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

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

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

#rockbox log for 2010-06-11

00:02:42***Saving seen data "./dancer.seen"
00:11:33 Join fdinel [0] (
00:14:49 Quit jgarvey (Quit: Leaving)
00:17:52bluebrotherbieber: if a new windows binary is required you can always ping me (as well as fml, but afaics he already left)
00:18:42 Quit pamaury (Remote host closed the connection)
00:20:45bieberI'd like to try and get a complete build environment set up in my Windows VM, I've just got to figure out how to statically link in MinGW's runtime
00:21:29 Quit MethoS- (Remote host closed the connection)
00:22:02bluebrotherthe windows build environment is a bit ... well, in a change these days. MinGW now has an official release of gcc 4.5, but the new installer isn't ready yet (and still defaults to gcc 3). Qt, OTOH, prefers gcc4 with Qt 4.6
00:23:04bieberI remember that I unpacked all the MinGW executables by hand, but I think it was a little bit shy of 4.5, but it was 4.something
00:23:26bluebrotherI also have a setup with gcc 4.4 (which was called RC IIRC)
00:24:31bluebrotherthough I've got a gcc 4.5 in my current setup.
00:24:51bieberAnd 4.5 automatically static-links the runtimes?
00:25:01bluebrotherI'm planning to adjust the script a bit so its easier to get reused with the theme editor.
00:25:13bluebrotherwhich right now basically works.
00:25:32bluebrotherno, Qt defaults to dynamic linking. Static linking is rarely used (like in our case)
00:25:54bieberI've got Qt recompiled for static linking (that took a LONG time), it's just MinGW I have to worry about
00:26:11bluebrotheryou need to rebuild Qt for static linking using configure -static. To avoid the dependency to mingwm10.dll you also need -no-exceptions
00:26:41bieberOhh, okay
00:26:44bluebrotherthough I _think_ things changed here with gcc 4.5. Haven't found the time to look into that yet though.
00:27:06bieberSo if I just rebuild Qt with -no-exceptions in addition to -static, I'll get completely standalone executables?
00:27:57bluebrotheryou should, yes.
00:28:08bieberOkay, I'll get that started compiling, then
00:28:20bluebrothermight be the easiest to simply copy the line from the RockboxUtilityDevelopment wiki page though :)
00:28:34bieberOh, that would help :)
00:28:52bluebrotherwebkit takes a looong time (around the same amount as Qt itself), and you most likely won't need phonon as well.
00:29:23bluebrotheralso, the examples and demos take ages as well. On Linux you can use -nomake demos -nomake examples. Not sure if that works on Windows too (IIRC it didn't in the past)
00:29:50bluebrotheron Windows you can simply run make sub-src sub-tools instead of using the default target. Then you won't get examples and demos compiled.
00:30:08bluebrotherhmm, iirc the Qt 4.6.3 build with MSVC I did recently accepted -nomake
00:31:09bluebrotherbtw, I've updated the Windows binary of the theme editor. In case someone wants to try :)
00:31:58bieberIs RBUtil automatically built and uploaded, or do you just do it manually?
00:32:51bluebrotherit's done manually. Building is automated with
00:33:23bluebrotherI have a few scripts around it to upload it to my webspace, but those aren't public (they contain my passwords :)
00:34:08bieberOoh, that looks handy
00:34:32bluebrotherthe theme editor binaries I did are done with btw. It needs a few adjustments, but luckily I wrote the script with the hope to reuse it for a different project later.
00:35:16bieberCould you add that to utils/themeeditor in the SVN tree? Or perhaps in a /tools subdirectory?
00:36:11bluebrotherI want to get some things merged here. Haven't figured the best way to do that though.
00:36:31bluebrotherbut I'm definitely planning on doing some stuff there :)
00:37:23bluebrotherright now I'm wondering if the separation in rbutil/ and utils/ does make sense anymore at all. Was thinking about that in the past before, but haven't gotten around doing anything about it.
00:37:34bieberbtw, did you receive my email about the schedule change?
00:37:45bieberAnd yes, I did wonder why rbutil was in the root instead of /utils
00:38:05bluebrotherfor example, when it comes to zip handling it wouldn't make much sense to simply copy the classes from the rbutil tree. A common one would make much more sense.
00:38:20bluebrotherthat's simply historical :)
00:39:35bluebrotherwe didn't have utils back then
00:39:50bieberOh, okay
00:40:01bluebrotheryes, got that mail. Was busy with RL stuff this evening so didn't got a chance to reply earlier.
00:46:23bluebrotherbieber: just send you a reply for completeness. Not that I had anything surprising new to say :)
00:47:58bieberHave you been generally satisfied with the editor so far?
00:48:23bluebrotheryes, you're progressing very well IMO
00:50:57bluebrotherjust spotted a minor problem: the hilighting of the line with errors only hilights the first line that contains an error. Also, I find the red background rather disturbing. Now that there's line numbering in place, how about making the line number background red instead of the whole line?
00:51:23bieberI can do that
00:51:56bluebrotherbut IMO that's minor. Just looked for something to comment on ;-)
00:52:15bieberThe single-line issue is because the parser stops when it encounters an error. I suppose I could feed it the remainder of the text starting at the next line as a new document, and see if any errors occur there, and etc. until it gets to the end of the document
00:52:23bluebrotheras always, feel free to discuss such things with the community. My opinion is only my opinion :)
00:54:01bieberThe line-number highlighting is definitely a good idea, now that the line numbers are there
00:55:04bluebrotherhmm. The status line showing only the first error feels also a bit strange to me. How about showing the validation of the current line instead? The error lines are hilighted anyway, so finding them (i.e. the line number) isn't absolutely necessary.
00:55:37 Quit tchan (Quit: WeeChat 0.3.3-dev)
00:56:20bieberSounds good
00:56:24bluebrothernice. Reordering the tabs via drag and drop works :)
00:56:39bieberThat one is all Nokia's work ;)
00:56:52bieberI've been really impressed with Qt4
00:56:55bluebrotherhehe, still nice. One of the reasons why I like Qt
00:57:50bieberI just can't wait to get started in on the rendering. Their 2D graphics API is really impressive
00:58:07bluebrotherI'd really like to have more opportunities to work with it. But well, that's how it goes when you have a day job.
00:58:42 Join Rob2222 [0] (
01:00:23bieberQt has some nice examples to play around with, if nothing else
01:00:51 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
01:01:37 Quit Rob2223 (Ping timeout: 264 seconds)
01:02:34 Quit davharster (Quit: Leaving.)
01:07:16 Quit ender` (Quit: If I had only finished this sentence,)
01:07:22 Quit CIA-6 ()
01:14:04 Quit efyx (Remote host closed the connection)
01:14:07 Quit kugel (Remote host closed the connection)
01:16:14 Part domonoky
01:20:10 Quit liar (Ping timeout: 240 seconds)
01:24:01 Join kramer3d [0] (
01:24:02 Quit kramer3d (Changing host)
01:24:02 Join kramer3d [0] (~kramer@unaffiliated/kramer3d)
01:24:10 Join evilnick_BSMM [0] (~620ec24b@rockbox/staff/evilnick)
01:30:10 Quit bieber (Ping timeout: 240 seconds)
01:30:25 Join bieber [0] (
01:34:35JdGordonbieber: ok cool :)
01:36:43 Quit DerPapst (Quit: Leaving.)
01:40:09JdGordonbieber: also, I realised the line number is actually useful for a few tags so keeping it is probably better
01:40:32JdGordonalthough the ones that need it need the viewport line offset, not document, but we can work around that
01:46:01 Join CIA-6 [0] (cia@
01:51:22 Part toffe82_
02:02:44***Saving seen data "./dancer.seen"
02:14:08bieberJdGordon: I'd like to have line number in there at very least for display in the parse tree view, but feel free to #ifndef ROCKBOX all the line-number handling code if you don't need it in ROCKBOX
02:14:33JdGordonfor the time being I'll keep it
02:15:33JdGordonhave you given anny thought to your renderer yet? are you going to make it fully capable with sublines and condiotionals and all that?
02:17:22bieberI'll be dividing tags up into the type that control the rendering and the type that return values, so conditionals can just evaluate the tag they're fed and decide which option to select from it
02:18:18JdGordonbefore you get too far I think we need a doc to specify exactly how things should be done re sublines/line drawing... I'm going to be redoing the rockbox renderer from scratch also so we need to both do the same thing
02:18:27bieberMy plan is to implement one big settings panel that will let you specify anything you want about the current status of the player
02:19:28bieberSounds good there. You don't have to worry about me getting too far ahead: I'm going to solidify the rest of the GUI more before I start on rendering, so I've got a little extra time now before I get started on it. And once I do, my first task will be just getting the groundwork laid and displaying viewports,
02:19:30*JdGordon should probably write down at least how the current renderer works so we can get close to it
02:20:07bieberDo you want to enable %t in conditionals, btw? I've heard it's a feature that's been missed. I don't know about the Rockbox renderer, but it'll be pretty simple to implement in the theme editor
02:20:11JdGordonI was wondering if the conditional and subline handling code should be done as shareable code? or if the doc is good enough?
02:20:31 Quit notlistening (Ping timeout: 248 seconds)
02:20:42JdGordonyou mean subline in conditional options?
02:21:14bieberLike, allowing you to change the time a specific subline is displayed depending on the value of a conditional
02:21:46bieberA doc should be good enough. My entire rendering system is going to be based on Qt's graphics API, so it'd be really difficult to reuse any of it in Rockbox, I'm afraid
02:23:51JdGordon%t(%?mh<10s|5s>) somethine like that?
02:24:51bieberMore like
02:24:51bieber%?sometag<%t(1)|%t(2)|%t(0)>Line 1;Line 2
02:25:10 Quit storm` (Ping timeout: 240 seconds)
02:25:28bieberWhere the %t only counts if it actually gets evaluated. I remember S_a_i_n_t talking about having that feature in a previous version, and it being useful for disabling sublines based on conditional values
02:26:00JdGordonthat looks a bit messy... isnt %t linked to the current subline and not the whole line?
02:27:05JdGordonin svn it is
02:27:39bieberYes, that's the point
02:27:41JdGordonI think %?aa<a;b;g|f> is a bigger want
02:27:59bieberSublines are alread allowed inside conditionals, no worries there
02:28:07bieberI actually just fixed some bugs in parsing that kind of situation today
02:28:21JdGordoncool, pretty sure they are illegal in svn :)
02:28:37 Quit Zarggg (Quit: Zarggg)
02:28:42JdGordontimeout in a conditional like that wont work... when is it checked?
02:28:59 Join Zarggg [0] (
02:30:25 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
02:30:41bieberHmm, that would be problematic
02:32:06bieberGrab the newest theme editor from SVN and check out the parse tree for "%?bl<A;%t(5)B;C|ABC>"
02:32:19JdGordonok, one min
02:32:50bieberBefore the parse tree for even conditionals within sublines was horribly messed up
02:33:24JdGordonlooks good
02:35:00JdGordoneven that is going to be interesting to time correctly...
02:35:10JdGordonor maybe not...
02:36:17JdGordon(nitpicking) Is there a better word than "Alternator" for it? and can you set the default column width for the parser to have the first colum almost the whole width?
02:36:36 Quit hamish_ (Quit: Leaving)
02:36:45JdGordonand have it auto-expand?
02:36:55bieberYeah, I'll probably switch it to sublines or something
02:36:59JdGordon(and follow the mouse/cursor)
02:37:13bieberPrettying up that parse tree viewer will be one of my big goals over the next couple days
02:37:16JdGordonclicking on a tag and having it select that tag in the tree would be sweet
02:37:23JdGordonbut totally useless for users I think
02:37:41bieberMy big pet peeve is just having to resize it every time
02:38:02 Quit sinthetek (Ping timeout: 260 seconds)
02:38:04JdGordonI wouldnt really bother spending too much time on the tree view, I don't think users need it and it should be disabled once this is finished
02:38:25bieberHmm, that is true
02:38:31bieberIt's hard to imagine a really good use case for it
02:38:37bieberOther than looking for parser errors
02:39:00JdGordonno, looking for missing | or ; 's
02:40:10 Join sinthetek [0] (~sinthetek@unaffiliated/sinthetek)
02:41:57bieberOh yeah, there is that
02:42:20bieberI'll probably just have it hidden by default, but with the menu option still there
02:43:26bieberOh, and wrt sublines in conditionals, I'd just re-evaluate all conditionals every time some relevant player state changes. If a set of sublines gets switched out then just stop all their timers and bring in whatever goes there according to the new state
02:44:25JdGordonthats pretty much how sublines timeouts are done now... the subline timer is paused
02:44:54JdGordonexcept ; is illegal inside a conditional in svn :)
02:45:46bieberIn the current Rockbox parser or mine?
02:45:48JdGordonIt had a fun bug when sublines where inside a conditional viewport, the timer kept going while it was hidden.... so when it was re-enabled it would quickly shoot through all the options it missed
02:45:57JdGordonsvn = current parser
02:46:15bieberAh, okay
02:46:44bieberDo you want to make it work in the new parser?
02:46:58JdGordonmake what work?
02:48:21bieberSublines inside conditionals
02:48:47JdGordonit does work, it was always a misssing feature
02:49:27bieberOh, okay
02:49:32bieberSounds good
02:49:34JdGordonlooking at the parse tree it is going to be dead simple to render :)
02:49:53bieberWant to just make a Wiki page to set down all the rules for rendering
02:51:11JdGordonyeah, I dont know how much coding time I'll have today so I'll do that
02:51:31bieberSounds good
02:51:32 Quit evilnick_BSMM (Quit: Page closed)
02:51:47bieberI'll be working more on the data models tonight, by about mid next week I should be starting on the rendering platform
02:54:00JdGordonok cool
02:54:17JdGordonbreaky time
03:09:50 Quit mt (Ping timeout: 240 seconds)
03:10:08 Join mt [0] (~mtee@rockbox/developer/mt)
03:11:45 Quit bluebrother (Disconnected by services)
03:11:48 Join bluebroth3r [0] (~dom@rockbox/developer/bluebrother)
03:17:12 Join storm` [0] (
03:23:17 Join CaptainKwel [0] (
03:24:51 Quit mt (Ping timeout: 258 seconds)
03:25:32 Join mt [0] (~mtee@rockbox/developer/mt)
03:36:36 Join aim__ [0] (~hamish@
03:36:56 Quit aim__ (Client Quit)
03:37:40 Join simonrvn [0] (
03:42:15S_a_i_n_tbieber: What did I do? ;)
03:42:32S_a_i_n_ts/do/want/talk about/whatever/ :P this about me wanting conditional timeouts for sublines?
03:43:42S_a_i_n_tI read some of the backlog a little, seems to be that way. Can it be done?
03:44:41 Join anewuser [0] (anewuser@unaffiliated/anewuser)
03:44:51JdGordonyes and no...
03:45:19S_a_i_n_tis that yes and no for "Can it be done?"
03:45:29JdGordon%?aa<%t(1)a;%t(2)b|%t(2)a;%t(1)b> is how you'd have to do it
03:45:43JdGordonbut you can now put sublines in conditionals which is what you really want anyway
03:46:22S_a_i_n_tA conditional timeout would be cool also.
03:46:25JdGordonLlorean: does changeing the theme fix your problem?
03:47:19S_a_i_n_tlike %Nn<%t(1.0)|%t(2.0)>
03:47:30S_a_i_n_tthat's something I'd really like to have.
03:48:02S_a_i_n_tMore possabilities.
03:48:10S_a_i_n_tewwww (sp?)
03:48:35Dhraakellianso I'm curious about something... If I were to make a female ipod to male sansa connector cable, would there be a way to get a rockboxed sansa to work with an iPod dock (which would be supported by Rockbox on an actual iPod)?
03:48:39Dhraakellian...if that made any sense
03:51:06JdGordonS_a_i_n_t: that is imo a nonesense request, and not really workable for the renderer anyway... how is it supposed to handle when the tag changes?
03:51:10LloreanJdGordon: I'm suing the default theme, and someone has posted on the tracker a specific revision that causes it
03:51:23JdGordonyes, that rev is a skin change
03:51:55*S_a_i_n_t fails to parse "how is it supposed to handle when the tag changes?"
03:52:02LloreanWhat is it? I mean, I'm using the default skin, and skins shouldn't be able to prevent the music from playing anyway, so it's still a bug whether changing skins resolves it or not..
03:52:36S_a_i_n_tI don't think it's nonsense to want different subline timeouts based on specific events.
03:53:05JdGordongee Llorean, no shit its still a bug.. I wouldnt ask if I didnt have a good reason...
03:53:40 Nick krazykit` is now known as krazykit (
03:53:43LloreanJdGordon: Well, maybe you could tell me what theme I should try or what tag I should try changing in the theme?
03:54:07LloreanI don't have any other themes on the player, so I can't really answer your question without downloading or changing it, and it'd waste both our efforts if I make stupid changes unrelated to the problem...
03:54:09JdGordonI couldnt because I dont know what the issue is..
03:54:22JdGordonit sounds like a beast-cabbie issue
03:54:27S_a_i_n_t"If W is happening, alternate every X seconds, if Y is happening, alternate every Z seconds etc." perhaps it makes more sense to me, or perhaps I've just gone insane...
03:54:31Llorean"do random stuff and see if it gets better" is something I can do, but not something that seems the most efficient.
03:54:34LloreanIf you want, I will though
03:55:43JdGordonS_a_i_n_t: when should the timeouts be checked? lets say last check we wanted to have the line up for 2 sec, but now we want it for 3, do we wait another second? what if we want it for 1? do we instantly change?
03:55:53LloreanJdGordon: iCatcher doesn't work either
03:56:36LloreanJdGordon: Nor rockbox default
03:57:20JdGordonok thats very strange
03:57:40S_a_i_n_tJdGordon: Hmmmm, now you've made me think of that, I can see how awkward it would be to implement.
03:57:52S_a_i_n_tI don't really have an answer for that question sorry...
