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

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

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

#rockbox log for 2009-07-10

00:00:56 Quit mirak (Remote closed the connection)
00:01:24 Quit n1s (Read error: 110 (Connection timed out))
00:06:41obokugel: okay - that took a little too long to sink in here. Will look up what I need to send.
00:08:58CIA-71New commit by bertrik (r21735): S5L8700: use wakeup_wait/wakeup_signal instead of polling for i2c communication
00:09:52BagderZagor: I think we should let the client exit if it had no server-traffic in 30 seconds or so
00:11:33ZagorBagder: yeah
00:11:54Zagorhave you looked at the server ping logic?
00:12:11Bagderno, but I think the problem is client-side
00:12:19Bagderie it doesn't "see" the server drop off
00:12:24Bagderso it just sits there waiting
00:12:37Zagorbut the server doesn't disconnect either, does it?
00:12:46BagderI don't know
00:12:54Bagderit should
00:12:59Zagoranyway, it can't hurt checking on both sides
00:14:38Zagor"make zip" fails for jjim. that's why there are missing builds in the size table. I
00:14:47ZagorI'm adding a check for that now.
00:15:19Bagderaha
00:15:36Bagderthat reminds me of a 'test' option for the client
00:16:31Bagderbut perhaps probing for most of the tools at startup should be enough
00:17:22 Quit BdN3504 ("CGI:IRC (EOF)")
00:17:32 Join Thundercloud [0] (i=thunderc@persistence.flat.devzero.co.uk)
00:18:57kugelZagor, Bagder: Why are builds killed before the round ended?
00:18:59 Join wincent [0] (n=wincent@host-091-097-067-213.ewe-ip-backbone.de)
00:19:09Bagderkugel: when another client completed it
00:19:50kugelWhy are clients getting the same build in the middle of the round? I thought this speculative parallel build was about when the round ends
00:20:15Zagorit starts as soon as all builds are handed out once
00:20:30kugelI could easily have two builds, but now I get 3 of 4 killed
00:20:34Bagderit'll happen pretty soon as some clients are really fast
00:21:19Bagderit's all about how the builds are distributed to what clients
00:21:28Bagderand that's why a better speed meter is useful
00:21:39kugelShouldn't it give builds that have not been assigned yet, and only assign builds of other clients if no builds are left?
00:22:18Zagorthat's what it does
00:22:31Bagderit gives the builds with the lowest number of assigns
00:22:39Bagderthat aren't completed
00:23:28kugeloh, I think all my builds got killed
00:23:38Zagoryep, looks like it
00:24:00ZagorI'll be adding some stats info so we can measure how much cpu time was "wasted"
00:24:37kugelI mean, instead of giving someone my ipodnano build, the other cliend should rather have taken what's next (or even last) in my list
00:24:54 Quit soap ()
00:24:54Bagderkugel: you need to read my explanation
00:24:57Bagderwe already do that
00:25:41Bagderwe hand out builds in a fine order to the clients we think are most suitable
00:25:50Bagderthe order is pretty easy to understand
00:26:02kugelyea, that's why I get ipod nano and ipodvideo64mb?
00:26:12kugelwith lowest bogomips?
00:26:42Bagderwhat are you suggesting?
00:27:05kugelwhy am I getting the hardest build if it's meant to suit the clients?
00:27:19Bagderthey're not
00:27:20Zagorkugel: you don't. your low bogomips means you don't get the heaviest builds in the first handout.
00:27:46kugelipodnano was my first build, ipodvideo64mb the 2nd
00:28:26kugelthat was the first handout, and those are some of the hardest, aren't they?
00:28:48SyriusBagder http://www.youtube.com/watch?v=EIyixC9NsLI&feature=channel
00:28:55Bagderipodnano is yes
00:28:55kugelI actually never got a quick bootloder or so
00:28:59Bagderipodvideo64mb is not
00:29:08BagderSyrius: what are you talkign about?
00:29:50Bagderbut yes kugel should not get ipodnano first I'd say, if joined from the start
00:29:50Zagorkugel: you're right. It looks like the order is reversed :-)
00:30:10Zagorkugel got the heaviest, gevaerts the easiest!
00:30:13gevaertsYou mean I'm not going to get all the bootloaders anymore?
00:30:16Bagderthat also explains why gevaerts gets to many bootloaders
00:30:19kugelomg.......
00:30:23*gevaerts is not happy!
