17:07:30lxsysS_a_i_n_t: hmmm... I've got screendump enabled but I can't see any files in the root dir that are screendumps
17:07:52S_a_i_n_tyou need to plug a USB cable to "take the screendump"
17:08:21lxsysoh ok
17:08:58 Join Dreamxtreme [0] (~Dre@
17:11:52 Quit leavittx_ (Ping timeout: 240 seconds)
17:13:04 Join factor [0] (
17:13:28pixelmalooks like the WPS backdrop appearing in the menu
17:13:49pixelmawhich version of Rockbox did you install exactly?
17:14:45lxsysSVN 28300
17:17:38pixelmadoes it also happen with other themes?
17:19:15lxsysI tried on "rockboxed" and it left used bitmap over my screen
17:19:54S_a_i_n_twhat is the last build you had on the device that *didn't* display this behaviour?
17:20:25lxsysI put 234something on there ages ago, that worked fine
17:20:59S_a_i_n_tyes, that was indeed ages ago. (if 23xxx is correct)
17:21:12lxsyscabbiev2 seems to work ok
17:21:46pixelmathat's weird as it also has a WPS backdrop
17:21:49 Join bmbl [0] (
17:21:50 Quit bmbl (Changing host)
17:21:50 Join bmbl [0] (~bmbl@unaffiliated/bmbl)
17:23:28lxsysI have to go in a minute, its my uncle's 3C'th birthday. :(
17:23:37pixelmalxsys: could you try e.g. yesterday's daily build (not sure how you updated)?
17:23:56lxsyspixelma: yeah ok downloading it now before I go
17:26:00lxsysOk i'm off now, I'll check back later and report success/fails.
17:26:04 Quit lxsys (Quit: back later)
17:34:18 Join Luca_s [0] (
17:34:37 Nick Luca_s is now known as Luca_S (
17:44:03 Join grawity [0] (
17:46:43 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
17:59:08 Join CGL [0] (~CGL@
18:02:08 Join kevku [0] (
18:04:36TheSevenbinsize: 832 bytes, ramsize: 904 bytes...
18:05:09TheSevenand completely independent of the rest of the rockbox binary
18:05:49TheSevenbut for some weird reason it still doesn't connect when triggered through a panic
18:05:56TheSeven(it does when run through embios though)
18:08:14S_a_i_n_tgevaerts: ping?
18:08:28 Quit krazykit (Ping timeout: 264 seconds)
18:08:32S_a_i_n_tI should be in bed, but I'll forget if I don;t mention it now.
18:08:51S_a_i_n_tthe resistor plugin has some quirks, I see you've been workin gon it.
18:09:26 Join earcar [0] (
18:09:27gevaertswhat sort of quirks?
18:09:32S_a_i_n_t1: - No way to "har exit" the help text (must go from start to finish...annoying if accidental)
18:09:40S_a_i_n_t*hard exit, even
18:10:24S_a_i_n_tyou also can't scroll to a previous page in the text, which I find odd, every keypress goes to the next page.
18:11:08S_a_i_n_t2: LED Resistence flashes the colour code for a millisecond, then exits.
18:12:09S_a_i_n_t3: - first keypress is eaten after exiting the LED resistence calc back to the resistor plugin menu
18:12:57S_a_i_n_t(this is all on the Nano2G btw, my only device presently)
18:14:28n1spixelma: cabbiev2 also has a menu backdrop which i guess most other shipped themes lack
18:14:46pixelmaah, true
18:14:51gevaertssasquatch: 2 and 3 sound like FS #11675
18:15:05gevaertsS_a_i_n_t I mean...
18:15:17S_a_i_n_tI've been called worse ;)
18:15:51gevaertsI'll try to improve :)
18:16:28 Quit Staphylo (Quit: Bye les gens =))
18:16:55 Join Staphylo [0] (
18:17:08S_a_i_n_tgevaerts; Aha, yes.