03:58:13JdGordonhaha I was thinking as I typed that thet they are non issues and have obvious answeers :p
03:58:41LloreanS_a_i_n_t: What are the sorts of situations where you'd want to change the timeout?
03:59:17 Quit sinthetek (Read error: Operation timed out)
04:01:43LloreanI mean, I imagine a lot of times you can accomplish it by careful design. If you need a 3 second timeout and a 2 second timeout for one, and a 2 and a 1 for another, you can just have a lot more alternating lines adding 2s+1s to get 3s (basically, repeat information on some timeouts)
04:01:50LloreanIt may not be impossible to accomplish what you want now, just more complex
04:02:14JdGordonyeah, I dont realy see the point
04:02:47***Saving seen data "./dancer.seen"
04:03:02S_a_i_n_tA conditional timout would be easier (and a lot less code) than having different viewports that are essentially the same except for the timeout
04:05:02JdGordonit might be doable like %t(%?aa<1|2|3>)text;
04:05:12LloreanS_a_i_n_t: I don't know how the timeouts work behind the scenes, but I'd imagine figuring out what to draw when becomes a lot more complex when several timeouts can change at one on a line.
04:05:19S_a_i_n_tI know that conditional timeouts can already be done in that way (multiple viewports), but I was just thinking of an easier way to achieve the same effect in the one viewport.
04:06:11JdGordonbieber: shows how the current renderer works... I'll do do the same listing for the new one soon
04:06:44S_a_i_n_tI never expect "I've asked for X/think X would be usefull, so implement it", I was just putting it out there...
04:08:43JdGordonthe problem with %t(%?aa<1|2|3>) is that there is no way to force those numbers are numbers
04:09:04JdGordonwhich means that will (and does) pass the first parser step but not the second if that 2 was a a
04:09:18 Join sinthetek [0] (~sinthetek@unaffiliated/sinthetek)
04:10:26JdGordonbut if we were to add this I would want it in that form instead of %Nn<%t(1.0)|%t(2.0)>
04:11:14JdGordonI think...
04:12:42JdGordonboth give pretty nasty parse trees to deal with
04:16:16S_a_i_n_tYeah, as I said...I don't *expect* it, it's just one of those things I think that if it *could* be done it could potentially make coding a lot easier. Primarily due to the fact that it would eliminate the need to have multiple viewports that only differ in the timeout field.
04:16:39S_a_i_n_tIf it can't be done, or is too difficult, or will mess things up...I can live without it.
04:21:56JdGordondisabled conditional viewports arnt cleared are they?
04:22:24JdGordonunless you draw over them..?
04:24:05LloreanJdGordon: The task for my bug just got updated. He said "everything after rX" before, but he meant "rX and everything after has the problem" so you may be looking at one revision later than he meant.
04:24:42JdGordonno, I looked at rX
04:24:57LloreanAh, okay.
04:25:48 Join tchan [0] (~tchan@lunar-linux/developer/tchan)
04:27:25 Join funman [0] (~fun@rockbox/developer/funman)
04:35:14S_a_i_n_tJdGordon: Regarding the viewport clearing thing...yes, that is correct.
04:35:26S_a_i_n_tThey are not cleared, they need to be drawn over.
04:38:40JdGordonbieber: you clever bugger :)
04:38:54JdGordonI'm looking at how commantes are handled and the parser gets it spot on :)
04:39:23JdGordonI was worried I'd have to check if the first element in a line is a COMMENT or not
04:40:05JdGordonoh wait.. no
04:40:43JdGordon"abc#comment\nABC" should have abc and ABC on two different lines no?
04:40:59bieberWell, the way the current parser works, a comment is supposed to include its newline
04:41:12bieberThat's why you can do things like sticking a comment at the end of a conditional to extend it to a new line
04:41:23S_a_i_n_tJdGordon: I'm looking at how the skinupdater converted my %Vi line, and I've looked at the wiki, what is the first field for exactly? in "%Vi(-,2,18,172,108,1)%Vf(FFFFFF)%Vb(000000)" am I right in thinking that the "-" in the first field means that this UI Viewport gets used by default?
04:41:36JdGordonS_a_i_n_t: yes
04:41:45LloreanJdGordon: Comments are supposed to eat the line ending, as far as I know.
04:42:07*JdGordon checks svn parser
04:43:45JdGordonyes, thats what svn does
04:43:49JdGordonok, /me stops worrying
04:45:14*JdGordon sees if he can make the new parser not store COMMENT elements at all
04:46:46 Quit amiconn (Disconnected by services)
04:46:47 Join amiconn_ [0] (quassel@rockbox/developer/amiconn)
04:47:09 Nick amiconn_ is now known as amiconn (quassel@rockbox/developer/amiconn)
04:47:10bieberIt would be easy to just skip them altogether, I just want them in there so that generated code will keep comments
04:47:27bieberi.e., I don't want the theme editor to trash all your comments if you just open a WPS and resize a couple viewports ;)
04:47:43 Quit pixelma (Disconnected by services)
04:47:44 Join pixelma_ [0] (quassel@rockbox/staff/pixelma)
04:48:03 Nick pixelma_ is now known as pixelma (quassel@rockbox/staff/pixelma)
04:48:57JdGordonI understand... I've already changed it to not store the comment on ROCKBOX builds, I want to remove the tree element completly... looks simple enough
04:49:13bieberAh, okay
04:49:34bieberJust do me a favor and ping me if you commit anything, I can be bad about forgetting to svn update
04:49:56JdGordonof course I could just as easily skip the element in the renderer... but thats no as nice
04:52:13 Join Barahir_ [0] (
04:52:40JdGordonhehe infinite loop :p
04:55:38 Quit TheSeven (Ping timeout: 248 seconds)
04:55:38 Quit Barahir (Ping timeout: 248 seconds)
04:56:05 Join RoC_MasterMind [0] (
04:59:50 Quit fdinel (Quit: Miranda IM! Smaller, Faster, Easier.
05:00:27 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
05:11:44 Join perrikwp [0] (
05:19:41JdGordonbieber: is the LINE type used for alternator or the logical line?
05:20:01bieberLogical line, but sublines will have LINE elements as their children
05:20:40DhraakellianI have a rockboxed sansa. I want to be able to use it with an iPod dock. If I were to make a female iPod−−>male Sansa adapter cable, would such a thing be possible with Rockbox?
05:21:20S_a_i_n_tDhraakellian: I guess that is one of the things that only trial and error will answer.
05:22:20S_a_i_n_trasher: Ping-a-ding-a-ling?
05:23:13*Dhraakellian ponders
05:28:35S_a_i_n_tscorche: Pingerooney?
05:29:21 Quit chrisb (Ping timeout: 245 seconds)
05:30:44 Quit RoC_MasterMind (Quit: Leaving)
05:32:43vaguerantgevaerts: OK, finally done.
05:32:57vaguerantThat patch made the build go perfectly.
05:33:01vaguerantLooks like somebody broke something.
05:33:10vaguerantTalking about
05:33:29funmanvaguerant: i'll commit the patch then
05:33:31 Quit panni_ (Quit: ( :: NoNameScript 3.81 :: ))
05:34:40 Join panni_ [0] (
05:36:34funmangrr.. i can use gas macro in .S files but not in inline asm
05:40:30 Quit panni_ (Quit: ( :: NoNameScript 3.81 :: ))
05:42:01JdGordonbieber: /me is very confused :p
05:42:12bieberWhat's confusing you?
05:43:02JdGordonitems in a LINE are not put in the children array?
05:43:21bieberNo, each line is stored as a linked list
05:43:26S_a_i_n_tJdGordon: Does %ax (mirror viewport according to language direction) not work for %Vl? If not, bugger...
05:44:06bieberIt's only when you have items that are logical children of each other, like SUBLINES and CONDITIONAL, that each subelement goes into the children array
05:44:36JdGordonarg, ok
05:45:00JdGordonso line->children[0]->next->next->next...?
05:45:05funmanvaguerant: can you use the cross compiler correctly ?
05:45:38bieberIf you wanted, you could have the parser traverse the list after it's been generated and put it in the array instead
05:46:02bieberI just didn't want to have the parser trying to divine how many elements it would need room for ahead of time
05:47:24S_a_i_n_tI wanted to update one of my themes to be *fully* language aware, support RTL alignment as well as translation but if the "%ax" tag doesn't work with "%Vl" then I'm screwed.
05:48:58JdGordon%ax should work with Vl
05:49:21S_a_i_n_tOk cool, the wiki says only %Cl and %V
05:49:31JdGordonall the %V tags are the same
05:49:54S_a_i_n_tI wasn't sure if it meant "%V(and family)"
05:51:06JdGordonbieber: ok, thats a bit confusing but makes sense....
05:51:32vaguerantfunman: Yep, seems to be working perfectly, I just compiled the latest SVN.
05:52:03bieberIt is a little bit of a kludge. In my C++ model, I just copy all the elements into a QList when loading it so I can avoid dealing with it
05:54:11 Quit tchan (Quit: WeeChat 0.3.3-dev)
05:54:49CIA-6New commit by funman (r26754): arm/memset: don't use ldm to load pc when it has been stored with str ...
05:54:56CIA-6New commit by funman (r26755): rockboxdev: build with -U_FORTIFY_SOURCE ...
05:57:06 Quit Rob2222 (Quit: Rob2222)
05:57:46 Join Rob2222 [0] (
06:00:57 Quit anewuser (Quit: for SELL 2 by the price of 1 now!)
06:02:48***Saving seen data "./dancer.seen"
06:07:18 Join CGL [0] (~CGL@
06:08:03JdGordonbieber: and children[0] of a VIEWPORT will only ever be a LINE or SUBLINE right?
06:16:20 Quit kramer3d (Quit: Leaving)
06:25:10 Quit CaptainKwel (Quit: Ex-Chat)
06:27:12JdGordonbieber: (guess you're gone now?) looks like child[0] of a VIEWPORT is a the TAG if it isnt the default viewport? thats fine but why not put that straight into the element like conditionals?
06:31:08*S_a_i_n_t is rather surprised that "mute" isn't in english.lang :/
06:31:45S_a_i_n_talso, along the same line I'm rather surprised that "line level" isn't either.
06:32:14S_a_i_n_tnot *as* surprised by the latter, but still surprised.
06:32:22funmanUnhelpful: opinion on the gas macros in ?
06:33:21funmanuse: ldrpc; ldrpc cond=eq; ldmpc regs=r10; ldmpc regs="r4, r6"; ldmpc cond="ne", regs=r10-r11 ....
06:33:51funmanI could make regs the first argument and get rid of the "regs=" but it would make order unnatural when a condition is used
06:35:02*JdGordon has a "working" skin renderer :)
06:36:19*S_a_i_n_t applauds JdGordon loudly!
06:36:35S_a_i_n_tWell done man, congrats.
06:36:41Unhelpfulfunman: they seem essentially sound and reasonable to me. i take it we're looking to do mixed thumb/arm?
06:36:51JdGordonand by renderer I of course mean something that just walks the tree and spits out the tag names :)
06:37:05JdGordonand always uses the first value for a conditional and subline :p
06:37:26funmanUnhelpful: yep, gives ~100kB binsize gain on AMSv1 and hardly noticeable speed loss in decoders (even some speed increase in some cases)
06:38:01funmani'll commit this, wait for kugel to make the AMS build with eabi, and make those AMSv1 with 2mb of ram use thumb
06:38:18funman(some?) non-eabi gcc hasn't a libgcc built with interworking
06:41:39CIA-6New commit by funman (r26756): FS #11335 by me: make ARM assembly functions thumb-friendly ...
06:44:57funmanhmm, it built fine with eabio
06:47:29funmans/eabi/the recent binutils we build with eabi/
06:47:36 Join shai_ [0] (
06:47:48CIA-6New commit by jdgordon (r26757): Add my new parser playground to svn. ...
06:48:01 Quit rvvs89 (Changing host)
06:48:01 Join rvvs89 [0] (ivo@pdpc/supporter/base/rvvs89)
06:50:23 Quit shai (Ping timeout: 265 seconds)
06:50:53CIA-6New commit by funman (r26758): drop 'req' from gas macro, not available in binutils 2.16 ...
06:57:22funmanhm screw me, mpegplayer isn't built on clipv1
06:58:03Unhelpfulwhat, wanted to see if it could build/run now?
06:58:24funmanjust build, i'm not sure how CPP macros & gas macros fit together
06:58:34funman(see table, disregarding the final " i forgot)
06:59:58funmanthere's no substitution inside ""
07:00:59Unhelpfulthey fit just fine, your source goes through CPP first, gas macros are part of assembly.
07:01:28funmanyep but if there's a ',' inside a gas macro argument I must enclose it in ""
07:01:54funmanI can cheat and change HIGH_REGS=r9-r11 into HIGHEST_REG=r11, so no ',' is needed
07:05:30 Join notlistening [0] (
07:06:29 Nick shai_ is now known as shai (
07:15:20CIA-6New commit by funman (r26759): CPP substitution isn't made inside " ", but we need " " when using , in a gas macro argument ...
07:31:35 Join LinusN [0] (linus@rockbox/developer/LinusN)
07:35:18 Quit joeyg (Ping timeout: 276 seconds)
07:37:40CIA-6New commit by funman (r26760): FS #6734 by me: optionally build with -mthumb on ARM ...
07:39:25Unhelpfuldoes gcc have any option for heuristically deciding whether functions should be arm or thumb, or can you only tell it an entire C file should be one or the other?
07:40:49funmanentire C file (-mthumb or default -marm)
07:40:50bieberJdGordon: That makes sense, I'll take care of that when I get back in a little bit
07:41:44JdGordonok, I've added a new folder under utils which has my "renderer"
07:41:45funmanyou need -mthumb-interwork on every file though
07:42:08 Join joeyg [0] (
07:42:20funmanbtw -flto (in gcc 4.5) could reduce size even further by removing unused exported functions/veneers
07:44:04Unhelpfuli don't think there should ever *be* unused veneers, if you mean the ones inserted for long calls.
07:44:16Unhelpfulthose are already done at link time
07:44:22funmannope for thumb/arm switches
07:44:47funmanIIUC those are added when building with -mthumb-interwork
07:46:00Unhelpfulah, i see... shame those aren't already generated at link time as with the long-call veneers.
07:46:34funmanhm well, but then they aren't added in .S files so they must come from linker ... but then where's the binsize increaase added by -mthumb-interwork ? :/
07:47:02funmangrep from_arm only matches .elf files so it comes from linker
07:47:46funmanwell thumb-interwork doesn't generate the veneers, only 'code which supports calling between ARM and Thumb'
07:47:59Unhelpfulok, then, they're probably already pretty optimal.
07:48:28funmanTODO: investigate the gas bug which happens on PP, and why -mlong-calls isn't enough to workaround it
07:50:16funmanalso, test on more targets ;)
08:01:16 Join Buschel [0] (
08:02:51***Saving seen data "./dancer.seen"
08:04:21funmanhm FM recording seems to work on c200v2 but not microphone
08:10:04funmanit's broken on clipv1 too
08:14:56funmanand fuzev1, damn!
08:23:40 Join wpsprob [0] (
08:25:35wpsprobwhat does Sx; Sr; mean ?. I have searched here
08:27:52JdGordonSx() is translate a string I tihnk
08:28:33JdGordon%Sr is used to find out if the language is left-to-right or right-to-left
08:28:41JdGordonboth need to be added apparently
08:28:47S_a_i_n_tyep, %Sx(string_to_be_translated)
08:31:30funmanI can't say what's wrong in mono2stereo() but the C version works, the ASM version just powers the device off ..
08:31:43funmanit's reported to work, so the problem might only appear when using eabi
08:32:01funmanah i see it
08:32:28funmangcc gives the same register to 2 variables :/
08:33:37 Quit wombat23_ (Ping timeout: 264 seconds)
08:35:55 Quit Buschel (Ping timeout: 248 seconds)
08:38:05 Join intrados [0] (
08:39:36CIA-6New commit by funman (r26761): as3525v1: drop the asm version of mono2stereo ...
08:39:51wpsprobI want to display play count as 1/100 would this be correct, %pe %Sx | %| |
08:40:40JdGordonany ideas why "buffer_front = (void *)(((unsigned long)buffer_front + 3) & ~3);" works in the sim but not a stand alone program? (64bit os)
08:40:50funmanJdGordon: define 'work'
08:40:51JdGordonbuffer_front is a unsigned char[]
08:40:54 Join flydutch [0] (
08:41:10funmanif you do pointer alignement you should use uintptr_t not unsigned long
08:42:56funmaneh. recording works fine on c200v2. Now the only feature missing on AMS/2MB (that I know) is voice (the .voice file doesn't fit in the audiobuffer when doing nothing, let alone while playing)
08:43:00S_a_i_n_twpsprob: For a start, %pe isn't even playcount, it's playlist entries.
08:43:18JdGordonI cant define work.. all I know is when I use that I get messed up pointers.
08:43:27JdGordonanyway, /me is off
08:43:40wpsprobthats ok, I'm just rying to get the syntax right for play count as 1/100 would this be correct, %pe %Sx | %/ | %| %pp|%pp %Sx | %/ | | %pe
08:44:25S_a_i_n_twhay are you throwing %Sx in there?
08:44:38S_a_i_n_tyou're not adding any translatable string.
08:45:04S_a_i_n_tyou want "%rp%/100
08:45:16S_a_i_n_tsorry... "%rp%/100"
08:45:26S_a_i_n_twithout the " "s
08:46:00S_a_i_n_tthough, why do you want the playcount listed as "out of 100"? It can quite easily exceed this number.
08:46:13 Join ender` [0] (
08:46:32S_a_i_n_t123(for example)/100 will just look weird.
08:46:49S_a_i_n_twpsprob: ^
08:47:03wpsprobwhat i'm rrying to display is "x" / "y"
08:47:38 Join Zagor [0] (bjst@rockbox/developer/Zagor)
08:47:42wpsprobwithout the ""
08:48:14S_a_i_n_tdo you want playcount, or "number in playlist"?