00:30:35kugelgevaerts: you could've told that earler
00:30:36rasherOh dear
00:30:46 Quit robin0800 ("Leaving")
00:30:48 Quit J-23 (Read error: 113 (No route to host))
00:30:54ZagorBagder: that's on sub sortbuilds, right?
00:30:56Zagor*in
00:31:01Bagderyes
00:31:08Bagderor possibly the client sort
00:31:22kugelbtw, when are we getting build times in new.cgi?
00:31:40Zagor"soon" :)
00:31:52 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
00:33:08*JdGordon| thinks he remembers some discussion about wanting to give big builds to slow clients
00:33:25ZagorJdGordon|: why would we want that?
00:33:32gevaertsJdGordon|: maybe in the old system
00:33:40kugelalthough, my initial point remains. My build was cancelled 4-5min before the build round ended. It seems to me that the handout doesn't prioritise builds that another client already started lowest
00:33:47JdGordon|yeah, it doesnt make much sense.. but I'm sure i rmember the discussion
00:34:09krazykiti can see the reasoning on a relatively small fleet of computers
00:34:24krazykityou get several builds out of the faster ones in the time it takes the slow one to do the one
00:34:29Bagderkugel: why not? we have very fast clients involved
00:34:38kugelshouldn't it work like: client requests new build -> give him a unhanded build; if there's no unhanded build give him a unstarted instead, if there's no unstarted give him a started one
00:34:46gevaertsIt does make some sense in the old system, in that it could avoid them starting a second big build
00:34:51rasherkugel: That's what it does
00:34:58rasherWell, barring any more sorting errors
00:35:12Bagderthe list is sorted on "handed out count" and then on "weight", excluding the ones already completed
00:35:27bertrikgevaerts machine is just very fast and probably responsible for a lot of kills
00:35:34kugelit doesn't recognize already started builds?
00:35:44Bagder"handed out count" is exactly that
00:35:46gevaertsI must say that I haven't seen many killed builds on it :)
00:35:47Bagdergiven to a client
00:35:57Zagorkugel has a point. we don't know when builds are started, only handed out. that means the builds we hand out again are likely builds that are already running on other targets rather than those in their queue
00:36:01rasherBagder: handed out count also includes builds that haven't been started yet though
00:36:02kugelhanded out != started. I mean started as in actively compiling
00:36:21Bagderwe have no concept of "started" in the server
00:36:24*rasher still doens't quite understand the need for the queue
00:36:40Bagderrasher: upload in the bg is the first explanation of course
00:36:56rasherBagder: Why doesn't the client just request a new build before uploading?
00:37:01 Join webguest93 [0] (n=d8870b32@gateway/web/cgi-irc/labb.contactor.se/x-c9deab1e8c2d7d48)
00:37:07kugelBagder: if we wouldn't hand out 3 builds at once at the start, then handed out == started
00:37:08rasherOr after starting the upload. Whichever
00:37:16Bagderwell, because clients don't request builds at all ;-)
00:37:21 Quit webguest93 (Client Quit)
00:37:26kugelwasn't that the plan, initially?
00:37:30rasherThis seems sub-optimal
00:37:32Bagderkugel: and why would that improve anything?
00:37:40BagderI don't get it
00:37:48 Quit HBK (Read error: 104 (Connection reset by peer))
00:37:57rasherBagder: Because there's less of a chance that a client gets its running build killed, rather just one picked from its queue
00:38:01 Join HBK [0] (n=hbk@pool-71-96-74-73.dfw.dsl-w.verizon.net)
00:38:14Bagderin theory perhaps
00:38:19BagderI doubt it is like that in real life
00:38:19rasherI *really* don't understand why the client just doesn't say "I'm ready for another one"
00:38:26kugelas Zagor said, it's not unlikely to hand a build out that a client is about to complete, instead of giving one that's only in their queue
00:38:46Bagderbut it's not a likely scenario
00:38:56kugelBagder: my ipodnano build was killed near the end even though I had 2 other not even started builds
00:39:01Bagderclients do the builds in the order they get them
00:39:03ZagorI think we'll need a "gimme more" command anyway, once we start doing single-thread builds.
00:39:24Bagderkugel: that doesn't prove anything
00:39:32Bagderon the contrary
00:39:32rasherBagder: But why risk it? Why have any builds that are not started in the queue?
00:39:40Bagderrisk?