18:17:23S_a_i_n_t2 and 3 are indeed FS #11675
18:17:54S_a_i_n_tsorry, I was just messing with the plugin, and am very tired so I needed to get it out before I forgot it all again.
18:18:12 Join krazykit [0] (
18:18:26 Quit krazykit (Client Quit)
18:22:29 Join GLiBERN [0] (
18:24:46TheSevenhm, what could be preventing the post-mortem stub from enabling USB?
18:24:58TheSeventhe code works fine when executed from embios, but not from a rockbox panic
18:25:21TheSeventhe neccessary clocks all seem to be enabled
18:25:31TheSevenresetting the OTG succeeds
18:25:38*grawity gives up trying to take comparison photos for the LCD issue.
18:25:47TheSevenbut the host doesn't even see that a device is plugged
18:26:21TheSevengrawity: did the build without that patch work properly?
18:28:04 Quit grawity (Read error: Connection reset by peer)
18:28:37kugelhm strange
18:28:54 Join grawity [0] (
18:32:07 Join krazykit [0] (
18:32:08 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
18:32:45kugelthe asm optimizations in mp3 are faster on fuzev1 and fuzev2, but not on my phone
18:33:50*TheSeven is out of ideas
18:34:03TheSeven[18:26] <TheSeven> grawity: did the build without that patch work properly?
18:34:11grawityTheSeven: yep
18:34:29TheSevenso it's definitely that patch?
18:34:31*grawity outputs some curses in the general direction of his flaky power cable.
18:34:34kugelboth fuzev2 and my phone are armv5 (well, phone is actually v6 but rockbox is built for v5)
18:34:35grawityTheSeven: yes
18:35:26TheSevenkugel: maybe more pipeline stalls because of new constraints on v6?
18:35:28n1skugel: afaiu arm v6 and armv5 cores are quite different with regards to stalling and register interdependencies
18:36:02grawityTheSeven: maybe the voltage change could be done only for non-ILI9320 displays?
18:36:07kugelit would be nice if someone checked it on the beast
18:36:32grawityTheSeven: although I don't completely understand what is the "buzzing sound" issue
18:36:42TheSevengrawity: in theory, yes. actually i'm more in favor of reverting that whole thing
18:36:59n1skugel: any codec in specific
18:37:45TheSevengrawity: there are at least three lcd types, with only one of them tested with that patch
18:38:09kugeln1s: mp3
18:38:19kugeln1s: disables the asm for it
18:38:22n1skugel: i'll try to test later
18:39:44bertrikWould anyone here like to help with getting a radio signal-strength indicator into the FMS?
18:41:11grawityTheSeven: is the noise being output to the outside world directly, or to the earphones? (because if it's the former, then there's no difference −− and no noise either −− on my device.)
18:41:33TheSevengrawity: no idea, ask Buschel
18:41:54TheSeveni haven't ever noticed noise on my LDS176 either
18:43:11kugelspeaking of fuzes, I cannot reproduce FS #11608
18:43:15 Join fragilematter [0] (~fragilema@
18:44:43n1sit seems no developer can
18:44:57 Join {phoenix} [0] (
18:46:14 Join pamaury [0] (
18:46:14 Quit pamaury (Changing host)
18:46:14 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
18:49:27 Join leavittx_ [0] (~lev@
18:49:30kugeljust got an undefined inst
18:50:23 Quit antil33t (Read error: Connection reset by peer)
18:50:25n1shmm the beast's backlight doesn't come on after usb disconnect it seems
18:50:35 Join antil33t [0] (
18:54:26n1slibmad with asm is faster on the beast, 12.00s to decode the 128k test file with asm and 14.29 without asm.
18:56:46 Quit z35_11 (Quit: Leaving)
18:57:02 Quit z35 (Remote host closed the connection)
19:03:33kugeln1s: any idea how to explain that?
19:04:00 Quit factor (Read error: Connection reset by peer)
19:04:34TheSevenkugel: asm-enabled armv6 builds are slower than non-asm armv6 on your phone?