08:48:42S_a_i_n_tI can quite easily show you what you need, but I don;t understand what you want.
08:49:26 Join Buschel [0] (
08:49:46S_a_i_n_tthat's "current position/of X entries in playlist"
08:50:04S_a_i_n_tis *that* what you want?
08:50:11wpsprobit's how to use the / as the divider that I'm stuck on
08:50:27wpsprobir seperator rather
08:50:57S_a_i_n_toh, mistake "%pp/%pe"
08:51:19wpsprobcurrently it's X of Y, I want to change that to X / Y
08:51:19S_a_i_n_tit's simple...if you want a "/", just add one.
08:52:02S_a_i_n_tso, you want "total playlist entries/current number of playlist"?
08:52:33wpsprobthank you
08:52:37S_a_i_n_tno problem
08:53:06 Join wombat23 [0] (
08:56:29 Quit funman (Quit: free(random());)
08:57:58 Quit intrados (Quit: WeeChat 0.3.2)
09:02:21 Join petur [0] (~petur@rockbox/developer/petur)
09:09:27 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
09:20:12 Quit detaos (Ping timeout: 245 seconds)
09:35:21 Join efyx [0] (
09:39:10 Join DerPapst [0] (
09:41:10 Join kugel [0] (~kugel@rockbox/developer/kugel)
09:53:37 Quit wpsprob (Quit: CGI:IRC (EOF))
09:53:40 Quit FlynDice (Ping timeout: 265 seconds)
09:54:00 Join pondlife [0] (~Steve@rockbox/developer/pondlife)
09:55:07 Join Rob2223 [0] (
09:58:32 Quit Rob2222 (Ping timeout: 245 seconds)
10:00:18 Join FlynDice [0] (
10:02:17 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
10:02:52***Saving seen data "./dancer.seen"
10:03:23 Quit flydutch (Ping timeout: 248 seconds)
10:03:33CIA-6New commit by bieber (r26762): Theme Editor: Modified parser to integrate the %V tag into the VIEWPORT element, in the same style as CONDITIONAL
10:03:42 Join flydutch [0] (
10:04:28 Join swilde [0] (
10:08:28kugelhm, eabi as configure option, thumb as advanced build. we should maybe make it consistent
10:13:50 Join detaos [0] (
10:16:23kugelI'd say target/arch specific things should be a configure option, and general stuff advanced build options
10:19:55 Join dfkt [0] (dfkt@unaffiliated/dfkt)
10:22:22 Quit kugel (Remote host closed the connection)
10:29:24 Quit S_a_i_n_t ()
10:29:46 Join S_a_i_n_t [0] (S_a_i_n_t@
10:34:51 Join B4gder [0] (~daniel@rockbox/developer/bagder)
10:37:41 Join Kitr88 [0] (
10:41:47 Quit Kitar|st (Ping timeout: 258 seconds)
10:42:22 Quit Kitr88 (Ping timeout: 264 seconds)
10:47:16 Join Kitar|st [0] (
10:57:30 Quit TheSeven (Ping timeout: 252 seconds)
10:59:36 Join pyro_maniac [0] (
11:00:49 Join esperegu [0] (~quassel@
11:00:50 Quit simabeis (Remote host closed the connection)
11:05:02simonrvnlatest build won't boot. target sansa fuze2, eabi toolchain. stalls at booting message.
11:11:18jhMikeSfunman: I think pcm-pp.c has a bug. (take a good, close look at the change for thumb) :)
11:30:32 Join wodz [0] (
11:37:42wodzI have philosophycal question - where to put setup of GPIO responsible for wall charger detect - into button_init() or into power_init() ?
11:45:55 Join stooo [0] (
11:53:17wodzpixelma: ping
11:57:44Stummiwhich version of qmake do i need to build the theme editor?
11:59:10 Quit stooo (Quit: Leaving.)
11:59:19 Join liar [0] (
12:00:33 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
12:02:55***Saving seen data "./dancer.seen"
12:05:17 Join watto [0] (~watto@
12:14:19 Quit pamaury (Ping timeout: 265 seconds)
12:27:26 Join esperegu_ [0] (~quassel@
12:27:38 Quit esperegu (Ping timeout: 264 seconds)
12:28:24 Join kaylinsigswort [0] (
12:33:35 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
12:37:53wodzTL350A lcd controller used in HD200 (and iaudio M3) has a command which puts it in power saving mode (which reduces current consumption by about 1mA). I was thinking if there is a hook in rb currently to enter this mode when hold switch is on ?
12:38:15 Quit S_a_i_n_t ()
12:38:25 Join S_a_i_n_t [0] (S_a_i_n_t@
12:40:58 Join funman [0] (~fun@rockbox/developer/funman)
12:41:00 Quit mt (Ping timeout: 258 seconds)
12:41:27 Join bmbl [0] (~Miranda@unaffiliated/bmbl)
12:42:13funmankugel: i'm ok if you want to make thumb a configure option
12:43:14funmanjhMikeS: thanks for the hint, s/, #-4// should do the trick
12:44:04 Quit feisar_ (Ping timeout: 276 seconds)
12:45:56wodzfunman: I was thinking about indicate which grid scale is used in view_battery(). What about put it in the same line as scale string and make it scroll if it do not fit on screen? (small screen width)
12:46:30jhMikeSfunman: just caught it while looking about trying to figure this S thing, which only affects pcm playback, not recording dma or disk dma. (somethings trashing something during init now)
12:46:50funmanjhMikeS: bissecting didn't show a particular commit?
12:49:38jhMikeSfunman: it did, r26718, which makes no sense at all unless it's revealing another underlying problem
12:49:45 Join mt [0] (~mtee@rockbox/developer/mt)
12:50:34funmanwhat if you revert only the commit where you changed locking?
12:52:57funmanwodz: why not if scrolling works
12:53:45 Join kugel [0] (kugel@rockbox/developer/kugel)
12:53:47jhMikeSwas considering that (weighing my options)...but if it is due to movement of areas, and that works, then it might still not be revealing the real cause...
12:53:58funmansimonrvn: eabi fuzev2 build works fine here
12:54:14kugelfunman: any specific reason you added USE_THUMB to autoconf.h and not simply to EXTRA_DEFINES (with -DUSE_THUMB)?
12:54:25 Part watto
12:54:35funmani wasn't 100% sure if EXTRA_DEFINES was applied to all assembly files
12:54:52LloreanjhMikeS: 26718? I thought that the post on the FS task mentioned he got a different number bisecting.
12:55:26kugelfunman: I we just call gcc for all files (for the final link too) so it should be ok
12:56:12funmani'll test if moving it works
12:56:15jhMikeSLlorean: oh, there's a FS? :) after a certain revision, the problem could show up periodically then. bisecting got me there, but I never when revision by revision to see.
12:56:37kugelfunman: I think gcc defines something when compiling for thumb on it's own but I don't know if that helps us
12:56:50LloreanjhMikeS: FS #11390
12:57:27funmankugel: __thumb__, but that doesn't help because it's only defined when using -mthumb. -mthumb is not used for some files but they need to be aware that the whole build will use thumb
12:58:22wodzwhat does HAVE_LCD_SLEEP, HAVE_LCD_ENABLE and HAVE_LCD_SHUTDOWN mean ?
13:00:03funmankugel: can you check if playback works with on PP ?
13:00:30jhMikeSwodz: has to do with shutting down the LCD driver IC, putting in various saving states
13:00:44kugelfunman: I'm not at home, but later sure (if nobody beats me)
13:00:57funmankugel: ah, we need to check a specific variable anyway (to modify $CC)
13:02:21funmanbut well, we can set config_thumb to something and modify extradefines
13:02:57jhMikeSLlorean: one thing's for sure, mpegplayer or any other pcm user is affected the same way
13:02:57funmanwodz: dunno about LCD_SHUTDOWN but I have read somewhere that LCD_SLEEP is like LCD_ENABLE, but it sleeps stronger
13:03:14funmanIMO this is nuts, but perhaps there is some explanation why we don't always use the strongest sleep
13:03:48jhMikeSfunman: yes, it "sleeps stronger". we don't use it that way because some lcd's take a long time to bring back out of a deeper mode.
13:04:31 Join feisar [0] (
13:04:59 Nick feisar is now known as Guest43341 (
13:05:07wodzso if I have command that enters powersave mode of lcd ic I should define what to utilize this?
13:05:32jhMikeSI first introduced that, shutting it all down, but some didn't like the extra delay. so it became: just shut off the visible display, if a timeout expires, go into standy mode
13:05:34funmanif coming back from powersave is nearly instant, just lcd_enable i think
13:07:00jhMikeSon gigbeat f and s, it's fast enough to come back that it simply waits 5 or so secs before going into sleep mode. they have no option, they stay fully active for a few second (to allow fades to complete first).
13:07:26kisakdo the arm builds on the site use eabi?
13:07:54jhMikeSkisak: soon maybe, not now though by default
13:08:08kisakok, thanks
13:08:35*S_a_i_n_t whistles for someone with dev-access to the themesite to help him out...
13:09:03S_a_i_n_trasher: scorche: ping-a-ding-a-ling?
13:10:47jhMikeSwodz: I'd look at x5 or e200 if your display is like the usual renesas stuff, which seems nearly ubiquitous.
13:12:00 Join robin0800 [0] (
13:13:04CIA-6New commit by uchida (r26763): text viewer: can select the indent spaces when the line mode is reflow.
13:13:29funmansomeone can try playback with on PP ?
13:13:53wodzmy question was a bit more general. if I define LCD_ENABLE I have to provide lcd_enable() function which is called when?
13:14:16 Quit soap (Ping timeout: 265 seconds)
13:14:20wodzwhat this function should do? enable/disable lcd content instantly?
13:15:19funmanwodz: it's called by backlight functions
13:15:29wodzI know
13:15:39wodzbut what should be the desired effect
13:15:44 Join soap [0] (~soap@rockbox/staff/soap)
13:16:44funman'enable' LCD / put it in powersave mode
13:16:57jhMikeSwodz: lcd_enable blocks drawing operations and visible operations (like powering off the panel). lcd_sleep puts it in its deepest state.
13:17:00funmanif the backlight is off you won't be able to read the screen anyway?
13:17:21wodzcool that's what I wanted
13:17:37 Quit perrikwp (Quit: CGI:IRC (EOF))
13:18:26wodzfunman: I can read screen with backlight turned off just fine - its greyscale
13:19:03CIA-6New commit by funman (r26764): make thumb build a configure option (−−thumb) ...
13:20:08pixelmawodz: is that the greyscale lcd of the MPIO - you want to shut it off?
13:20:26jhMikeSwodz: it's the usual to leave it on, if it's perfectly readable. (not that kind of readable when you need direct sunlight to barely make out the contents :) )
13:20:52wodzpixelma: yes
13:21:06wodzIdealy I would like to turn it off on hold()
13:21:30wodzIf I use hold switch player goes to my pocket so I don't need display
13:21:56pixelmaah yes... I wouldn't want to turn that off as it is usually perfectly readable without backlight (except in complete darkness) and none of the other monochrome or greyscale targets disable the lcd...
13:22:22Torneyou can do backlight/lcd independantly, no?
13:22:23Tornelike ipodvideo?
13:22:31Tornedefaults to only disabling the backlight, but you can also disable the lcd if yo uwant
13:22:50wodzTorne: yes exactly
13:22:54funmani can't find the vectors in e200 rockbox.elf :/
13:23:09Torneso it's still worth implementing disabling the lcd, if there's any power saving at all from it
13:24:10wodzlowlevel stuff is easy - I don't get how high level framework is organized
13:24:28 Quit soap (Ping timeout: 248 seconds)
13:24:40Tornewell, that's why the rest of us are here :)
13:25:18CIA-6New commit by funman (r26765): pcm-pp: revert r26756 ...
13:25:47 Join soap [0] (~soap@rockbox/staff/soap)
13:25:50funman(my PP patch can't work as is)
13:26:29wodzso coming to the point - if I implement lcd_powersave(on_off) type function and I want it to be called on hold() what should I define :-)
13:27:11funmani would say lcd_sleep()
13:27:44funmankugel: i want your eabi patch !!!
13:27:53funman(in svn)
13:27:55kugelwodz: lcd_enable() and it's automatically called when backlight goes of
13:28:06kugelfunman: commit it :)
13:28:29funmani thought you were polishing it or something
13:28:56funmani can add explicit thumb for the AMSv1 lowmem at the same time if you can't commit from where you are
13:29:03kugelno need to I think. it's temporary until all ARM run eabi anyway
13:29:19 Join ucchan [0] (
13:29:21wodzI want it to turn off lcd on hold() not on backlight timeout
13:29:38kugelwodz: backlight should go off when the hold switch is enganged
13:30:08wodzkugel: as well as when backlight time expires
13:30:36kugeland that's a problem?
13:31:16wodzfor me yes
13:31:45funmankugel: ah, don't you think it would be better to make a blacklist and not a whitelist, to show which models haven't been tested with eabi ?
13:32:06funmani.e. make eabi the default but will all but AMS blacklisted
13:32:21pixelmakugel, wodz: there is a seperate setting for backlight on hold - off, on and same as usual backlight setting
13:32:56kugelwodz: you can set backlight = on in the settings but backlight on hold = off in the settings, there's no need to hardcode it
13:34:00 Join simabeis [0] (
13:34:18 Quit esperegu_ (Read error: Connection reset by peer)
13:35:02wodzThis is still not what I am thinking about. I want to have backlight timeout and lcd on whet hold button is off and lcd off when hold button is on
13:35:34Tornehm, you mean never turn off the lcd, only the backlight, except when hold is on?
13:35:41Torneyeah we don'thave that combination of settings
13:35:56 Join esperegu [0] (~quassel@
13:35:58Tornebut in any case, don't try and implement this kind of logic in the firmware level
13:36:19pixelmaI can't follow
13:36:45Zagorpixelma: in normal use, only backlight goes off. lcd stays on. when hold is engaged, also lcd goes off.
13:36:59Zagorright wodz?
13:37:20gevaertsthat does seem to make some sense
13:37:35TorneYeah, maybe we should make that work
13:37:41pixelmaaha, still I don't get why anyone would want to disable the lcd on greyscale or monochrome targets
13:37:43Tornebut in any case, the actual firmware code is the same
13:37:51Tornepixelma: if it uses less power, why not?
13:38:04Tornei disable the lcd on ipodvideo, which is perfectly readable in reasonable lighting with no backlight
13:40:12pixelmait depends on how much it really saves in real use - you are then forced to do something to enable the lcd (and backlight) if you want to look at the dissplay. If you don't switch the lcd off you can just look at it if the light situation allows without the backlight - which is likely on greyscale or mono targets
13:40:39Torneyes, but we already have all the infrastructure to do this
13:40:43Torneit's just a matter of implementing one functoin
13:40:50Torneso i'm not sure why you wouldn't do it :)
13:40:57wodzpixelma: why do use HOLD?
13:41:16Tornei almost never look at the display of my player at all, whether hold is on or not
13:41:26CIA-6New commit by wodz (r26766): HD200 - small cleanup in button_init_device()
13:41:30CIA-6New commit by wodz (r26767): HD200 - power_init() rework based on latest power consumption measurements
13:41:31Tornei don't take it out of my pocket to operate, mostly :)
13:41:34pixelmawodz: to prevent accidental button presses
13:41:35CIA-6New commit by wodz (r26768): HD200 - clean up usb related functions
13:43:47CIA-6New commit by funman (r26769): Make EABI gcc the default for ARM platforms ...
13:45:00pixelmaTorne: Ipod Video is a bit different because the lcd probably draws more power than a 128x128 greyscale display (or even smaller resolution) and even though the lcd is quite readable without backlight it is still a bit worse than greyscale or monochrome LCDs. Although if you ask me, I wouldn't have implemented that setting there as well
13:45:38Torneipodvideo it saves a lot of power, because it means we can power down the BCM video chip
13:45:41Tornewhich sucks up loads
13:45:52Torneit's not the LCD itself that's the saving there
13:46:11 Quit Buschel (Ping timeout: 258 seconds)
13:46:41Torneit extends battery life by a couple of *hours*
13:46:43pixelmaok, that adds to that
13:47:05Tornein fact that change was what pushed us past the OF's battery life on that device, i think ;)
13:47:15kugelmaybe just implement a lcd_sleep on hold setting
13:47:43pixelmaI thought it was half an hour on Buschel's Video when he measured but ok
13:48:15Tornekugel: well, lcd shutdown has immediate/never/varioustimes
13:48:15funmanah, clients don't benefit from ccache :)
13:48:32Tornewe could just add another option there for "when hold switch on"
13:48:56wodzthat would be perfect for me
13:49:06pixelmayay, more settings ;\
13:49:20gevaertsfunman: actually, part of the slowness could well be that not that many clients have the eabi toolchain yet
13:49:29 Quit kugel (Quit: exit(0);)
13:49:43*S_a_i_n_t wonders if the Nano 1/2G can disable the LCD and it's just not an available setting, or if it hasn't been implemented on those targets yet.
13:50:21CIA-6New commit by funman (r26770): Make −−thumb the default for AMSv1 with 2MB of RAM ...
13:50:37TorneS_a_i_n_t: if it was implemented the setting would show up..
13:51:26pixelmawodz: I'd be really curious about some measurements in a somewhat "real use" setting
13:52:11Tornepixelma: theoretical and real use are quite close for this, no?
13:52:16Tornemost people don't interact with their player a lot
13:52:23S_a_i_n_tnot could be ifdeff'ed out for some odd reason ;)
13:52:31 Join watto [0] (~watto@
13:52:37Torneso if the lcd is off all the time they're not using it, then that's pretty close to getting the maximum saving
13:52:52CIA-6New commit by funman (r26771): ata-creative-zvm: make sure return value is defined
13:52:54pixelmadepends on how often you look at the display for just reading the track info on your display
13:53:35Tornea little, but if you kill it on hold, that's, what, ten seconds every five minutes?