00:39:43Bagderit's not a risk
00:40:05 Quit FlynDice_ (Remote closed the connection)
00:40:13Bagderthe one you've worked on the longest is the one your building/uploading
00:40:13rasherWell, it's not as fast as it could be if you kill a build even if there are unstarted builds around
00:40:30Bagder_that_ one was the highest prio for you to complete
00:40:31Zagorkugel: what happens at one client doesn't really reflect what happens at the server
00:40:32Bagderand you failed
00:40:36Bagderkill
00:40:39JdGordon|FAILED!
00:40:42Bagderit doesn't matter if you introduce a command
00:40:47Dhraakellianls
00:40:50Dhraakellianwrong window
00:40:56kugelgiving a build that's currently compiling instead of a queued one is just illogical
00:41:03rasherkugel: agreed
00:41:12Bagderits perfectly logical
00:41:16JdGordon|no its not... it could be stuck on a really slow server
00:41:22ZagorI agree too. I don't quite see your point Bagder
00:41:35kugelJdGordon: that's no problem
00:41:42 Join AndyIL [0] (i=AndyI@212.14.205.32)
00:41:44JdGordon|you could add some logic to only do that if the <score> is better though
00:41:45rasherJdGordon|: So what? If there are other builds around, let the slow server finish rather than *wasting* the time that server's spent on the build
00:41:59kugelif there's no queued build left, currently compiling builds would be handed out. whoever finished first wins
00:41:59JdGordon|ah, yeah ok
00:42:01Bagdermy point isn't so much that I dislike the idea of changing this
00:42:22Bagderit's mostly that these guys argue as if this is something that'll make a big diff
00:42:33Bagderand I don't believe that
00:42:42bertrikit's not about spending time most efficiently, but getting the builds done quickest
00:42:43JdGordon|only one way to find out for sure though :D
00:42:44kugelI believe it does.
00:42:49rasherIt just seems more complex and less efficient than simply "give me a build" "okay"
00:42:53ZagorBagder: well, possibly not. but in theory it can make a difference.
00:42:54kugelI often see even my first build getting killed
00:43:09BagderZagor: yes, and I'm not against it
00:43:10rasherIt won't make the round end twice as fast
00:43:22rasherBut it'll invariably shave off some amount of time
00:43:25gevaertslet's first see what happens with the fixed sorting
00:43:43Bagderrasher: the thing is that we don't add things we don't need, hence we never had such a command
00:43:47Zagorgevaerts: well the underlying issue will remain
00:43:57Bagderthe point with a queue was to allow the client to build in parallell
00:44:10Bagderfor single-threaded builds, such as manuals and voices
00:44:14rasherBagder: I don't see why this is an argument? The client can still build in parallell surely
00:44:24rasherJust ask for a new build when it's ready..
00:44:25kugelI recall this command was sort of planned anyway
00:44:29gevaertsZagor: of course, but I expect the symptoms to look a bit different
00:44:34BagderI'm just explaining
00:44:43ZagorBagder: but with 8-core servers, we'd need 8 builds in queue. It would be simpler to just have the client say "I can take one more, please"
00:44:47Bagderas we saw no need for such a command, we made none
00:44:50Bagderindeed
00:44:54BagderI'm just explaining
00:44:57Zagorok
00:45:07gevaertsThe server maintains a queue per client, right?
00:45:16Bagderyes
00:45:17rasherI guess I'm just slightly surprised you came up with the most complicated solution
00:45:24BagderI disagree
00:45:28kugelI actually thought the clients requesting builds was one of the main ideas initially
00:45:30*JdGordon| also diagrees
00:45:37JdGordon|there is always a mroe complicated solution
00:45:40Zagoruh, the most complicated? I can think of _lots_ more complicated solutions! :)
00:45:45Bagderkugel: it wasn't from me at least
00:45:54Zagorme neither
00:45:55rasherAnyway!
00:46:05kugelthen my memory is wrong
00:46:06gevaertsMaybe you could increase the allocation count for the first build on every queue?
00:46:08rasherI'm willing to stop and look forward
00:46:27rasherOr ignore the whole thing. Whichever.
00:46:33Bagderadding a "gimme next" command for the clients should be fairly easy
00:46:41kugelI'm thinking the whole queue'ing could be dropped for client requested builds
00:46:50Bagderno
00:46:56rasherHuh?
00:46:58Bagderit still wants a new before the previous is done
00:46:59Zagorgevaerts: queuing is the problem, really. we should just skip the server queues and add the "gimme more" command
00:47:04Bagderthat's a queue
00:47:14Bagderisn't that a queue?