19:04:54kugelwell only mp3 is slower
19:05:46TheSevenmaybe your phone has faster RAM than the beast?
19:06:19 Join krabador [0] (
19:06:52TheSevenhuh? how can armv6 builds be slower than armv5 ones for some codecs?
19:07:02kugelI wondered that too
19:07:31***Saving seen data "./dancer.seen"
19:09:45TheSevenwv_normx4.wv looks as if some columns were mixed up
19:11:04TheSevenand why does nero/lame make such a big difference?
19:12:17soapbertrik, what kind of help do you request?
19:12:18TheSevenn1s: did you use the lame test file?
19:12:39kugelTheSeven: ah yes, I messed up wv_normx4.wv
19:12:58kugelswap the values for the armv6 tests
19:13:46 Quit b0hoon (Ping timeout: 255 seconds)
19:14:30bertriksoap, getting some tags for the fm signal strength into the parser and the renderer, and creating a wps for the fms using the new tags
19:14:31 Join b0hoon [0] (
19:14:54soapahh, jd work
19:15:20*TheSeven is still clueless why his post-mortem-dumper fails to initialize USB
19:18:54 Quit Staphylo (Quit: Bye les gens =))
19:21:37 Join factor [0] (
19:25:18*kugel grmls at FS #11676
19:26:04 Quit krabador (Read error: Connection reset by peer)
19:26:45 Join krabador [0] (
19:26:56 Join saratoga [0] (463f90ed@gateway/web/freenode/ip.
19:27:29TheSevenkugel: is it really a feature that it unpauses after the call ends if it was paused before the call stareted?
19:28:25kugelhm, I read that wrong
19:28:39kugelI implemented it so that it only resumes if the audio was playing
19:29:03TheSevenat least i interpreted that report differently
19:29:06saratogafor what its worth, the libmad asm is probably scheduled all wrong for arm11 and cortex A8
19:29:17kugelresume = (audio_status() & AUDIO_STATUS_PLAY) != 0;
19:29:21saratogai don't think the armv5 or armv6 instructions make any difference though
19:29:35kugelso, either audio_status() or the reporter is wrong
19:32:19TheSevenkugel: that should be easy to check :)
19:33:27kugelhm, it seems audio_status() gives both PLAY and PAUSED if paused
19:34:28 Quit leavittx_ (Ping timeout: 240 seconds)
19:42:47*TheSeven needs a new lcd for his nano2g
19:43:28 Quit saratoga (Quit: Page closed)
19:43:44n1sTheSeven: yes, used the 128k sample from test_files
19:43:57TheSevenlame_128 or nero_128?
19:44:21n1skugel: no, not really any idea, is the android sdk using about the same gcc as we do?
19:44:39 Part GLiBERN ("Leaving")
19:44:48kugelit's using 4.4.0, with some modifications
19:44:48n1sTheSeven: lame, nero is aac
19:44:53TheSevenoh, right
19:44:57TheSevenso maybe their compiler is just more clever than ours?
19:45:57TheSevendamn, at least 34 bad cells on that LCD now
19:47:27n1skugel: eh, is the android port setting the CPU var in the build system, many codecs have tuned optimization flags based on that
19:52:38CIA-7New commit by kugel (r28301): Fix FS #11676 - audio_status() returns AUDIO_STATUS_PLAY|AUDIO_STATUS_PAUSE if paused so only resume if AUDIO_STATUS_PLAY is the only bit.
19:52:56 Join fragilematter1 [0] (~fragilema@
19:53:24kugeln1s: I set CPU_ARM and ARM_ARCH for those tests
19:53:47kugelis there another var that I need to set?