13:54:27pixelmawell, if you interact with the player to only switch on the display, backlight comes on too
13:54:40 Quit B4gder (Quit: It is time to say moo)
13:54:47pixelmaand backlight is probably the bigger drainer
13:55:36Tornethat doens't negate the fact that 90%+ of the time both would be off, if you configure it that way
13:55:40pixelmabut ok, the test might still be interesting in the usual setting
13:55:59Torneand thus the difference between just backlight off vs both off has a lot of time to add up
13:58:20pixelmaI have a monochrome player with poorer batter life (the Ondio - about 7-8 hours on their usual NiMHs, a bit more on alkaline). Still I never wished to disable the LCD
13:58:35pixelmabattery too
13:59:02liarS_a_i_n_t: it's possible on the nano2g, but its not implemented in rockbox
13:59:56wodzpixelma: Ondio - that remind me to ask You for testing FS #11189 :-)
14:00:06Tornepixelma: yah, but since we have the facility there already, it seems reasonable to support it on as many players as possible if there's any measurable improvement at all..
14:00:06liari have a patch here but that needs to be cleaned up
14:00:12Zagorfunman: you forgot to change the builds list. now all clients are told to do eabi builds, even when they don't have the compiler
14:01:44funmanwhat's the builds list ?
14:01:50Zagoror they would, if the configure change had worked :)
14:02:14pixelmaTorne: the latter is what I am curious about...
14:02:24Torneright, it's worth measuring..
14:02:53funmanZagor: can you handle that?
14:02:57***Saving seen data "./dancer.seen"
14:02:59Zagorfunman: sure
14:03:12pixelmawodz: ah yes. I'll do that
14:03:16funmanon my side i'll ask configure to not build the sims & checkwps in thumb ;)
14:03:47Zagorfunman: it seems the default eabi is not working. most (all?) builds are still using the old compilers.
14:04:07funmanZagor: well yeah I blacklisted all stable targets not tested with eabi
14:04:15funmanso far only Sansa AMS use eabi
14:04:26funmanand unstable targets
14:06:13Zagoris a blacklist the best way to do it? we're spreading the compiler information in a lot of places already
14:06:41funmanI think it's the best way to remind developers they should test eabi builds!
14:07:44Zagorumm, how exactly are they reminded? the patch just silently uses the old compiler.
14:08:06funmani just hope they would see it the commit :/
14:08:33funmanI can make a warning appear on stdout, it won't show as yellow on the table
14:13:29 Join bluebrother [0] (~dom@rockbox/developer/bluebrother)
14:14:51LloreanWe've got until 3.7 to work out the kinks, why not just move everything over, see what breaks, and recommend that people who must have a working build use 3.6 for the time being?
14:15:27funmanThat's what I wanted to do too, but I think the DevConEuro crowd decided we'd need to test individually each target
14:16:16CIA-6New commit by funman (r26772): configure: don't use thumb on native builds (simulator/checkwps)
14:16:21CIA-6New commit by funman (r26773): Warn when building an ARM target with non-eabi default compiler
14:16:44 Quit bluebroth3r (Ping timeout: 248 seconds)
14:18:45 Quit vaguerant (Quit: Java user signed off)
14:20:21Zagorfunman: did we? I'm leaning more towards "change it and see who screams" too
14:20:41funmanthat's what I remember
14:22:08funmanMaking eabi default everywhere is very fine with me
14:23:05Zagoron the other hand, build rounds will take >10 minutes if we move all targets over now
14:24:03funmanI wonder about the fuzev1 red delta
14:24:12funmanand e200v2
14:24:22Zagorwe have 19 eabi clients right now
14:29:00Zagorfunman: so which models are eabi now? can you write me a list?
14:33:00 Join panni_ [0] (
14:35:36TheSevenbtw, is there anything holding back switching nano2g to eabi?
14:35:51ucchanfuman: thumbs version's iPod video Rockbox starts from ipodloader2 !!
14:36:00TheSeveni'm running eabi builds on mine for a while now, and all the bugs seem to have been sorted out by now
14:36:01funmanwow, nice!
14:36:05funmanucchan: does it boot ?
14:36:15funmanTheSeven: just remove it from the tools/configure list
14:36:23ucchanfunman: boot ok!! Rockbox logo displays
14:36:46Torneis there any reason to suspect that the thumb code will be better on targets like ipodvideo?
14:36:57 Join domonoky [0] (~Domonoky@rockbox/developer/domonoky)
14:37:00TheSevenucchan: does it actually work, or does it get stuck on the logo?
14:37:21 Quit robin0800 (Remote host closed the connection)
14:37:29amiconnfunman: The problem you fixed in r26761 is indeed an (obvious) bug in the constraint
14:37:37literaldo Sansa AMS devices have ARM processors?
14:38:03funmanTorne: what's specific to the ipodvideo ?
14:38:09funmanamiconn: ah, which one?
14:38:18amiconnSo if you want the one cycle per sample back, revert it, then add an '&' (early clobber) modifier to the 'left' constraint
14:38:24Tornefunman: nothing..
14:38:31Tornefunman: which is why i suspect thumb would be worse
14:38:34Tornethumb is, generally, worse.
14:38:44ucchanTheSeven: confirm logo displays normally.
14:38:49funmanTorne: well, one must try
14:38:53wodzamiconn: have You see my test of Your greylib patch on HD200?
14:39:12amiconnwodz: Erm, no?
14:39:14Tornethumb is originally intended for systems where 32-bit memory accesses are more expensive than 16-bit
14:39:43amiconnTorne: PP DRAM interface is 16 bit as far as we know
14:39:50funmanZagor: gogearhdd1630 gogearsa9200 samsungyh820 samsungyh920 samsungyh925 sansac200v2 sansaclip sansaclipplus sansaclipv2 sansae200v2 sansafuze sansafuzev2 sansam200v4 zenvision zenvisionm30 zenvisionm60
14:39:58wodzamiconn: I posted results on
14:40:18wodzamiconn: It is faster with Your change
14:40:21amiconnThumb makes more efficient use of the cache, so it may speed up things even though it needs a few more instructions
14:40:32funmanTorne: some codecs are actually faster when building what we can in thumb
14:40:45Tornefunman: well that's resonable if the dram interface is 16 bit..
14:41:00Torneand PP cache is kinda crappy
14:41:32funmanafaict dram interface is 32 bits on the models where i tested
14:41:46Torneso ok, pp probably does benefit i guess
14:41:51funmananyway now it's easy for someone to make some numbers
14:42:02amiconnwodz: Okay. Btw, your results look like the HD200 skips greylib interrupts when running unboosted
14:42:11funmanucchan: can you play some files with thumb build ?
14:42:22amiconnSo NEED_BOOST might in fact be appropriate for HD200 too
14:42:41Tornebut later ARMs generally don't; symbian builds stuff for thumb solely because it saves ram, and most things are generally a little slower that way despite the reduced icache pressure
14:43:06TheSevenhmm, seems like someone broke my pcm driver
14:43:08funmanAMS is 'later than PP'
14:43:26funmanTheSeven: check today logs : jhMikeS has a problem on gigabeats with PCM
14:43:37CIA-6New commit by zagor (r26774): Moved AMS sansas to eabi gcc.
14:43:56TheSevenfunman: nope, i just spotted some change some weeks ago in my code
14:44:03jhMikeSbetter yet, check FS #11390
14:44:10 Quit wodz (Quit: Leaving)
14:44:20TheSeveni'm just wondering what's the right way to fix it
14:44:33amiconnwodz: Further speedup might be possible by utilizing line transfers, since the greylib buffer is in dram
14:44:55amiconnDid you also check whether the greylib works properly?
14:45:08jhMikeSTheSeven: fix the PCM? it certainly seems like something is trashing PCM on the beast, making only playback inoperable for both plugins and core
14:45:35amiconnPossible failure modes would be inverse (negative) display, or pixel stripes swapped
14:46:26TheSevenjhMikeS: I'm talking about r26253
14:46:52 Quit esperegu (Read error: Connection reset by peer)
14:46:58amiconnEven when running at twice the rate, almost 50% cpu load when boosted is a lot. The H1x0 has one third of that, despite its larger lcd
14:47:35*amiconn thinks that 150Hz scan rate gives rather smooth looking greyscales
14:47:53 Join n1s [0] (~n1s@rockbox/developer/n1s)
14:47:55 Join chrisb [0] (
14:48:12S_a_i_n_tHmmmm, howcome the themesite reports the current build as being "r26761M"? (the part I'm querying is the `M`)
14:48:29TheSevenhas anybody experienced crashes/lockups when stopping playback on nano2g with builds since r26253
14:48:48TheSeveni have a strong suspicion that this will almost certainly happen, from looking at the change
14:48:55ucchanfunman: play ogg/mp3/wave ok. and text viewr plugin ok.
14:48:59*Torne has experienced crashes/lockups when stopping playback while the system is heavily loaded (loading dircache/tagcache/etc), on ipodvideo, *always* :)
14:49:04Torneif that helps
14:49:17Torneit doesn't do it frequently or repeatably enough to give me much of a clue
14:49:18S_a_i_n_tTheSeven: Not myself personally...the only thing I've noticed that seems to be behaving badly on the Nano2G is pictureflow while playback is running.
14:49:51 Join esperegu [0] (~quassel@
14:50:05S_a_i_n_tAnd by "behaving badly", I guess what I really mean is "crap" :/
14:50:30TheSevenit will almost certainly jump to a garbage address while stopping DMA, because someone removed a register initialization but not the jump using the register
14:50:39 Quit scorche (Read error: Connection reset by peer)
14:51:10 Join scorche [0] (~scorche@rockbox/administrator/scorche)
14:51:20funmanucchan: hmm I wonder if building e200v1 with thumb still fails then
14:51:41jhMikeSTheSeven: which pcm file? that's pp?
14:51:55S_a_i_n_tThe 2G is a pretty fast target, there's no reason I can think of that should have the Nano1G out-perform the 2G with pictureflow. Playback and pictureflow used to be fine, but the pictureflow-wps integration seems to have messed it up somewhat.
14:52:08S_a_i_n_tIt's all jerky and slow, and distorts the AA.
14:52:53S_a_i_n_tIt's fine if playback is stopped.
14:52:57n1sjhMikeS: do you know of any reason not to switch the beast to using ebai by default?
14:53:33funmanjust do it !!11!1!
14:54:28funmanbtw, kugel remembers the same thing than me (we decided to switch targets one by one after experimentation), but I can't see anything on DevConEuro2010 page
14:54:33jhMikeSTheSeven: it should just use the callback in pcm.c without condition (I suppose you mean the r0)
14:54:36 Join Szpila [0] (~sszpila@
14:54:49TheSevenit isn't
14:55:09TheSevenit's using dma_callback if *not* stopping, else a leftover value from some other code
14:55:26 Join mt_ [0] (~mtee@
14:55:47jhMikeSTheSeven: I know it isn't, and well, that's clearly not my intent, but rather a mistake, to leave r0 possible uninitialized.
14:55:47TheSevenso you say the ldrne should be a ldr?
14:56:00TheSeventhe other option would be bxne r0
14:56:26ucchanfunman: I don't undefstand. e200 specific code (driver?) are broken still.
14:56:28jhMikeSTheSeven: if stopping, it should just call pcm.c. pcm.c will call all the routines that are required to stop playback.
14:56:30mc2739S_a_i_n_t: See FS #11387 regarding pictureflow on nano2g
14:56:58TheSevenit should be a ldr instead of ldrne then
14:57:06 Quit mt (Ping timeout: 258 seconds)
14:57:06TheSeveni still don't get how that could ever work though.
14:57:07domonokyS_a_i_n_t: the themesite reports the modified version, because i forgot to revert some changes where the automated checkwps are built. Should be fixed on the next checkwps update (tomorrow)
14:57:11jhMikeSTheSeven: I'm guess that's what should have been done, yes
14:57:26jhMikeSn1s: no, I've tried it and it's ok. only thing is the problem that's there now.
14:58:11TheSevenjhMikeS: can you commit a fix? i don't have a checkout with me right now.
14:58:47funmanucchan: hmm if i build ipodvideo with −−eabi −−thumb, I still see the buggy __memset_from_thumb call in corelock_init
14:58:57TheSevenactually you could simplify it even further by just doing a bl dma_callback instead of all the ldr, mov, bx stuff
14:59:02jhMikeSTheSeven: sure. I'll take your word for it being the right thing. :) All I could do was guess before.
14:59:09funmanucchan: to verify: you ran tools/configure with −−thumb −−eabi ?
14:59:39n1sjhMikeS: what is the "problem that's there now"?
15:00:23 Join hebz0rl [0] (
15:00:26 Join einhirn [0] (
15:00:34TheSevenjhMikeS: I'll verify if everything works fine in the evening
15:00:55jhMikeSn1s: pcm playback is messed up on the beast. I don't know precisely where it started where it became unstable. See FS #11390.
15:00:56 Quit mt_ (Ping timeout: 258 seconds)
15:01:02jhMikeSTheSeven: ok, np
15:01:20*TheSeven is still confused by that code not breaking instantly
15:01:27ucchanfunman: 1) ../tools/configure −−eabi 2) make only (gcc is used arm-elf-eabi-gcc 4.4.4/arm-elf-eabi-ld
15:01:54funmanucchan: thumb is off by default, you need −−thumb. It is only activated by default for clipv1, c200v2, and m200v4
15:02:29funmanucchan: you can try with −−eabi −−thumb, and edit Makefile to add -mlong-calls, but kugel said it didn't work on e200v1
15:03:23n1sah, that doesn't look compiler related but i think it's a good idea to hold off the change untill that's fixed :)
15:03:39S_a_i_n_tTheSeven: Can you answer a Nano2G query of mine?
15:04:02TheSevendepends on the query ;-)
15:04:11ucchanfunman: oh! Because binsize is small, then I thought that it was thumbs version.
15:04:41 Join kugel [0] (~kugel@rockbox/developer/kugel)
15:04:43funmanucchan: eabi also helps with binsize because it removes some long calls
15:05:03kugelZagor: "funman: did we? I'm leaning more towards "change it and see who screams" too" yes, you did actually :)
15:05:10S_a_i_n_tI was wondering why "Charging: True/False" shows up in "Debug/View HW Info" and not in the (as I expected) "Debug/Battery"?
15:05:24S_a_i_n_tTheSeven: ^
15:05:26kugelI suggested to change all in one go, but you said we should make the change "more controlled and verify all targets"
15:05:26 Quit pjm0616 (Ping timeout: 265 seconds)
15:05:31funmankugel: ah thanks, i'm not mad ;P (or we both are)
15:05:47TheSevenS_a_i_n_t: historical reasons
15:06:10funmanucchan: I have a small test case for the GNU ld bug with PP + thumb, I'll try to speak with binutils developers
15:06:15ucchanfunman: ok. I try build "−−eabi −−thumb".
15:06:15TheSevenand maybe because we read it from GPIOs and not the PMU
15:06:21S_a_i_n_tIs that the other way of saying "Noone has changed it yet"? ;-P
15:06:48kugeloh, gcc 4.4 apparently produces smaller code for armv5+ than 4.0
15:06:52 Part LinusN
15:07:32 Join roolku [0] (
15:08:50kugelfunman: shouldn't iaudio 7 / cowond2 also use eabi? they aren't stable
15:09:00 Join pjm0616 [0] (~user@
15:10:09funmanyeah, I probably messed up my list
15:11:02CIA-6New commit by jethead71 (r26775): s5l8700 PCM: Always call dma_callback. (fix r26253 remnants)
15:11:17roolkuzagor: any idea why n02 - n18 are disabled? They only have the old (pre devcon) eabi install, but they could have build the other platforms?
15:11:41 Join vaguerant [0] (
15:11:47 Quit vaguerant (Changing host)
15:11:47 Join vaguerant [0] (~vaguerant@wikipedia/vague-rant)
15:12:28roolkuZagor: I can bring them down now and move them over to the new installation, unless you want me to leave them as ?
15:12:44jhMikeSwhat happened? I see the expected and logged build time shot up through the roof (eabi?)
15:12:58roolkuZagor: Server message: Fatal build error: Command not found. You have been temporarily disabled.
15:13:34roolkuZagor: that is after: Starting build archosondiospboot
15:13:57funmanjhMikeS: not all clients have eabi compiler
15:14:04roolkujhMikeS: my buildclients don't like the change for some reason
15:14:09funmanalso thumb builds are 50% longer (but OTOH there are only 3 of them)
15:14:25funman6*, but bootloaders are quite small
15:14:35funmanwhich reminds me that i didn't test thumb bootloaders
15:14:47*roolku goes to upgrade the clients
15:15:00TheSevendoes thumb make any sense at all for bootloaders?
15:15:19funmannot really
15:15:20kugelI think the culprit is that way more targets use the eabi now than just the ams ones (and Zagor only changed the ams ones in the builds list
15:15:24jhMikeSn1s: I have a patch to use the INIT section on the beast, but I need eabi for it. I suppose that ticked me off enough that now we're here with eabi. :)
15:15:53n1swhy does that need eabi?
15:16:03kugelso roolku's clients try to do eabi builds because the build master thinks they are old-gcc ones
15:16:42roolkukugel: but for some reason they don't build any of the other stuff either
15:16:48jhMikeSn1s: because beast uses short calls, and the init secion needs to be in a place where short call cannot reach
15:17:03kugelroolku: yea, ones they fail a build they're disabled completely
15:17:06 Quit esperegu (Read error: Connection reset by peer)
15:17:22kugelroolku: command not found makes absolutely sense, you have no arm-elf-eabi-gcc on the system :)
15:17:46n1sjhMikeS: oh
15:17:50jhMikeSn1s: most notably, the plugin buffer, we have 64MB (= way too distant)
15:18:05 Quit flydutch (Quit: /* empty */)
15:18:20n1swhere does the init section go?