00:47:20rasherBagder: Why does it want a new before the previous is done?
00:47:20Bagdertwo outstanding builds
00:47:25rasherWell, depending on threads
00:47:28Bagderbecause it is uploading
00:47:40kugelcompile->request new build->upload
00:47:40Zagorwell, it depends on definition. the queue would only contain running builds then
00:47:41Bagderthen it wants to start the next, before the ul is done
00:47:44gevaertsok, so you want to replace pre-emptive queue filling by filling on demand?
00:47:50JdGordon|I would argue that A and C in the wiki are "the client asking for another build"
00:48:11rasherBagder: Yeah, that's a definition. I mean a queue of not-even-running builds
00:48:14kugelBagder: then just request before uploading
00:48:30Bagderkugel: that's what I said, but that makes it a queue in my terms
00:48:39Bagderbut not in yours ;-)
00:48:48rasherSo we all agree
00:48:51rasherAnd cake for everyone
00:48:53kugelseems so :>
00:48:55Bagdercake!
00:49:18kugelBagder: it's also a queue, but not the same as the current one :)
00:49:29Bagderyes it is, just shorter ;-)
00:49:45Bagderthe server won't have to treat it any differently
00:51:06ZagorBagder: do you have time to do this, or should I dig in?
00:51:40Bagderplease go ahead, I'm stuck in another project atm
00:51:47Zagorok
00:51:54BagderI'll review the commits
00:55:04CIA-71New commit by zagor (r21736): Check that zip is in path. Check that make zip succeeded.
00:55:14***Saving seen data "./dancer.seen"
00:55:22 Quit AndyI (Read error: 110 (Connection timed out))
00:56:49 Join soap [50] (n=soap@rockbox/staff/soap)
00:57:06kugelbertrik: What does a wakeup offer what a simple volatile variable doesn't?
00:57:51rasherI guess doing away with the non-building-queue does mean that slow clients will only get one bootloader to build
00:58:20mcuelenaerekugel: yielding?
00:58:25 Quit Hillshum ("ChatZilla 0.9.83 [Firefox 3.0.3/2008092417]")
00:58:27mcuelenaere+ a sleeping thread
00:58:28bertrikkugel, it puts the thread doing the wait to sleep until it gets woken up by wakeup_signal, saving cycles
00:59:00kugelah, that's nice of course
00:59:24kugelbut if we do that in the lcd driver, it would be done in interrupt context
01:00
01:00:06kugelalthough we could special case that as it writes only 1px (count == 1), the wakeup thing woudnt be needed for count == 1 as the fifo doesn't fill enough
01:00:36Zagorrasher: no, not really. the slowest of 20 clients will still get the 20th hardest build in the first handout
01:01:12Zagorwe discussed trying to "spread" the builds over clients but didn't come up with a good way to do it.
01:01:28Zagorand besides we'd need a non-bogus speed measurement to have any real use for it
01:01:36kugelthe slowest clients get the hardest builds?
01:01:48Zagorkugel: eh?
01:01:51Bagderkugel: read again
01:02:03Bagder"the slowest of 20 clients will still get the 20th hardest build"
01:02:08kugelah, 20th
01:02:41kugelI was "the slowest of 20 clients will still get the 20 hardest builds"
01:02:56kugel+reading
01:02:59Bagderhaha
01:03:29rasherZagor: So with the current system, the slowest system doesn't get the 3 easiest builds?
01:03:46Bagderexactly
01:04:15Bagderto do that, we'd have to "leave" a very hard build for a while and hand out easy builds to the slower ones
01:04:27Bagderto risk that no fast client is left later
01:04:53Bagderwe instead focus on handing out builds in prio order, the heavier the earlier
01:04:57Zagorrasher: correct. they are simply last in the handout list.