19:54:01 Quit fragilematter (Ping timeout: 276 seconds)
19:54:02n1skugel: the buildsystem has it's own variables, set in the makefiles that configure generates
19:54:12kugelah I see
19:54:21n1sthose are tested to set optimization options for codecs
19:54:23kugelI cannot set this since it's used for the target tree
19:54:32CIA-7r28301 build result: All green
19:55:01n1sah, you can test hardcoding O1 for libmad though to see if this is the cause
19:59:34kugelcan I do || in makefiles?
20:00:03kugele.g. ifeq ($(CPU),arm) || ($(CPU),android)) ?
20:06:13 Join simonrvn [0] (
20:06:55 Join lxsys [0] (
20:07:31lxsysFollow up to the issue with themes: yesterdays daily build worked for me
20:09:17 Join _s1gma [0] (~d.d.derp@
20:09:26gevaertslxsys: it was a build from today that had the problem?
20:09:50lxsysI belive so.
20:10:09lxsysI'll take a closer look after tea, but I'm pretty sure its a recent build.
20:10:30gevaertsHaving exact revisions for the working and non-working builds would be useful
20:10:56gevaertsDo you have this working build installed now?
20:11:18lxsysI have yesterdays daily build installed and it appears to be working
20:11:39gevaertsCan you check the exact revision of that one?
20:11:56TheSevenkugel: that line will at least need another ifeq
20:12:05gevaertsSystem->Rockbox Info will have that
20:12:28lxsys28284 is the working build
20:13:58 Join moos [0] (moos@rockbox/staff/moos)
20:14:15lxsysbrb, tea
20:15:05 Quit stoffel__ (Remote host closed the connection)
20:19:30gevaertslxsys: could you also try (or have you tried) this morning's daily, r28294?
20:19:45gevaertsThat would reduce the number of possible culprits a bit
20:22:19 Quit GeekShadow (Quit: The cake is a lie !)
20:24:18 Quit Luca_S (Quit: CGI:IRC (EOF))
20:25:09kugeln1s: -O1 is *a lot slower* (asm not enabled)
20:25:45kugelthe % realtime drops by 500-800
20:29:03kugelheh, but now the asm is faster :)
20:29:14kugelstill slower than plain C with -O2
20:29:37 Part fragilematter1 ("Leaving.")
20:30:25*kugel tries -O2 on the fuzes
20:30:46CIA-7New commit by moos (r28302): Update the polish translation. ...
20:31:29lxsysgevaerts: ok I'm downloading now
20:31:32 Join s1gma_ [0] (~d.d.derp@
20:31:49grawityAww, crap. Forgot a / in foreach
20:32:33CIA-7r28302 build result: All green
20:34:20 Quit _s1gma (Ping timeout: 245 seconds)
20:35:14kugel-O2 seems faster on fuzev1 as well
20:45:52 Join leavittx_ [0] (~lev@
20:54:08lxsysgevaerts: I think I found the issue.
20:55:24lxsysthere isnt an issue with the build, as far as I'm aware
20:56:02 Quit CGL (Ping timeout: 260 seconds)
20:56:03 Quit TheSeven (Read error: Connection reset by peer)
20:57:01 Quit krabador (Read error: Operation timed out)
20:57:57lxsyswhen I install additional themes, they seem to be broken. I think I may have overwritten the working "icatcher" theme file
20:58:43 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
21:01:01 Quit moos (Quit: ciao ciao soccer party)
21:07:24gevaertshm, ok I guess
21:07:35***Saving seen data "./dancer.seen"
21:07:46 Join chattr [0] (
21:07:47gevaertsI'm still not sure that it isn't a bug though
21:12:01*lxsys goes to compare the theme files
21:24:14lxsyshmmm... I think I may have found it
21:25:41lxsysI've added changed "iconset:" to "iconset: -" and "viewers iconset:" to "viewers iconset: -" in iCatcher.cfg from and it appears to be working
21:26:19lxsysdisregard my added, I must have added it somehow.
21:27:48gevaertsr28286 and r28288 seem related
21:28:47gevaertsmainly r28288
21:28:49TheSeventhe stkov i used to test my post-mortem dumper just went away by itself?