15:20:10 Join evilnick_B [0] (0c140464@rockbox/staff/evilnick)
15:20:39kugeljhMikeS: you cannot use the plugin buffer anyway
15:20:56kugeljhMikeS: you need the codec buffer (skin engine grabs the plugin buffer during startup)
15:21:07kugelbut the codec buffer is probably too far away anyway :)
15:21:25 Part Szpila
15:22:53ucchansorry all developpers, does not boot iPod video.
15:23:02jhMikeSkugel: hmmm...well, then the codec buffer. when does it grab the plugin buffer? it *could* have something to do with the pcm problem since SDMA code is put there, but it's done with it *very* early in init
15:23:39kugeljhMikeS: theme loading at the end of the startup
15:23:49kugelit loads the .wps/.sbs files into it
15:24:28 Quit pyro_maniac (Remote host closed the connection)
15:24:42 Join esperegu [0] (~quassel@
15:29:58*jhMikeS thinks it's impossible it'll matter since sdma is done with the plugin buffer before kernel_init even returns
15:33:42chrisbdoes iPod's line out from a dock work?
15:33:45kugeljhMikeS: you could use init_attr for that sdma as well
15:33:49chrisbclassic ipod
15:34:11kugelit should be compatible with data too
15:34:45Tornechrisb: should do, if you turn it on
15:34:53Tornechrisb: there's a setting to enable/disable it because it takes some power
15:35:05Tornenote that it is really a line out, i.e. the volume is fixed
15:35:06jhMikeSkugel: in the patch, I did exactly that
15:36:24funmanI just sent a bug report to bug-binutils ML
15:36:39kugelfunman: brave!
15:36:50jhMikeSkugell: it didn't matter at the time about the plugin buffer, but that's an easy alteration
15:40:23chrisbTorne: thanks
15:40:50 Quit TheSeven (Ping timeout: 260 seconds)
15:41:49Zagorroolku: builds broke temporarily because tools/configure was modified without changinging the buildserver build table. so all clients were asked to build eabi arm targets.
15:42:13ZagorI think the ondio line was unrelated
15:42:45Zagorthat's also why the build time skyrocketed: clients are disabled on failed builds, so only a few clients remained in the round
15:43:19roolkuZagor: yes, seems to be okay now. :) hopefully all the n?? nodes will be up in a short while.
15:48:05 Quit lev_ (Ping timeout: 258 seconds)
15:48:39 Join saratoga [0] (9803c6dd@gateway/web/freenode/ip.
15:48:44jhMikeSfunman: if I just remove the model from models_not_checked_with_eabi, it uses eabi as the default?
15:49:11saratogaam I correct in thinking that no longer requires me to change the path if I install eabi now?
15:50:22*jhMikeS still has to build 4.4.4 (still has 4.4.3)
15:50:25funmanjhMikeS: yep
15:50:53funmansaratoga: if you have /usr/local/bin in PATH, nope, but you should remove the path to old compilers from PATH
15:55:46kugelZagor: can you check it doesn't update anymore
15:56:25 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
15:57:38roolkubuildclients n01-n18 now upgraded...does anyone have anything to commit? :)
15:57:39 Join GeekShado_ [0] (
15:57:47Zagorroolku: yay
15:57:58*jhMikeS was going to do something soon, but not in five minutes
15:58:10Zagorroolku: better now?
15:58:11kugelis the builds file fixed?
15:59:00funmani'll try something
15:59:10roolkui'll hang around a little longer
15:59:52kugelZagor: locally? I see no commit
16:00:11roolkuI'll leave the other clients without eabi, shouldn't make much of a difference
16:00:12Zagorkugel: www commits are not listed on the front page
16:00:28kugelI know, I was looking here:
16:00:37mc2739kugel: r26774
16:00:44 Quit GeekShadow (Ping timeout: 248 seconds)
16:00:46 Quit pjm0616 (Ping timeout: 276 seconds)
16:01:09kugelmc2739: that one wasn't sufficient
16:01:20Zagorkugel: why not?
16:01:44kugelthere are more targets than the ams sansa that use eabi
16:02:29Zagorfunman said he blacklisted all non-ams targets
16:02:42funmanall _stable_ non ams
16:03:00***Saving seen data "./dancer.seen"
16:03:00funmanthe yh and zvm are built with eabi too
16:03:15 Join pjm0616 [0] (~user@
16:03:16funmanand gogear too
16:03:31funmani have a commit ready
16:03:35kugelexcept for iaudio7 and cowond2, those use old abi still (but ought to use eabi as well)
16:04:11Zagorwell, I did ask for a list...
16:04:20 Quit ucchan (Quit: Leaving...)
16:04:32CIA-6New commit by funman (r26776): pcm as3525v1: correct ASM constraint
16:04:36funman.. which i gave
16:04:50CIA-6New commit by funman (r26777): pcm as3525v1: revert 26761 to use faster ASM code ...
16:05:04 Join GeekShad__ [0] (
16:05:12Zagorfunman: time?
16:05:30jhMikeSpcm_postinit should *not* be forced into the init section, it completes asynchronously
16:07:08funmanjhMikeS: woops, i suppose i am to be blamed ?
16:07:33Zagorfunman: excellent, thank you
16:07:36jhMikeSno idea who? I just saw that in the mapfile and realized that cannot work reliably
16:08:02funmanjhMikeS: that must be me, but i remember checking if the path that lead to this func was fully in init functions
16:08:41funmanaudio_thread() calls it, which itself is created in audio_init(), which itself is an init func
16:08:46 Join DataGhost [0] (~dataghost@unaffiliated/dataghost)
16:09:10funmanbut then the thread execution might be postponed?
16:09:10 Quit GeekShado_ (Ping timeout: 258 seconds)
16:09:37kugelif audio_thread calls audio_init it shouldn't be INIT_ATTR
16:09:50funmannah that's the reverse
16:09:58funmanaudio_init() creates the thread
16:10:03kugeloh sorry
16:10:22funmanwtf is this fnarfbargle-uk-BradC client without a tools/configure ?
16:12:44 Quit pjm0616 (Ping timeout: 276 seconds)
16:15:21CIA-6New commit by zagor (r26778): Fixed missing eabi builds.
16:15:51 Join kugel_ [0] (
16:17:00 Join pjm0616 [0] (~user@
16:20:37 Quit kugel (Ping timeout: 272 seconds)
16:23:12 Quit vaguerant (Quit: Java user signed off)
16:24:30 Quit antil33t (Read error: Connection reset by peer)
16:24:37 Join antil33t [0] (
16:27:40 Join kugel__ [0] (
16:27:52 Nick kugel__ is now known as kugel (
16:27:58 Quit kugel (Changing host)
16:27:58 Join kugel [0] (~kugel@rockbox/developer/kugel)
16:29:55 Quit funman (Quit: free(random());)
16:30:13 Join TheSeven [0] (~TheSeven@rockbox/developer/TheSeven)
16:31:28 Quit kugel_ (Ping timeout: 240 seconds)
16:33:18jhMikeSafter changing the compiler, gigs seem have magic pcm functionality again for apparent good reason...hmmm.
16:37:34 Nick GeekShad__ is now known as GeekShadow (
16:38:37 Quit kugel (Quit: exit(0);)
16:39:38CIA-6New commit by jethead71 (r26779): Configure Gigabeat S with EABI compiler by default. Implement the INIT section that this enables (due to selective need for long calls). Remove ...
16:40:19 Join crk [0] (
16:41:15ZagorjhMikeS: please update www/buildserver/builds too
16:43:09 Join komputes [0] (~komputes@ubuntu/member/komputes)
16:43:43jhMikeSZagor: excuse me?
16:44:57Zagorwhen you change compiler for a target, you need to update the buildserver too otherwise it will send eabi builds to non-eabi clients
16:45:08TheSevenhm, fnarfnbargle keeps killing builds?
16:45:17 Join toffe82 [0] (~chatzilla@
16:45:42TheSevena single client on rampage can apparently hurt the build system quite a bit
16:45:54Bagderwe can block single clients
16:46:24*TheSeven wonders if the build system should do that itself with this kind of errors
16:46:34jhMikeSZagor: no idea where to find that setting :)
16:46:48ZagorjhMikeS: www/buildserver/builds
16:47:06 Quit toffe82 (Remote host closed the connection)
16:47:16TheSevenespecially if a single client completed like five builds in ten seconds, it shouldn't be assigned anything for the rest of that round at least
16:47:19 Quit crk (Quit: CGI:IRC (EOF))
16:47:55ZagorTheSeven: clients are temporarily disabled if they fail to upload binaries. unfortunately fnarfbargle was assigned mostly non-upload targets.
16:48:03*TheSeven is about to bring his build client up again
16:48:07Zagormany clients can complete >1 build/second
16:48:40TheSevenZagor but having 5 bad builds in a row at that rate means that something is wrong with either the code or the client
16:49:09ZagorTheSeven: certainly. but if the code is bad, all clients will be disabled...
16:49:22 Join loveless [0] (~loveless@
16:49:30TheSevenwhich won't hurt that much if they are only disabled for the rest of that round and the code is bad anyway
16:49:36 Join dfkt [0] (dfkt@unaffiliated/dfkt)
16:50:01TheSevenwhat exactly do i need to do to update my pre-eabi client? remove all the old compilers from my path and /usr/local/whereever they were installed and run again? what else?
16:50:05Zagoryes it will, because the non-broken targets will not be built
16:50:21ZagorTheSeven: you only need to remove any old compilers you plan to replace
16:50:26knittlZagor: got a little time? ^^
16:50:28Zagoryou can build just the eabi one if you like
16:50:39Zagorknittl: I found the git bug: permission issue
16:50:42 Join bucko [0] (~bucko@
16:50:46TheSevenwhat's the current full list of architecture names for the buildclient script?
16:51:17ZagorTheSeven: arm arm-eabi-gcc44 sh m68k mipsel sdl
16:51:26buckoHey doodz! I just installed Rockbox onto my Sansa V2. Nice stuff there! But I'd like to add a new shuffle mode which plays albums from start to finish. How would I do this?
16:51:29knittlZagor: yep, works now :)
16:51:38Zagorknittl: yay :)
16:51:45knittlcooool :)
16:52:02AlexPbucko: I think random folder advance will do this if each album is in a folder. If using the database, no idea
16:52:20buckoI need the database as I have a micro SD card, too.
16:53:09buckoAlso can I turn off USB transfer support? When I plug in USB it reverts to old firmware when normally I just want to charge it. I understand there is no USB file transfer support but I don't want that mostly.
16:53:16AlexPWhy does that matter?
16:53:20lovelessAnyone around for the GPN conference in Karlsruhe this weekend? Would love to meet up.
16:53:26AlexPYou can get to the card through the filebrowser too
16:53:39buckoAlexP: It appears as a subfolder so in terms of folder shuffle it'd miss it, right?
16:53:46*jhMikeS honestly still has no idea exacly where this setting is since it never concerned him before\
16:53:55AlexPbucko: Don't think so, check the manual
16:54:02buckoOh, it's over all folders in device? :(
16:54:15AlexPbucko: Also, check the manual - it tells you what button to hold when inserting USB to charge
16:54:52AlexPbucko: It builds a list of folders, and then you can remove ones you don't want. It'll then play through those folders randomly
16:55:07 Quit petur (Quit: *burp*)
16:55:15buckoWhere would I look in manual?
16:55:35AlexPI'd get the pdf one and search
16:55:42AlexPrandom_folder_advance is a plugin
16:55:52AlexPFor the charging there is a section on it early on
16:57:37 Join toffe82 [0] (~chatzilla@
16:57:46 Join ssorgatem [0] (~ssorgatem@
16:57:46buckoAh right, found that. Was using onling manual and wasn't obvious it would be in 'quick start'. :)
16:57:53AlexPah yeah
16:58:07AlexPThe pdf one is much nicer in that it allows searching :)
16:58:11bucko(And the FAQ on batteries is really broken.)
16:58:30AlexPThe wiki needs general cleanup/fixing/removing etc.
16:58:42AlexPIt is a user driven resource, except nobody does anything
16:58:57jhMikeSZagor: well, I would update it, but I can find nothing in my checkout by those names
16:59:03buckoHeh. :)
16:59:08buckoWhere do I find plugins?
16:59:15AlexPjhMikeS: You need to checkout www as well as trunk
16:59:22AlexPbucko: They are all installed
16:59:34buckoOh right, it's an 'application'.
17:00:07amiconnZagor: It's arm-eabi-gcc444 according to your mail (and what worked when I set it up)
17:00:27jhMikeSAlexP: heh, well, never did :), never seemed to matter to what I was doing either
17:00:40Zagoramiconn: oops, yes. arm-eabi-gcc444 is correct.
17:00:55AlexPjhMikeS: First time for everything :)
17:01:30 Part Zagor
17:01:32buckoAlexP: It made a playlist of the results?
17:01:51jhMikeSAlexP: exactly. I ignore what I don't care about until I have to. :)
17:01:57amiconnZagor: Btw, the error message you get when using a wrong (i.e. nonexisting) target string in the client is kinda irritating
17:02:05amiconnMaybe you can do something about this?
17:02:08buckoI suppose that's a reasonable approach.
17:02:17AlexPbucko: No, it is a list of all the folders (sub folders included) on the device - I imagine it is stored in .rockbox somewhere
17:02:37buckoAlexP: It makes a playlist too once you tell it to 'play shuffled', I think.
17:02:49AlexPbucko: The plugin lets you edit the list (well, you remove things)
17:03:04TheSevenS_a_i_n_t: hm, the keyclick patch has rotten again?
17:03:06AlexPbucko: Could be, I only every used it once years ago :)
17:03:15buckoWell, it does what I want so. :)
17:03:22AlexPGood to hear :)
17:03:57buckoHow easy is theme making? Current theme displays lots of nice info but long track names and so on get cut off and I'd like to make them wrap or something.
17:04:16AlexPThey won't wrap, but you could make them scroll
17:04:25AlexPAlthough most themes already do this
17:04:28AlexPThe default does
17:04:35buckoOh, they already scroll. No way to make them wrap?
17:05:03AlexPnot currently anyway :)
17:05:05buckoCan't even tell one line to display say 20 chars and the next to display the next 20 or something?
17:05:06 Join esperegu_ [0] (~quassel@
17:05:18AlexPNot of a tag
17:05:47buckoBut it can play Doom.
17:05:56AlexPYou never know though - mention it is JdGordon's presence and if he is bored he might put it together
17:05:59S_a_i_n_tTheSeven: Has it? Oh...bugger.
17:06:08 Quit hebz0rl (Quit: Ex-Chat)
17:06:13TheSevenat least it just conflicted during an svn up
17:06:26jhMikeSjust replace "arm" with "arm-eabi-gcc444" at the beginning?
17:06:31TheSevenseems to be rather easy to sort out though
17:06:31buckoI think this means my MP3 player is a better computer than my first PC.
17:06:41 Quit esperegu (Ping timeout: 276 seconds)
17:07:05S_a_i_n_tI would *really* like someone to look at it and figure out why it doesn't work for anything except the Nano2G.
17:07:26buckoAnyway, now I have to see if I can install it on my parents' iPods.
17:07:28S_a_i_n_tI'd *love* HW Key-click on my Nano1G also.
17:08:02TheSevenactually i would like to see this committed
17:08:27saratogai have 'arm-elf-eabi-gcc-4.4.4' but the email says the build script takes ' arm-eabi-gcc444', is that correct?
17:08:47ssorgatemHas the migration of the arm tootlchain to eabi as default been completed for sansa AMS?
17:09:12S_a_i_n_tI've looked at why it doesn't work for the other iPods anymore...but it's totally beyond me.
17:09:52CIA-6New commit by jethead71 (r26780): Update build server selection for Gigabeat S to select for eabi.
17:11:20S_a_i_n_tTheSeven: I know you're a busy man...but, if you can figure it out I'll build a shrine out of iPods and proclaim you to be my new God.
17:11:21TheSevenis "subs pc, lr, #4" (as an FIQ return) thumb-safe?
17:11:22jhMikeSwell, hope I got it looked right :)
17:18:06 Join Tarvis [0] (~Tarvis@
17:18:21Tarvishey guys\
17:18:33TarvisI noticed my H300/fuze/whatever theme BoXamp is missing from the site
17:18:40TarvisIs it just because of the format change
17:18:41TorneTheSeven: yes, because subs restores CPSR from SPSR_irq
17:18:47Torne..i think
17:18:54S_a_i_n_tTarvis: Yes, most likely.
17:19:09TarvisI'm guessing the converter utility didn't work with it
17:19:14TarvisNo matter - I'll just update it myself
17:19:40TheSevenTorne: oh right, we don't need to watch out for bit 0 there, as the cpsr will be thrown away anyway
17:19:58TorneTheSeven: yah, writing to PC with S flag does it :)
17:20:09TheSevenhowever, I'm not so sure about the subtraction of 4
17:20:35TheSevenor is pc always pointing 4 bytes ahead, even if that's two instructions?
17:21:47TarvisIt is there, just not in the listing because it doesn't work with curent build
17:22:26TorneTheSeven: PC doesn't "actually" point anywhere consistently, but it's okay to assume that the address to return to is lr-4
17:22:37TorneTheSeven: because it's architecturally defined as being so :)
17:22:59Tornein normal code PC points to two instructions ahead
17:23:11TheSevenjust wanted to make sure that my pcm assembler code is thumb safe :-)
17:23:14Tornein exceptions, lr ends up being whatever ARM say it is ;)
17:23:24Tornewhich is always a fixed offset
17:23:31Torneyou don't have to do anything to interwork exceptions :)
17:23:50*rasher waves at bucko
17:24:35 Quit antil33t ()
17:27:31buckoOh it's Bacon Boy.
17:27:52buckoOr another rasher.
17:28:21buckoSo it mentions in the manual that I can improve battery life by using an efficient codec.
17:28:37buckoIs there any details on what codecs are indeed the most efficient?
17:28:59 Join stoffel [0] (
17:29:29saratogabucko: CodecPerformanceComparison on the wiki
17:29:46TheSevenbucko: this also depends heavily on whether you use a flash- or hdd-based device
17:30:31TheSeveni suspect for flash it will be flac
17:30:54buckoI was mostly wondering between MP3 and Vorbis.
17:31:26saratogaprobably vorbis on nearly all devices
17:31:30TheSevenprobably vorbis then, but i have also seen vorbis files which were worse than mp3 for some unknown reason
17:31:49saratogaTheSeven: do you have such a file handy?
17:31:59buckoVorbis on that page seems to consistently need more megahurts.
17:32:09TheSevenno, i only vaguely remember seeing that some months ago
17:32:20saratogabucko: are you sure you're reading it correctly?
17:33:28TheSevenbtw, how do mp3 and vorbis compare in terms of audio quality vs. bitrate?
17:33:38buckosaratoga: I don't know? The MHz column in the top one is "MHz needed for realtime", which seems like a sensible measure of processing requirements. And on the first few at least, Vorbis needs more MHz.
17:33:45saratogavorbis is more efficient below 128k, not much difference above it
17:34:01saratogabucko: which player are you using?
17:34:19buckoSansa V2. I'm not sure details exist for that on said page.
17:34:52buckoVorbis seems to win on MIPS though.
17:35:31buckoOn the V1, MP3 still seems to win.
17:35:37saratogabucko: D2 is probably closest to your CPU, in which case its 28MHz for MP3 and 22 for Vorbis
17:36:07TheSevenhm, my build client is still compiling compilers...
17:36:13 Quit pamaury (Remote host closed the connection)
17:36:16saratogathat would be "Telechips TCC7801" which is arm9e like the fuzev2
17:37:04buckoSo they appear about the same on that one. Is there a way for me to run these benchmarks myself?
17:37:19saratogabucko: on the V1 vorbis is faster too
17:37:26saratogaalthough its a different CPU so not important to you
17:38:22buckoe200 v1 MP3 128 20.71, Vorbis 128 25.10MHz. MP3 seems faster there.
17:39:01buckoMarginally faster on the Fuze though.
17:39:17saratogabucko: ah e200v1, but no, vorbis is much faster
17:39:27saratogathe mp3 decoder on the e200v1 uses both cores
17:39:33saratogaso its really about 38-39MHz
17:39:50buckoOh. That's something I'd hope to see indicated on said page. :)
17:40:11saratogayou're welcome to add it, but its not really important
17:40:34buckoI don't really feel authoritattive enough to do so.
17:40:59bucko(And I'd want to do everything if I did so for fairnesses sake.)
17:41:11saratogaif all you care about it runtime, then knowing the mp3 is really fast on PP is all you need to know
17:41:42saratogaknowing that its fast because i split decoding across 2 CPUs only matters if you're working on the decoder, which you'll probably realize anyway while working on the decoder :)
17:41:48saratogathe e200v1 CPU is a PP chip
17:42:16buckoOh but I'm on V2. which you said was like the D2.
17:42:22 Part S_a_i_n_t
17:43:21buckoHow come FLAC's so fast?
17:43:41saratogabecause it has to do a small fraction of the work of most lossy formats
17:44:56buckoMuch more I/O, though.
17:45:22TheSevenI/O only really hurts on HDD-based targets, that's why I asked ;-)
17:45:36buckoOh, Flash I/O is basically free?
17:45:55TheSevennot really free, but not much compared to the actual decoding either
17:46:14TheSevenhard drives use way more power while spinning than flash chips do
17:47:38TheSevenbtw, what shall we do about iLoader vs. the rockbox bootloader now?
17:48:07buckoAlso Musepack and WMA appear to beat Vorbis. I assume that Vorbis is more convenient on Linux than WMA though.
17:48:17TheSeveni would actually like to keep iloader restricted to advanced users and get the rockbox bootloader up to speed somehow
17:48:53saratogawere you thinking about merging iloader into our svn?
17:49:05 Quit roolku ()
17:49:09TheSevennope, I don't think that could work out
17:49:51TheSevenit's a completely different codebase
17:50:01TheSevenit only shares the FTL and NAND drivers to some degree
17:50:50TheSeventhe biggest problem is probably integrating nandfsck into the rockbox bootloader
17:51:18 Quit esperegu_ (Read error: Connection reset by peer)
17:51:44 Quit sinthetek (Quit: leaving)
17:51:54 Join sinthetek [0] (~sinthetek@unaffiliated/sinthetek)
17:52:19 Join esperegu [0] (~quassel@
17:57:10TheSevenand we will of course need to think about a better installation/uninstallation method
17:58:22 Join pamaury [0] (
17:58:22 Quit pamaury (Changing host)
17:58:22 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
17:59:24 Quit loveless (Ping timeout: 245 seconds)
18:00:11 Quit amiconn (Remote host closed the connection)
18:00:12 Quit pixelma (Remote host closed the connection)
18:03:02***Saving seen data "./dancer.seen"
18:03:33 Join pixelma [0] (quassel@rockbox/staff/pixelma)
18:03:38 Join amiconn [0] (quassel@rockbox/developer/amiconn)
18:04:06 Part pondlife
18:06:19TheSevenhuh, is the most recent still using /usr/local/arm-elf-eabi/?
18:06:44TheSeveni'm seeing loads of such paths while running it
18:07:28TheSevenand i did an svn up immediately before running it
18:12:32TheSeven"bl <label>" isn't thumb-safe, is it?
18:16:24TheSevenis it, or isn't it?
18:16:49Torneboth, amusingly
18:16:57Tornein general no, a branch instruction isn't sufficient to set the T bit
18:17:04Tornebut the EABI linker will insert shims
18:17:37Torneassuming it has sufficient information to do so
18:17:38TheSevenok, so i can use that, i'm 90% sure that the target will be arm anyway
18:20:48CIA-6New commit by theseven (r26781): Remove an unneccessary long call from the S5L8700 PCM FIQ handler
18:20:57 Quit GeekShadow (Quit: The cake is a lie !)
18:21:40 Quit Battousai (Ping timeout: 240 seconds)
18:23:11 Join flydutch [0] (
18:24:23 Quit bmbl (Read error: Connection reset by peer)
18:30:31TheSeven[18:06] <TheSeven> huh, is the most recent still using /usr/local/arm-elf-eabi/?
18:30:32TheSevenany ideas?
18:32:13gevaertsTheSeven: you should probably move the old directories out of the way before building
18:32:14bluebrotherTheSeven: no, it was changed
18:32:19TheSeveni did so
18:32:29TheSevenit installed to both locations from what it looks like
18:32:43Torneit installs to /usr/local
18:32:51Tornethis includes some files in a directory called arm-elf-eabi
18:32:59Torneif you look in an old toolchain you will note that it also contains that directory :)
18:33:04 Quit swilde (Quit: ERC Version 5.3 (IRC client for Emacs))
18:33:08gevaertsThose directories are still used, yes. They will have gcc and things, as opposed to arm-elf-eabi-gcc in /usr/local/bin
18:33:25Tornegcc install creates directories named after the target, which contain all the files for that target which aren't prefixed :)
18:33:46TheSevenso it installs two copies of itself?
18:33:54TheSeventhey aren't even sym- or hard links
18:34:09TorneThey're copies of the stub, yes
18:34:28Tornethe actual binaries are in neither of those places
18:34:39TheSevenso those 250KB things are stubs?
18:35:26Tornegcc is, yes
18:35:41Tornebinutils isn't, so i guess there are two copies of that, yeah
18:36:01TheSevenhm, that's a bit confusing
18:36:19Torneanyway, this is just how gcc toolchains look.
18:36:42Tornethis is why you have to delete the old /usr/local/foo first
18:36:42Tornebecause otherwise there wouill be interesting effects ;)
18:38:05 Join S_a_i_n_t [0] (S_a_i_n_t@
18:39:03 Join pamaury_ [0] (
18:39:40 Join Battousai [0] (~bryan@gentoo/developer/battousai)
18:40:51 Join funman [0] (~fun@rockbox/developer/funman)
18:42:31funmanTheSeven: BL doesn't switch, you need the armv5+ BLX
18:42:49TheSevendoesn't BLX *always* switch, no matter what the address is?
18:42:50 Quit pamaury (Ping timeout: 252 seconds)
18:43:01 Quit Battousai (Remote host closed the connection)
18:43:05Tornebx and blx always set T based on the bottom bit of the address
18:43:05funmanit sets the T bit to (reg & 1)
18:43:06 Nick pamaury_ is now known as pamaury (
18:43:13simonrvnfunman: ok, thanks (re: eabi fuzev2 build works fine here)
18:43:22 Quit pamaury (Changing host)
18:43:22 Join pamaury [0] (~quassel@rockbox/developer/pamaury)
18:43:27funmanthe binutils people answered: the bug is already fixed in their svn, they pointed me to the patch so perhaps I can backport it
18:43:47 Join wodz [0] (
18:43:51funmannext 2.21 release is scheduled for October
18:44:13 Join Battousai [0] (~bryan@gentoo/developer/battousai)
18:46:09 Join domonoky1 [0] (
18:46:27bertrikpamaury, I'm doubting a little whether to put the clip+ mystery radio in the existing si4700 driver (some registers/bits are nearly identical), or create a separate driver for it
18:46:48 Quit domonoky (Ping timeout: 276 seconds)
18:47:24funmanbertrik: does it really matter at this point? I'd go the simple way and post a patch which only modifies the code to work with this specific radio chip, clean code can come later
18:47:56pamaurybertrik: if there share code, use the same driver and if it becomes too messy, split it
18:48:06simonrvni'll just go fetch the daily build
18:48:12 Join GeekShadow [0] (~Antoine@reactos/tester/GeekShadow)
18:48:25 Quit elinenbe (Read error: Connection reset by peer)
18:48:43 Join elinenbe [0] (
18:49:47 Join lhn555 [0] (
18:50:02lhn555how do i refresh song list?
18:50:53 Part esperegu (" - Chat comfortably. Anywhere.")
18:52:05lhn555on ipod video
18:52:30saratogawhich list?
18:52:45wodzhow to use lcd_puts_scroll? It prints string but do not scroll at all
18:52:53wodzshould I set up something?
18:53:24 Join kugel [0] (~kugel@rockbox/developer/kugel)
18:53:29bertrikI think it starts scrolling when the tick tasks are active
18:53:58lhn555Database music list
18:54:17kugelfunman: I'd rather backport it, but everyone would need to rerun again
18:55:01funman+ if they want to make thumb builds for PP
18:55:05lhn555saratoga: Database music list
18:55:07saratogait should update automatically provided you have the auto update enabled (which I believe is the default)
18:55:18lhn555huh? it isnt
18:55:34 Part knittl
18:55:44saratogaactually, why do we even have an option to disable autoupdate?
18:55:50saratogawhat is the use case for that
18:55:52lhn555i didnt
18:56:17wodzdo I understand correctly that recent do not provide facility to install toolchains in arbitrary location?
18:56:39 Join Jerom [0] (~Jerom@
18:56:41kugelwodz: no, you don't
18:56:48wodzah ok
18:57:13saratogaif you run it with no options it lists the options
18:57:16saratogaone is where you want to install it
18:57:23kugelwodz: it should give you some hits if you run it
18:57:24saratogainstall_prefix or something like that
18:57:44n1ssaratoga: since we have no way to disable the database, always updating would be annoying
18:57:59saratogan1s: just don't enable it in the first place?
18:58:13kugelfunman: I don't have good experience with non x.yy.1 releases of binutils :\
18:58:45lhn555saratoga: what did you say?
18:58:53wodzso if I want to use lcd_puts_scroll in debug_menu.c do I need to do something special?
18:58:53n1ssaratoga: also, it scans for files on every startup, wich is useless on most startups unless you update your music very often
18:59:14saratogalhn555: what is your question?
18:59:29saratogan1s: doesn't it just read the FAT to see if anything is new?
18:59:42 Join blackromeo [0] (
18:59:55lhn555how do i refresh song list?
19:00:06saratogano i mean the question you asked me
19:00:10bertrikfunman, yeah probably best to worry about integrating later
19:00:16n1ssaratoga: don't think so, but i'm not 100% sure
19:00:30blackromeosup peeps
19:00:33n1slhn555: use "update now" in the database menu
19:00:34lhn555saratoga: i didnt
19:00:44blackromeothanks that helped the other day
19:00:45lhn555n1s; what?
19:00:51blackromeoto all i mean
19:01:13Jeromdid the default toolchain for arm change ?
19:01:18n1slhn555: to update the database, use the "update now" function in the database menu
19:01:24blackromeobut now i have a sloight problem on my fuze
19:01:29lhn555n1s: what do u mean?
19:01:33saratogaJerom: yeah it changed
19:01:39saratogafor some targets anyway
19:01:47n1slhn555: i mean that to do what you want, do what i said
19:01:58n1sor to get the result you want, rather
19:02:02lhn555n1s: what database menu?
19:02:11saratogamaybe you should skim the manual
19:02:43lhn555dont have time
19:02:48n1sgeneral settings->database
19:02:55lhn555oh thx
19:03:24blackromeowhen i turn my fuze on it displays the message "free 90mb for musicDB"
19:03:25 Quit lhn555 (Quit: CGI:IRC (EOF))
19:03:33blackromeohow do i do this?
19:03:40funmanquick answer: format it
19:04:18blackromeook coolies
19:04:48blackromeoi have a mac
19:05:34blackromeowhich i got on monday
19:05:43n1sgood fro you but off topic in here
19:05:57blackromeohow do i format lol
19:06:05pixelmasaratoga: database is not enabled by default and it still has show stopping bugs for me (e.g. it fails if there is no music file it can see at all). Auto-update is not enabled by default because of that and the fact that it is only quick enough with dircache enabled which isn't by default either (also because of show-stopping bugs IMO, e.g. it chokes on too many small files - and it is not available on low mem targets)
19:06:24blackromeoim still getting used to it
19:06:34saratogathis probably isn't the right place to ask how to use a mac
19:07:03saratogapixelma: auto update isn't enabled by default?
19:07:39pixelmanot that I know of, and I just gave some reasons
19:07:39JeromCan someone tell me in which files I can find the Ipod Nano2G usb stack please ?
19:07:54saratogaah you're right
19:08:01saratogawell thats probably that guys problem then
19:09:39bluebrotherJerom: firmware/usbstack maybe?
19:09:58funmanuh oh, crash on clipv1 thumb build
19:10:21*funman turns backlight always on
19:12:21ssorgatemfunman: mkamsboot has the toolchain compiler hard-coded in the makefile... is it safe to use the eabi toolchain for mkamsboot?
19:12:24amiconnwodz: ping
19:12:58pixelmablackromeo: I think you can't find solutions for your problem in sandisk's own forum (can't remember the URL or something, I don't need it). It's a message of the original firmware though and so it's off-topic here and you probably won't find people who would know around here anyway ;)
19:12:59funmanssorgatem: no idea, probably it would be a good idea to build mkamsboot with the 2 toolchains and compare md5sum of the files it outputs
19:13:26funmanssorgatem: well i think we only use it to assemble files, so the dualboot.* output should not change
19:13:41simonrvnfunman: something must be up with something (the bootloader perhaps), because the daily build worked fine for me.
19:13:52wodzamiconn: pong
19:14:01 Join stripwax [0] (
19:14:02CIA-6New commit by bluebrother (r26782): Convert uninstallation to use signals / slots for logging.
19:14:23simonrvnoh well. could be my toolchain too.
19:14:24ssorgatemfunman: then the makefile should be fixed? in order to be able to compile all stuff having only the eabi toolchain installer
19:14:47saratogait should be fixed after its verified to work
19:15:11JeromI'm looking for a more target specific code
19:15:16pixelmablackromeo: I meant to say that you *can* find a solution over at sandisk's
19:15:17amiconnwodz: thru 14:47
19:15:19funmanssorgatem: you don't need a cross compiler to build mkamsboot, just use the dualboot.c/h ; if you want to modify dualboot.S you should know what you are doing
19:16:15wodzamiconn: cube looks ok if that's what You refer to
19:16:20ssorgatemfunman: Mmm true. My fault, I was using a script which already rebuilt dualboot.c/h
19:16:47 Quit stoffel (Remote host closed the connection)
19:17:15wodzamiconn: btw cube runs smooth unboosted also
19:17:39amiconnYes, but then it won't draw 150 fps
19:17:46funmanwodz: now you are in line with the other coldfire for power consumption ?
19:17:51Jeromfunman: your wrote "You can also look at the (working) code for Ipod Nano2G which looks similar.", where can I find the target specific code of the nano2G ?
19:18:01amiconnSince CPU load is already almost 50% at 124MHz, it will be >100% at 45MHz
19:18:15wodzfunman: I do slightly better than OF at least on HD200
19:18:31amiconnSince this isn't possible in practice, it means that it will skip timer interrupts, running at 75 fps effectively
19:18:32saratogaJerom: firmware/target
19:18:33funmanJerom: I don't know right now, I would look at firmware/SOURCES for which files are built on NANO2G
19:18:45funmanwodz: nice!
19:19:13amiconnWhile you might not notice the difference in cube, you probably will with more complex images, e.g. when viewing a .jpg
19:19:19funmanwodz: next step is draw a negative amount of current so the battery recharges while using rockbox.. ;-)
19:19:48amiconnVery light and/or very dark shades will look more grainy when unboosted
19:19:51funmannano2g is built on a s5l8xxxx SoC, right ?
19:19:56wodzamiconn: so maybe it is worth to reduce framerate and allow to run unboosted?
19:20:03saratogapictureflow actually works pretty well for playback now
19:20:06saratogai'm impressed
19:20:08bertrikyes, a samsung s5l8701 IIRC
19:20:40funmanbut it uses USB from another SoC then ? like the AMSv2 ?
19:21:05amiconnwodz: Imo that shouldn't be done
19:21:16saratogai think pretty much every USB controller is on lots of different SOCs
19:21:22wodzamiconn: btw I was thinking if line transfers from dram will help - isn't 8bit width lcd interface the bottleneck?