21:28:57TheSevenhow can that happen?
21:30:09gevaertslxsys: I'm pretty sure that if recent changes influence this, it's r28288. I don't know that part of the code well enough to comment more though
21:31:01lxsysgevaerts: to me the issue appears to be in
21:31:37gevaertslxsys: yes, I think I agree
21:32:11lxsysalthough I'll try and get something before 28288 running
21:35:19 Quit Horschti (Quit: Verlassend)
21:37:03 Quit elcan (Ping timeout: 276 seconds)
21:44:34lxsysI've downloaded and installed R28282, then folled the same process that gave me errors with R28300, and it appears to be working
21:45:05gevaertsok, then my bet is that r28288 changes things a bit
21:46:34lxsysafter 28288 doing "iconset:" to "iconset: -" and the other one I mentioned, seems to also fix the issue
21:46:47n1skugel: that probably a) means we should tune for specific arch versions, not just one tuning per arch and b) beast svn uses O1, assuming that is the same as fo you and a lot slower than O2 it's no surprise the asm is better
21:47:11*n1s tries just for completness
21:49:32 Join fdinel [0] (
21:51:46 Join Horscht [0] (
21:51:47 Quit Horscht (Changing host)
21:51:47 Join Horscht [0] (~Horscht@xbmc/user/horscht)
21:51:52 Join saratoga [0] (9803c6dd@gateway/web/freenode/ip.
21:52:06saratogatried 3 different AMSv1 players and all decode ogg fine
21:52:12saratogaso i guess i'm no use looking into that bug
21:52:15n1sdecoding time on the beast for the 128k mp3 sample goes: O1 no asm > O1 + asm > O2 + asm > O2 no asm
21:53:05kugelso, O2 no asm is fastest?
21:53:12n1swith both with asm settings are pretty close naturally and the O2 no asm setting is about 10% faster
21:53:26kugelthat's what I experience on my phone too
21:53:57saratogatheres a couple different ASM bits that could be at fault
21:54:04saratogathe most likely is synth_full
21:55:17 Quit S_a_i_n_t (Ping timeout: 265 seconds)
21:56:27saratogayou could try disabling the code in synth_full_arm.S
21:56:34n1ssaratoga: yes, that uses a lot of ldr with the result used right after, that's a 2 c stall on v6, and most of the users of the ldr results seem to be smull/mslal which has the multiplicands as early regs so the stall is 3c IIUC
21:56:48saratogayes its a mess on later on arm
21:56:55saratogabut i wouldn't be surprised if other bits are bad too
21:57:10saratogathe mdct code is taken from the original libmad, circa 2003
21:57:22saratogai suspect it was not written with arm11 in mind
21:57:53n1si think arm11 was introduced around 2004 so you are probably right
21:58:36n1sbut if we disable this asm we also need per arm version compilation options
22:00:12saratogaits probably better to just disable it now on the beast, and then later on rewrite it with neon if it turns out that it matters on Android
22:01:10n1syeah but just disabling it makes libmad slower unless we use -O2... which is slower on arm7
22:01:15saratogaon arm11 is it bad to write back to a source register?
22:01:37saratogacheck for the beast model number in the makefile
22:02:04saratogaa lot of this looks bad on arm9 too
22:02:51n1show do you mean write back to a source register?