19:21:24bertrikfunman, you'd have to ask TheSeven
19:21:29amiconnEither force boost, or optimize transfer so that it achieves 150Hz unboosted as well
19:22:17n1syeah, the beast uses the same usb controller as pp
19:22:17TheSevenfunman: it's some synopsys thing, definitely not the one from the s5l8700
19:22:20 Quit einhirn (Quit: Miranda IM! Smaller, Faster, Easier.
19:23:05amiconnOn iriver H1x0, introducing line transfers helped a lot
19:23:42amiconnIt seems that on HD200 the LCD interface is even less of a bottleneck, comparing (ordinary) lcd update speed of H1x0 vs. HD200
19:24:00wodzok I'll try that approach
19:24:20funmankugel: you want to try an e200v1 build with the patched ld ?
19:24:29*FlynDice is trying to run to install eabi toolchain and I get this error: ROCKBOXDEV: "makeinfo" is required for this script to work. Do I find "makeinfo" from linux or Rockbox....?
19:24:50funmanFlynDice: linux
19:25:13funmanhmm don't we use −−disable-docs ?
19:25:39kugelfunman: sure
19:28:09funmankugel: (binutils 2.20.1 patched with
19:28:51amiconnwodz: Should be easier on HD200 than on H1x0 because you don't need head/ tail loops, just single head/ tail handling
19:30:16 Join grndslm [0] (
19:32:50 Quit Zarggg (Quit: Zarggg)
19:32:53 Join lev_ [0] (~lev@
19:33:46ranmafunman: I didn't fix any rockboy warnings, I think it doesn't get compiled on c200v2
19:35:48funmanranma: it gets now ;) tetris even runs fine
19:36:29funmanranma: if you haven't check the audiobuffer size of current svn on your c200v2 (in debug->buffering thread), you might have a little surprise
19:36:29 Quit BHSPitMonkey (Ping timeout: 245 seconds)
19:38:15kugelfunman: \o/
19:38:31ranmaLatest one I compiled is 26670
19:39:05 Join Luca_S [0] (
19:40:11kugelfunman: very worky :)
19:40:28funmaneheh :P
19:40:33funmanwhat's the binsize difference ?
19:40:39*funman looks for a swede
19:40:57kugelfunman: ~170k
19:42:19 Join Horscht [0] (~Horscht2@xbmc/user/horscht)
19:42:47ranmafunman: Let me guess, it's compiling thumb by default now? *g*
19:44:36funmanyeah, also -Os, and the plugin buffer is much smaller
19:45:17Luca_Shello everybody! trying to compile latest svn from vmware dev env. i'm running should I choose the 'eabi' option or the 'arm' option for a fuzev2?
19:47:56funmanBagder: ping, can you upload a file on the download server?
19:48:05BagderI can!
19:48:26funmancan you put on please?
19:48:27Luca_Sthankyou funman. btw, the last commit related to uSD fixed my crashes when copying a song from uSD to uSD - great job!
19:49:15Bagdermaybe we should move that dir to the download server
19:49:27kugelfunman: aren't you supposed to commit it to www/gcc?
19:49:33funmanoh perhaps
19:49:44*kugel doesn't know really
19:50:01Bagderyes the others are committed there
19:50:06funmanBagder: is it worth it? it isn't likely to be downloaded a lot of time
19:50:22funmanindeed, i'll do it myself :)
19:50:22Bagdernah, commit it in www/gcc instead
19:50:43 Quit kugel (Quit: exit(0);)
19:50:54 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:51:31CIA-6New commit by funman (r26783): Add the patch for binutils 2.20.1 thumb interworking fix
19:52:39CIA-6New commit by funman (r26784): rockboxdev: patch eabi binutils to fix thumb interworking mixed with long calls ...
19:53:01 Quit kugel (Client Quit)
19:53:02BagderI svnupped to the site
19:53:07 Join kugel [0] (~kugel@rockbox/developer/kugel)
19:53:17*funman now expects some numbers thumb vs. normal
19:54:01gevaertsDo we need to rebuild eabi?
19:54:23funmanif you want to run it on PP yes (perhaps some other targets too, i can check if you want)
19:54:36funmanthough you can skip the gcc build, binutils alone is quite fast to build
19:55:29gevaertsI mean for the build clients mainly
19:55:37 Join Zagor [0] (bjst@rockbox/developer/Zagor)
19:55:46kugelfunman: maybe put a comment that the patch is only needed for 2.20.x so that someone trying to use newer binutils doesn't get trapped?
19:55:53funman will skip gcc
19:55:58funmankugel: i have put it in the .diff itself
19:56:09funmanpatch will fail anyway on newer binutils
19:56:11kugeloh ok quits with error "mkdir: cannot create directory build-binutils: File exists". what should I do?
19:56:22funmangevaerts: well for now we only build AMSv1/2MB with thumb by default
19:56:30funmanLuca_S: rm -fr /tmp/build-binutils
19:56:52kugelfunman: codecs/plugins seem to work fine also, though I haven't tested a representative set
19:57:01*pamaury is now pretty convinced that the usb controllers of the ipodnano2g and amsv2 might the same, with only a few register mismatch. Even though the rockbox driver seems suspiciously short and the linux one suspiciously long :) And there are some differences in the code :(
19:57:13pamauryWho wrote the ipodnano2g driver ?
19:57:15funmanwell I didn't test each codec/plugin either on AMSv1, but then those are unstable
19:57:18funmanpamaury: TheSeven
19:57:50pamauryTheSeven: how did you write the usb driver of the nano2g ? Did you get some doc somewhere ?
19:57:53ranmafunman: Nice: usefl: 0/543168
19:58:27funmanranma: i think you should make a 1.0 bootloader, unfortunately i can't test it on the other variant because the OF can only run USB connection, nothing else
19:58:33Jerompamaury: Where did you get all the defines in usd-drv...h ?
19:58:44pamauryfrom a linux patch
19:58:53TheSevenyes, I used the S3C6400X datasheet as a base, and corrected some things by analyzing apple code
19:59:08pamauryis the datasheet public ?
19:59:51JeromI think the linux code is for host
20:00:03JeromAnd we are a device ?
20:00:21pamaurynope, this is a gadget driver
20:00:24TheSevenpamaury: it's leaked pretty well
20:00:26ranmafunman: 1.0 bootloader?
20:00:27TheSevenjust google it
20:01:12funmanranma: if we put a tagged binary bootloader on the server rbutil could support c200v2
20:01:20TheSevenand btw my driver is probably much smaller because it doesn't support isochronous, at least not properly
20:02:06pamauryAnd it has hardcoded constants everywhere :)
20:03:05***Saving seen data "./dancer.seen"
20:03:22 Join perrikwp [0] (
20:04:25pamauryTheSeven: is your driver bulletproof or are there things not working ?
20:05:06TheSevennot really bulletproof, but working stable enough to transfer files on an everyday basis
20:05:25TheSevenas soon as you enable the HID however, strange things start to happen, which we haven't tracked down yet
20:05:29Stummiwhere are the regions for the radio-app defined?
20:05:41Stummi(with their ranges and so
20:06:36Luca_Suhm... since the eabi switch the vmware environment has become much more difficult to use for a non linux expert :( what should I do if it says arm-elf-eabi-gcc: command not found? I already ran successfully with the eabi parameter
20:06:57Jerommaybe your PATH is wrong ?
20:07:03StummiPATH=$PATH:/usr/local/arm-elf (or so)
20:07:27Bagderthe vmware image probably doesn't have the eabi compiler installed
20:07:28Stummii am not sure if an export is neccessary here
20:07:53 Quit stripwax (Quit:
20:07:54Zagorthe eabi compiler is in /usr/local/bin, so that needs to be in your path
20:07:58funmanStummi: export PATH=$PATH:/usr/local/bin
20:08:05funmanLuca_S: ^
20:08:13Stummiah ok
20:08:26bertrikStummi, firmware/tuner.c
20:08:34 Join Zarggg [0] (
20:10:07Stummithe third number gives the step, doesn't it?
20:10:15 Join loveless [0] (~loveless@2a00:1328:e1ca:246:216:cbff:feb9:8c8d)
20:10:32bertrikyes, the struct is probably defined in tuner.h
20:12:36 Quit blackromeo (Quit: CGI:IRC (Ping timeout))
20:13:16ranmafunman: Tested and works for me:
20:14:41funmanranma: i think you should build it with modified to print "1.0", and put it on the forum thread for official test builds, if it works on the other variant you can tag the revision you used and ask a swede to put it on the download server
20:17:04Jerompamaury: Have you started working on the USB for as3525v2 ?
20:18:27 Quit loveless (Quit: loveless)
20:20:16pamauryYes,but I'm hestitating. Perhaps we could kind of merge the ipodnano2g driver with the one I'm writing but as I don't have an nano2g, I don't want to break it. I think I'll currently write a new one from scratch. Perhaps with the S3C6400X datasheet + linux driver I can make everything works. The datasheet only contains the register definitions so prehaps the linux code can help
20:21:19wodzhow do we treat datasheets that leaked to the net?
20:21:23 Join steve|m [0] (
20:21:56funmanwodz: with great care, though if they are recent it's better to keep it private and not put them on the wiki
20:23:13wodzI have rockchip rk27xx datasheet but will not upload it than :-)
20:24:34ranmafunman: Where should I post it? "Rockbox Development > Official Test Builds"?
20:25:25 Quit funman (Quit: free(random());)
20:26:23ranmaHmm, I think I have to get the forum developer badge first before I can post there :)
20:26:39pixelmawodz: your latest patch in FS #11189 doesn't seem to apply cleanly even with r26704, if I didn't do anything stupid
20:27:27wodzpixelma: that strange, I'll check when my children finaly will sleep() :-)
20:28:34wodzThere were some talks about tracking connections on clip PCB. I think I can arrange Xray photos of PCB if I get one
20:28:57wodzI don't know if someone is interesting though
20:29:46saratogait would be neat to see what dock pins are wired where
20:29:52saratogaoh the clip doesn't have the dock
20:30:17wodzbeware that Xray photos are not that easy to interpret
20:30:35bertrikWhat could we use xray photos of a clip for?
20:31:01Stummisomebody here with some graphic/pixel-skills?
20:31:15wodz - there are some xray photos of HD200 I took
20:31:55AlexPranma: (Sorry for asking) do you have commit access?
20:32:00 Join Buschel [0] (
20:32:03pixelmawodz: it fails with a few "Reversed (or previously applied) patch detected! Assume -R? [n]" in e.g. firmware/drivers/dac.c . I stopped after that one this time as it resulted in a messed up tree before. I cleaned it up and checked out r26704, an svn diff shows nothing related in these files
20:32:38pixelmathen tried to apply again and got this message too
20:33:12 Quit n1s (Quit: Lämnar)
20:35:30bertrikIt helps sometimes if you know which svn version the patch was against, so you can svn up to that specific revision, apply the patch and then svn back up to the very latest revision. This doesn't work if the patch wasn't made with svn, because then you generally don't know which revision the patch was against.
20:37:17pixelmabertrik: that's what I did (r26704) ;)
20:37:36pixelmaas stated in the flyspray description
20:40:33 Join petur [0] (~petur@rockbox/developer/petur)
20:40:53 Join stripwax [0] (
20:42:17 Join jgarvey [0] (
20:43:03ranmaAlexP: Yes
20:43:31AlexPranma: Then I can give you said shiny developer badge :)
20:44:02ranmaGreat :) I was about to look into the mail to see whom I have to ask, but got sidetracked elsewhere :)
20:44:06AlexPranma: What is your forum username?
20:44:30AlexPah, complicated :)
20:44:38ranmaVery :)
20:46:22wodzpixelma: the problem with that patch is with removing files
20:46:37AlexPranma: There you go
20:47:44pixelmawodz: how did you remove them? I thought a proper svn remove would do it...
20:47:52pixelmabut you used git...
20:48:08 Quit stripwax (Quit:
20:48:09wodzit is created with git so there is some glitch probably
20:48:17kugeldo we have a plan on how to do the transition if we ever need the patched binutils?
20:48:40kugelright now creates a different binutils than 1 hour ago
20:48:48gevaertskugel: I'd say rebuild and call it arm-eabi-gcc444-1 ?
20:48:50Jerompamaury: I will let you do, you seems to have a lot more clue then I do ^^
20:48:51pamauryTheSeven: could you try to run some code on your nano2g to check something about the usb controller ?
20:49:02TheSevenyes, of course
20:49:13kugelgevaerts: configure needs to be aware too
20:49:22gevaertsdoes it?
20:49:28 Join hebz0rl [0] (
20:49:45kugelpeople could complain about non booting builds
20:50:19pamauryOk, it would be nice if you could dump the content of five registers: *(REG32_PTR_T)(OTGBASE + 0x040) and so one: 0x44,0x48,0x4C,0x50
20:50:22pamauryTheSeven: ^
20:50:26 Join stripwax [0] (
20:50:36 Join spyrothehuman [0] (
20:50:55kugelalso, there would be the possiblity to just add the -1 to without actually rebuilding. nobody would notice. But I think we can trust our build clients enough
20:50:56spyrothehumanhey, i got pacman running on rockbox. can i get any other games to work?
20:51:14pamauryTheSeven: this might have undefined effect or even crash it, I don't know if those registers are implement on the nano2g :)
20:51:39TheSevenreading nonexistant memory just returns garbage on that target
20:51:40spyrothehumanit runs like a dream
20:52:12spyrothehumanAnyone here working on arcade emulation on rockbox?
20:52:20 Quit simonrvn (Ping timeout: 272 seconds)
20:54:40wodzpixelma: I uploaded new version of the patch created with git diff -M. This should preserve history of the files and make svn happy
20:54:54 Join simonrvn [0] (
20:54:56AlexPspyrothehuman: Nobody to my knowledge
20:54:57TheSevenpamaury: which range of regs do you need?
20:55:25 Quit kaylinsigswort (Remote host closed the connection)
20:55:37spyrothehumanIf anyone in this room has any knowledge of porting arcade games on rockbox, please contact me
20:56:06pamaurythe one I wrote you above: five regs starting at *(REG32_PTR_T)(OTGBASE + 0x040)
20:58:37pamauryTheSeven: ideally, the first one should value 0x4f54260a if it's the same controller
20:58:50spyrothehumanwhat happened to the application on rockbox that was kind of like an oscilliscope. it was called something like ffs
20:59:03spyrothehumandid it get taken off?
20:59:24wodzfft You mean - its fourier transform not osciloscope
20:59:43spyrothehumanyes, that was what i meant
21:00:05spyrothehumanwas fft canned?
21:00:17wodzI was wondering how fast would work dosbox if ported to rb :-)
21:00:22wodzspyrothehuman: no
21:00:33TheSeven[0x38800040] = 0x4F54220A
21:00:33TheSeven[0x38800044] = 0x00000264
21:00:33TheSeven[0x38800048] = 0x228DD9D0
21:00:33DBUGEnqueued KICK TheSeven
21:00:33TheSeven[0x3880004C] = 0x050004E8
21:00:33TheSeven[0x38800050] = 0x01F08001
21:00:45TheSevenno exact match, but really close
21:01:11TheSevenpamaury: ^
21:01:14pamauryok, I'll analyse them and told you
21:02:08spyrothehumanwhere can i find a copy of fft? i really like that program
21:02:14pamauryOk, my guess is that the nano2g also has a synopsis one but a older one
21:02:16ranmaAlexP: Thanks, posted as
21:02:32Zagorspyrothehuman: it's in the default builds of rockbox
21:02:41spyrothehumanranma was your name inspired from the anime ramna 1/2
21:02:50spyrothehumanthanks zagor
21:02:59ranmaYes, it was
21:03:02spyrothehumanwhen i become rich and famous you will get a bently
21:03:16spyrothehumannow, i dont know where the keys are
21:03:43ranmaDamn, it's getting brighter outside already...
21:03:45spyrothehumanperhaps you could use a wrench to force the car on
21:03:52CIA-6New commit by bieber (r26785): Theme Editor: Switched error highlighting to the line numbers
21:03:52spyrothehumanperhaps not
21:04:32spyrothehumanThe sansa e200 series are becoming really rare
21:04:44spyrothehumanwow, horrible grammar
21:05:10wodzI'll repeate my earlier question - how to use lcd_puts_scroll to do it scroll actually?
21:05:33spyrothehumananyway, sandisk is not making the e200 series and in a couple of months the price of the sansa e200 line will skyrocket
21:05:39TheSevenwodz: what exactly is the problem?
21:05:44TheSevenare you testing some early code?
21:05:50CIA-6New commit by jethead71 (r26786): gigabeatfx to eabi buildservers (trunk change coming shortly).
21:05:51TheSevenit might be that you never yield
21:06:04 Quit chrisb (Ping timeout: 245 seconds)
21:06:18wodzI want to scroll text in view_battery() from debug_menu.c
21:06:26spyrothehumando you think an atari 2600 emulator could be ported on the sansa e260
21:06:58wodzI repleaced lcd_putsf with snprintf & lcd_puts_scroll and nothing changed (no scroll)
21:07:05CIA-6New commit by jethead71 (r26787): Build Gigabeat F/X with EABI tools.
21:07:23wodzTheSeven: ^
21:07:26saratogaspyrothehuman: i don't think anyone could answer that without looking at existing emulators
21:07:35saratogabut assuming a suitable one exists, probably
21:07:41 Join BHSPitMonkey [0] (~stephen@unaffiliated/bhspitmonkey)
21:07:49TheSevenhmm, no idea
21:08:04saratogawhy not find a suitable one and see for yourself
21:08:05spyrothehumani know that stella couldnt work because it is meant to be used with a mouse
21:08:31spyrothehumanbut what if the cursor could be controlled with the up down left right keys
21:08:55spyrothehumanhmm, a programming i will attempt
21:09:04spyrothehumanill google it
21:11:32TheSevenare rasher's buildgraphs dead?