22:03:04saratogae.g. smull r6, r7, r10, r7
22:03:33kugeln1s: O2 vs O1 doesn't make a huge difference on arm9
22:03:41n1saha, i don't know, don't think i've read anything about that
22:03:59*TheSeven is observing weird things
22:04:28n1skugel: i don't remember the numbers but O1 was definitely fastest on arm7 maybe it should be retested to see how big the diff is
22:04:42TheSevenmy dumper is working flawlessly on my linux box (usb1.1) and not even enumerating on my laptop (windows 7, usb2)
22:05:38kugelsaratoga: the asm is a lot faster than the C on arm9
22:06:07 Quit t0rc (Remote host closed the connection)
22:10:07 Quit lxsys (Quit: leaving)
22:10:27kugelon my fuzev2 all combinations are about the same
22:10:27n1ssaratoga: since the source regs are "early regs" for multiplication instrs i assume the contents are already fetched by the time the actual execution starts so writing back to them shouldn't hurt
22:13:48 Quit froggyman (Ping timeout: 240 seconds)
22:14:05kugelok, fuzev2 (armv5) O1+no asm > O2
22:14:17saratogawhat does ldrd do
22:14:28saratogai'm baffled at the worthlessness of arms site
22:14:48kugelO1+no asm > O2+no asm > O1+asm == O2+asm
22:14:59n1sldrd loads a 64 byte value at once
22:15:06n1sso like a double ldr
22:15:07TheSevenbigger in terms of time needed or performance?
22:15:19kugeldecoding time
22:16:04kugelO1+no asm is noticeably slower, the others are almost equally fast
22:16:08 Join benedikt93|AFK [0] (~benedikt9@unaffiliated/benedikt93)
22:17:14 Join froggyman [0] (
22:17:14 Quit froggyman (Changing host)
22:17:14 Join froggyman [0] (~seth@unaffiliated/froggyman)
22:22:45 Quit domonoky1 (Read error: Connection reset by peer)
22:25:21saratogalooking at synth_full closer, I think we could actually use the armv6 stuff to make it faster
22:27:12saratogai think the D0ptr and D1ptr values could be stored in IRAM as packed 16 bit values with almost no loss of precision
22:28:53saratogathen the SMALxx type instructions could be used to do 32x16 = 64 bit multiplies on them
22:28:59saratogathis would at least halve the number of loads
22:31:46saratogais SMULWy much faster then doing a full multiply?
22:32:58 Quit bmbl (Quit: Verlassend)
22:35:57n1sSMULWy is 1c while regular mutiply is 2c on arm11 they have the same result latency
22:36:29 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
22:37:40n1sit's a bit weird that they didn't add a subtracting variant of smlal as every other weird multiply accumulate they added seems to have one
22:41:07saratogathe latency on all those mul instructions is huge
22:41:15saratogaall those back to back load/mul pairs must stall like hell
22:42:03amiconnNot because of the mul though
22:42:27n1smuls require multiplicands early so ldr-mul will stall for 3c
22:43:05n1smultiply-multiplyaccumulate only stalls for 1c
22:43:13saratogaoh yeah i misunderstood the sheet
22:44:10n1slrd r0, [r1]; smull r2, r3, r0, 5 takes 6c IIUC
22:44:23saratogabut if i understand correctly, if we could do packed 16 bit values, we'd be able to load one register, multiply the previous, then load the next, etc with little or no stall
22:44:24 Join wodz [0] (
22:44:52saratogasince we'd have enough mul instructions and registers to completely hid the load latency
22:44:59n1syeah i think so, a lot of stuff was added for dealing with packed 16 bit values in v6
22:46:04saratogathe only thing i'm not sure about is accumulating the filter into a 32 bit register
22:46:09saratogalibmad uses a 64 bit register now
22:46:20saratogai need to ask buschel if he thinks that actually needed
22:46:51n1sdoes the c code use a 64 bit acc too?
22:47:46saratogai think an 8 tap filter should be ok with 32 bit accumulation though
22:47:53 Quit t0rc (Ping timeout: 240 seconds)
22:48:03saratogathats not very many ads, and the filter taps are only 16 bits if I'm reading them right
22:48:26wodzAlexP: ping
22:48:57 Join rvvs89_ [0] (
22:49:10 Join t0rc [0] (~t0rc@
22:49:10saratogahuh the comments say they're 20 bit, but log2(max(D)/min(abs(D))) gives only 15 bits + 1 for the sign bit
22:49:14 Quit t0rc (Changing host)
22:49:14 Join t0rc [0] (~t0rc@unaffiliated/t0rc/x-5233201)
22:49:47 Quit rvvs89 (Ping timeout: 264 seconds)
22:50:46saratogaoh duh
22:50:54saratogayou need some bits to actually store the smallest value
22:51:42saratogaactually, since you know the filter taps in advance, I suppose you could do the smallest few in 32 bit precision and all the rest at 16 bit precision, thus avoiding rounding error
22:54:49 Quit t0rc (Quit: Give someone code, help them with one project. Teach someone to code, help them rule the world.)
22:54:52 Quit merbanan (Ping timeout: 272 seconds)
22:55:03saratogaheh, dct32 doesn't look great on arm either
22:56:15kugelsaratoga: dct32_arm.S seems unused
22:56:32saratogaactually i remember buschel saying something about that
22:57:04saratogawait AMSv2 is armv5e right?
22:58:22saratogaso we can use SMLALxy
22:58:37saratogathat gives some interesting posibilies
23:00:14 Quit DerPapst (Quit: Leaving.)
23:04:00saratogawhat is an "I cycle" in the arm reference manual?
23:04:23saratogawell besides "internal"
23:04:53 Quit petur (Remote host closed the connection)
23:07:38***Saving seen data "./dancer.seen"
23:08:32 Join robin0800 [0] (
23:09:51 Quit wodz (Quit: Leaving)
23:11:42 Quit kevku (Quit: KVIrc 4.0.2 Insomnia
23:14:51 Quit CaptainKewl (Ping timeout: 240 seconds)
23:17:41kugeln1s: couldn't we read mcpu switch in instead of $(CPU)?
23:18:09kugelthat way we get better info about the target cpu than just arm
23:20:30kugelhrm, we only ever use it for arm and mips so that doesn't work for sh/cf/x86
23:23:01 Quit b0hoon (Quit: - Chat comfortably. Anywhere.)
23:24:18 Quit {phoenix} (Remote host closed the connection)
23:25:44kugelif we switched to gcc 4.4 for all archs we could use #pragma gcc optimize instead of the makefile magic which enables preprocessor magic
23:27:54kugelthat would probably work nicer, also w.r.t. recompilation
23:29:03 Quit grawity (Quit: Good night.)
23:31:53 Quit earcar (Quit: bye)
23:34:39Horschthas the serial device code for ipod 5.5G 80Gb changed since rev. 100913?
23:35:30HorschtI just updated to the most current build and when I put my ipod in my alarm clock, rockbox doesn't start playing when the alarm goes off...
23:35:51kugelserial device code?
23:36:18 Quit chattr (Quit: gone)
23:36:27kugeldo you mean accessiory protocol? that hasn't changed in ages
23:36:30Horschtit used to work so I reverted to the build I had on before (100913), and my alarm clock seemed to simply send the "select button" code to the ipods serial port thingy
23:37:02Horschtkugel, the ipod accesory port
23:39:00Horschtso, it used to work like this: I put my ipod into my alarm clock and when the alarm goes off my ipod enters the menu that is currently highlighted on the screen.
23:39:54Horschtit does that with rev 100913 (maybe later ones) but not with the "latest" current build. With the current build it will simply sit at the main menu doing nothing
23:42:51 Join JdGordon [0] (~jonno@rockbox/developer/JdGordon)
23:42:55 Quit pamaury (Remote host closed the connection)
23:45:27 Quit n1s (Quit: Lämnar)
23:47:29 Quit benedikt93|AFK (Quit: Bye ;))
23:54:05pixelmarevision 100913?? Sounds more like a date...
23:54:22 Join CaptainKewl [0] (
23:57:54 Quit bertrik (Quit: :tiuQ)
23:58:15Horschtoh wait...
23:58:25Horschtyou're right :D
23:58:54Horschtcurrently syncronizing my music library, gonna get the rev. after it's finished

Previous day | Next day