21:11:46TheSevenand what the hell could have caused this!?;type=ipodcolorsim
21:12:24wodzspyrothehuman: rather unlikely to run stella on rb targets -> 32MB of ram highly recommended (only a few will fullfill this)
21:12:26saratogais that a mac?
21:12:50saratogaprobably half our targets have that much memory
21:13:37TheSevensaratoga: referring to my build log?
21:13:42TheSevennope, it's a P4 singlecore
21:13:56TheSevenubuntu karmic
21:14:15TheSeven(should probably update to lucid some day)
21:14:56TheSevenand everything worked fine before i rm'ed the /usr/local/<arch>/ directories and re-ran
21:15:00wodzmay it be that default viewport has configured scroll style to noscroll or something?
21:15:34pixelmawodz: same thing with the new patch
21:15:37CtcpIgnored 1 channel CTCP requests in 0 seconds at the last flood
21:15:37*kugel is happy to see the eabi transition going on
21:15:41ZagorTheSeven: did you remember to remove /usr/local/<arch> from your path?
21:16:03wodzpixelma: ok I'll take closer look than
21:16:03Zagorbefore running
21:16:45TheSeventhey should have never been there for the root account in the first place
21:16:46TheSevenlet me check
21:17:19Zagorwe have had a number of really weird compiler issues from people who have built them with /usr/local/arm-elf/bin in their path
21:17:39TheSevenOK, I can recompile the toolchains, but I strongly doubt that this was in the path at that time
21:17:56TheSeveni directly ssh'ed into it as root, and root never had those in his path, only the other accounts
21:18:09 Part watto
21:18:17CIA-6New commit by bluebrother (r26788): Rework Rockbox Utility update version number check. ...
21:18:20TheSevenwhich reminds me that i might have svn up'ed as root, let me check some permissions
21:19:26TheSevenoh, was updated within the last hours?
21:20:16jhMikeSkugel: I'm gonna run through checking the arm targets I can do this on atm, one by one
21:20:27ZagorTheSeven: yes, with a binutils patch
21:20:34TheSevenyes, i just saw it
21:20:43TheSevenshouldn't have caused any trouble over here though
21:21:49 Quit hebz0rl (Ping timeout: 248 seconds)
21:21:53TheSevenhm, getting loads of "/usr/local/arm-elf-eabi/bin/arm-elf-eabi-gcc: No such file or directory" during make reconf, so apparently something survived which shouldn't have
21:21:57TheSevenno idea why though
21:22:13domonoky1bluebrother: why is the rbutil mac rebuild named 1.2.5-1 and not ?
21:22:19TheSeveni didn't have that in my path, ran rm -rf /usr/local/whatever, and then
21:22:36domonoky1with all dots, the old version checking would have worked fine :-9
21:22:37TheSevenbtw, this looks weird: Using arm-elf-eabi-gcc 4.4.4 (404)
21:23:03TheSevenwhat does that 404 mean?
21:23:39Zagorlooks like buggy version number parsing
21:23:58domonoky1bluebrother: version check gets difficult if we invent a new schema evertime we do a bugfix release :-)
21:23:59kugeljhMikeS: checking e200 right now
21:24:43Zagorah no, it's actually intentional: # 3.3.X => 303, # 3.4.X => 304
21:24:51kugelTheSeven: major*100+minor
21:24:53*jhMikeS was going to check e200v1 (but actually already did before), has h10(both) and 3g ahead
21:25:27bluebrotherdomonoky1: because it's not a new version but just a new build
21:25:30kugelI don't know if the number is acutally used though
21:25:44domonoky1bluebrother: like those 1.0.7b versions ?
21:25:47bluebrotherdomonoky1: yes, it might have been a bad idea. However, version detection works no :)
21:26:05bluebrotherand it works both with 1.0.7b and 1.2.5-1 :)
21:26:17Zagorkugel: it's used to disable some warnings for higher gcc versions
21:26:39bluebrotherTheSeven: 404 means file not found :)
21:26:53kugelit's maybe because gcc had a different naming scheme in the 2.x series
21:27:36TheSevenrebuilding tools right now
21:27:55bluebrotherdomonoky1: the problem with the old version comparison is simply that it expected us to not use any other character for separation than a dot. However, someone (err ... who? ;-) did
21:29:47bluebrotherI created a small test for the new comparison. Have to clean that up a bit, but I'm planning to commit that later. Any ideas where to place it? I was thinking about a test/ subfolder below rbutilqt, and then creating a new subfolder for each unit test. Or put all unit tests (it's only one, but I'm planning to make more use of that in the future) in that test/ folder and distinguish them by filenames only?
21:29:56*TheSeven is rebuilding toolchains
21:30:35TheSeventhis time, /usr/local definitely didn't contain any files having "arm", "eabi", "mipsel", ... in their name any more
21:30:55*domonoky1 is fine with starting with rbutil/test, if it gets too many files, we can always create subfolders.
21:30:59pixelmahmm... why are the "channels" options and their explanation in a table? I don't think this way is used anywhere else in the manual
21:31:27kugeljhMikeS: the bootloaders should be verified as well
21:34:23jhMikeSkugel: true enough. if you have any of those I listed, you check loaders, or I'll check loaders and go with parallellism?
21:35:32CIA-6New commit by kugel (r26789): Build Sansa e200 with new eabi gcc
21:36:20CIA-6New commit by kugel (r26790): Build Sansa e200 with new eabi gcc.
21:36:43*bluebrother spots the upx website looking completely different
21:36:53pixelmathe channels settings explanation in a table seems to be an age old thing. Should it be converted to \item style?
21:36:55*TheSeven still likes the idea of a rednag bot ;-)
21:37:50pixelma(bluebrother, AlexP or others) ^^
21:38:19bluebrotherpixelma: if it's done differently in all other places, yes. Consistency is a good thing :)
21:38:28ZagorTheSeven: that idea isn't dead
21:38:48TheSevenwhat's holding it back?
21:38:57AlexPpixelma: See what bluebrother said :)
21:39:10TheSevenis there some reasonable hook from the build system for failed builds?
21:39:16AlexPTheSeven: We had red emails at one point that seem to have gone away
21:39:19bluebrotherrednag boot?
21:39:33ZagorTheSeven: we can always make one
21:39:42kugeljhMikeS: I have only 1 non-eabi target left (yh925), I verified that one a while ago
21:39:42AlexPbluebrother: I presume a bot like CIA-*** to nag on a red build
21:39:46bluebrotherah, misread bot for boot :o
21:39:53*jhMikeS does 3g then
21:40:01ZagorI didn't really want to create a custom bot for this though. I was thinking of having the cia bot(s) do this. you can send custom messages for them to say.
21:40:03*bluebrother would like a build result announcement
21:40:23bluebrotherlike "build completed in 1:23, 1 successful, 2 warnings, 3 failed
21:40:29AlexPbluebrother: that'd be good, it'd cover all bases
21:40:29 Quit perrikwp (Quit: CGI:IRC (Ping timeout))
21:40:29*TheSeven would like to automatically get highlighted on IRC if there is something wrong with his commits
21:40:45AlexPTheSeven: That'd need a SVN/IRC ncik lookup table
21:40:45bluebrothergood idea.
21:40:52AlexPbut that should be easy :)
21:41:05TheSevenAlexP: one could setup a highlight on his SVN name
21:41:24TheSevenand as both are identical for me, I don't need to care ;-)
21:41:36AlexPmine aren't :)
21:41:44AlexPclose though
21:41:53TheSevenjhMikeS' aren' either, which already confused me earlier today
21:42:02bluebrotherit could simply use the svn names, and devs using different nicks can setup the hilight on their side
21:42:22CIA-6New commit by kugel (r26791): Get rid of the STATICIRAM hack for eabi builds. ...
21:42:44pixelmaAlexP: ok... hooray for things you find when working on something else though ;)
21:42:53 Quit flydutch (Quit: /* empty */)
21:44:34 Quit spyrothehuman (Quit: CGI:IRC (EOF))
21:45:02jhMikeS3g bootloader = boots and loads (so it's still a bootloader)
21:45:33bluebrotherotherwise it would be a waitloader? ;-)
21:45:39amiconnwodz: Shall I commit my patch, or wait so you can optimize it further using line transfers?
21:45:41bluebrotheror bootwaiter?
21:45:49AlexPpixelma: indeed :)
21:46:08wodzamiconn: commit it
21:46:09jhMikeSa bootless freeloader?
21:48:13CIA-6New commit by user (r): logmsg
21:49:33wodzkugel: you have experience with git - how to produce patch from git after move/rename to be accepted by diff or svn diff
21:50:02kugelwodz: git handles these, so just do git diff or git format-patch
21:50:12kugelyou need to use git mv though
21:50:37bluebrotherwodz: afaik it isn't possible to make a patch that svn recognizes a move within
21:50:42kugelyou can commit from git directly too
21:50:56bluebrotherhowever, git-svn automatically handles moves / renames
21:51:14wodzkugel: I know about git-svn
21:51:35CIA-6New commit by bieber (r26792): Theme Editor: Fixed bug in parser handling empty lines and made ParseTreeModel handle the new VIEWPORT element properly
21:51:45pixelmait would be cool if I could test that patch but I'm not really motivated to learn and use git for that (on cygwin too)
21:51:57wodzthe problem is I want create patch that others can test
21:52:15bluebrotherpatches created with git are no problem to test for non-git users.
21:52:23bluebrotheryou just need to use -p1 instead of -p0
21:52:26kugelwodz: the patch will have the move/rename
21:52:50pixelmayes, but the one he has now doesn't handle the move correctly it seems
21:52:54jhMikeSbluebrother: I've had to convert those, -p1 just didn't do it. maybe the patch version is too old?
21:52:58bluebrotherand a file renamed will simply show up as getting removed and added again in a diff.
21:53:03kugelit's just that svn won't recognize as move/rename but as deleted 1 file, created 1 file
21:53:19CIA-6New commit by amiconn (r26793): Port greylib blitting optimisation to MPIO HD200. ISR speedup is ~10%; further speedup should be possible by using line transfers for accessing the ...
21:53:21bluebrotherjhMikeS: no idea. Works fine for me, and I'm using it since quite a while
21:53:23Stummii formated my sansa fuse v2 because i couldn't mount it with linux (debian) and installed the current rockbox. Now rockbox can't boot and brings the message "No partition found" on booting. Booting the OF ist possible. Any ideas?
21:53:37pixelmajust changed files apply correctly with -p1 and I knew that
21:53:39Stummibefore formating it worked
21:54:01bluebrotherthe only important thing is to manually svn mv the file when committing with svn
21:54:17Zagoris anyone comfortable with the filter syntax? I find it rather confusing...
21:54:44*bluebrother never seen that syntax
21:55:02kugelZagor: thanks for fixing the git mirror
21:55:51*Zagor bows :)
21:58:27amiconnDoes r26784 mean we have to rebuild the eabi toolchain?
21:58:43amiconnAlso, does the buildserver check that somehow?
21:58:49kugelnot necessarily but I guess it's recommended
21:59:24kugel(if you want thumb on PP you should)
21:59:42pamauryTheSeven: (comparison of clip+ vs nano2g usb controller)
21:59:44Zagorkugel: so the change is not used by any of the auto-built targets?
21:59:51literalso, I'm just curious, what effect does this toolchain switch have?
22:00:05kugelZagor: yes
22:00:13pamauryIF, I repeat it, IF the hardware config registers are the same, then they are very similar but there are a few changes here and there
22:00:24Zagorkugel: what does yes mean? :-)
22:00:57pamaurythe most important one would be that the clip+ has dedicated tx fifos (even though I'm unsure of the impact)
22:01:04CIA-6New commit by user (r): test log msg
22:01:04wodzkugel: could You look at FS #11189 last two patches - maybe you will have a clue why it produce rejects
22:01:07kugelonly some ams sansas use thumb, they don't need the change because the bug isn't exposed to to memory layout
22:02:27kugeldue to*
22:02:45kugelit only occurs with long calls. you never have long calls on them because all code is within 32MB
22:03:01CIA-6New commit by bluebrother (r26794): deploy-release: Make upx part optional and disable it. ...
22:03:08***Saving seen data "./dancer.seen"
22:03:10kugelon PP however, the IRAM is far away, so calls to it from RAM need long calls
22:04:25jhMikeSkugel: the other way too :)
22:07:21 Quit Buschel (Ping timeout: 260 seconds)
22:07:22jhMikeSahh....3g stkov 40004350 (1) <== ?? (perhaps too small on the idle stacks) :\
22:09:14jhMikeSyeah, e200 is using 93% of it now on core 1, probably it
22:11:46CIA-6New commit by bieber (r26795): Theme Editor: Made editor highlight all errors in a document, not just the first one
22:13:30 Join hebz0rl [0] (
22:18:31CIA-6Build result: test log msg
22:18:41CIA-6New commit by user (r): test log msg
22:18:49Zagoryesss, nailed it
22:21:48*jhMikeS is having issue with 3g...moving on for now...
22:23:22wodzpixelma: I think just for testing You can ignore rejects. SOURCES is modified so files which should be deleted (but are left due to rejects) are not compiled anyway
22:23:35CIA-6New commit by pixelma (r26796): Replace the table that was used for the channels settings option with the item list which is used for this kind of settings lists everywhere else in ...
22:24:38pixelmahrrmm... the message could have been worded a bit better :\
22:29:23 Quit dfkt (Quit: -= SysReset 2.53=- Ph'nglui mglw'nafh Cthulhu R'lyeh wgah'nagl fhtagn.)
22:31:40CIA-6New commit by pixelma (r26797): Rework the button table in 'Navigating the Main Menu' - use \newline to keep line breaks inside the cell which makes the alternating row colours work ...
22:37:01 Quit Topy44 (Ping timeout: 248 seconds)
22:37:07 Part Jerom
22:37:45 Join Jerom [0] (~Jerom@
22:38:19jhMikeSh10 20gig PLUGIN_IRAM overflows building mp3_enc.codec :\
22:39:36 Join Topy44 [0] (
22:40:44 Quit pamaury (Remote host closed the connection)
22:44:50 Quit BHSPitMonkey (Ping timeout: 245 seconds)
22:50:42 Join M3DLG [0] (~M3DLG@
22:51:51CIA-6New commit by bieber (r26798): Theme Editor: Fixed status bar update bug
22:58:33 Join anewuser [0] (anewuser@unaffiliated/anewuser)
23:04:40Jerommysql down ?
23:05:32gevaertshm, Zagor did something wrong apparently
23:10:52jhMikeSkugel: didn't the bootloaders need arm-eabi-gcc444 too in www/buildserver/builds? (I have to change that for the gigabeat f/x/s then too)
23:12:06 Quit DerPapst (Quit: Leaving.)
23:12:40 Quit DataGhost (Ping timeout: 265 seconds)
23:12:52CIA-6Build result: All green
23:13:55bluebrothernow make that green green :)
23:14:04gevaertsZagor: maybe add the revision?
23:14:13*TheSeven 's build client seems to be working again
23:14:14CIA-6bieber: r26798 build result: All green
23:14:28bieberWhoah, wat was that?
23:14:31TheSevenZagor: oh no, don't ping if it's green
23:15:03*bluebrother still misses the colors ;-)
23:15:15kugeljhMikeS: right, I forgot about htat
23:15:29jhMikeSkugel: I'll get it. I got H10 20GB coming anyway.
23:15:51Zagorbieber: build result spam coming up :)
23:16:23bieberOh, okay :)
23:17:06CIA-6bieber: r26798 build result: All green
23:17:17kugeljhMikeS: e200 needs it as well
23:17:23 Quit kugel (Quit: Ex-Chat)
23:17:39CIA-6r26798 build result: All green
23:17:47CIA-6New commit by jethead71 (r26799): Use EABI toolchain for H10 20GB. Also, arm-eabi-gcc444 wasn't specified on some bootloaders where it should be now.
23:17:53pixelmacool :)
23:17:59CIA-6r26781 build result: 726 errors, 807 warnings (theseven)
23:18:21pixelmasomeone commit something broken... ah yes ;)
23:18:23 Quit Tarvis (Quit: /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit /quit)
23:20:46 Quit CGL (Remote host closed the connection)
23:21:37CIA-6New commit by jethead71 (r26800): Build h10 GB with EABI toolchain. Remove some pointless ICODE attributes from mp3_enc.codec and, to make it build, move one function out of IRAM there ...
23:21:53pixelmabluebrother, AlexP: we have the same note about bookmarking not working under certain conditions in the main menu section (bookmark list button table) and in the extra section on bookmarks. Is that correct? I'm slightly worried that they could get "out of sync", the aren't currently though
23:23:25CIA-6r26800 build result: All green
23:24:39CIA-6New commit by bieber (r26801): Theme Editor: Made errors display in status bar when cursor is on error'd line
23:28:00bieberbluebrother: I got your per-line error display idea working :)
23:28:17gevaertsZagor: is the build system stuck?
23:28:37ZagorPerl error: Illegal division by zero at line 1075....
23:29:37Zagorweird: Total speed: 0 points/sec (63 clients)
23:30:29Zagorrestarting round now
23:31:46CIA-6r26801 build result: All green
23:32:13bluebrotherbieber: nice :)
23:40:01 Join Buschel [0] (
23:40:57pixelmabluebrother, AlexP: we also not only have the note in double but alsoi the button table it seems (bookmark list)
23:43:57rasherTheSeven: not dead. Recuperating
23:44:09kisakso alot has happened in the last several hours ... if I'm understanding this right, we're now using eabi except for stable targets that have not been tested yet
23:45:13jhMikeSwell, ipod 3g got tested and failed outright
23:45:25 Quit hebz0rl (Quit: Ex-Chat)
23:46:24kisaksounds like it'll soon be time to call for benchmarks on all targets
23:47:28JeromWell, it works on fuze v2 ^^
23:52:46 Quit toffe82 (Read error: Connection reset by peer)
23:57:06 Quit komputes (Remote host closed the connection)
23:57:21 Quit stripwax (Read error: Connection reset by peer)

Previous day | Next day