01:05:19rasherBagder: True
01:05:25 Quit kushalone (Client Quit)
01:05:32 Quit soap ()
01:05:46Bagderbut there are indeed many things we can tweak in this
01:06:54 Quit midgey|w (Ping timeout: 180 seconds)
01:10:07 Join soap [50] (n=soap@rockbox/staff/soap)
01:12:19 Quit mcuelenaere ()
01:14:46kugelBagder, Zagor: going for clients requesting builds would also make making use of ccache easier right? the client could mention the build he did last in the request message
01:15:11gevaertskugel: the build the client did last is already built :)
01:15:12Zagorkugel: no, clients don't choose builds. the will just say "I'm ready for another one"
01:15:41Zagorwe discussed prioritizing clients that did previous versions of each target, but that gets very messy very fast
01:15:56Bagderyeah, we get too many variables to take into account
01:16:59*gevaerts thinks that the best allocation strategy is to give out builds that require uploads before builds that don't, and then just big builds first
01:17:06kugelZagor: I didn't mean to make the client choose, but rather put a recommendation, as in telling the server "This is what I did last, it would be nice if you can give it me again due ccache, if not I'll silently accept any other build too"
01:17:20Bagdergevaerts: yes, but that's just a matter of "scoring" those builds higher
01:17:30Zagorgevaerts: yeah that was the plan for this last build, but it bugged. heavy zip builds should be first next round.
01:17:49 Join MrDuck [0] (n=kachna@r4ax178.net.upc.cz)
01:17:52kugelbut it might get complicated indeed
01:18:29*preglow_ applaudes the new build system work
01:18:34 Nick preglow_ is now known as preglow (i=thomj@tvilling2.pvv.ntnu.no)
01:18:34Bagderif all clients stay equally fast and they are the same set between builds, they will most likely get the same builds
01:20:00kugelheh indeed, getting the same build automagically seems rather likely
01:20:31rasherConsidering the amount of variables involved, I doubt it
01:21:06kugelhm, wait, they get the same starting build. but the build last is most likely not the starting build
01:21:31Zagorkugel: you see how it's getting messy?
01:21:33Bagderwell, the last _set_ would be the same
01:21:41kugelZagor: I do, yes
01:22:01Bagderbut yeah, towards the end of a round the order is not likely to be the same between rounds
01:23:00kugelwe could also have a custom ccache that only ccaches certain builds based on bogomips :D
01:26:26 Quit bmbl ("Bye!")
01:26:27CIA-71New commit by zagor (r21737): Added server connection check.
01:27:58 Quit kugel (Remote closed the connection)
01:29:13*shotofadds wonders if preglow has seen fs#10415, before running off to bed
01:29:41shotofaddsnight all
01:29:43 Quit shotofadds ("Leaving")
01:31:53 Quit Syrius ("/connect irc.freenode.org /join ##survival")
01:33:04 Quit gevaerts (Nick collision from services.)
01:33:13 Join gevaerts [0] (n=fg@rockbox/developer/gevaerts)
01:33:27preglowshotofadds, i have indeed, but i'm a bit here and there during the summer, and my d2 isn't with me, currently :/
01:33:43preglowthought the prospect of it alone of course has me very happy indeed :P
01:34:24 Quit Zagor ("Clint excited")
01:35:30Unhelpfulok, i had to use gcc-4.0.4, but i have an arm-elf-eabi toolchain. it doesn't really seem to be making a huge difference.
01:40:08 Quit dmb (Read error: 104 (Connection reset by peer))
01:40:27 Join dmb [0] (n=Dmb@unaffiliated/dmb)
01:41:49 Quit tessarakt ("Client exiting")
01:47:04 Quit Thundercloud (Remote closed the connection)
01:54:46 Part toffe82
01:59:04 Quit JdGordon| ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
01:59:54 Part wincent ("Kopete 0.12.7 : http://kopete.kde.org")
02:00
02:18:30 Quit r0b- (Read error: 110 (Connection timed out))
02:19:18 Join saratoga [0] (i=9803c6dd@gateway/web/freenode/x-7a55fb6249a2b3cf)
02:19:50saratogaBagder: my build client ended up in a loop saying this today: http://pastebin.com/m4bbfb404
02:19:54saratoganot sure if it matters
02:20:56 Quit efyx_ (Remote closed the connection)
02:38:07 Nick fxb is now known as fxb__ (n=felixbru@h1252615.stratoserver.net)
02:55:17***Saving seen data "./dancer.seen"
03:00
03:12:49 Quit saratoga ("Page closed")
03:14:12 Join fdinel [0] (n=Miranda@modemcable204.232-203-24.mc.videotron.ca)
03:34:09 Join toffe82 [0] (n=chatzill@adsl-99-130-7-236.dsl.frs2ca.sbcglobal.net)
03:40:42 Join ReKleSS [0] (n=ReKleSS@d58-110-112-4.meb3.vic.optusnet.com.au)
03:45:24 Quit HellDragon (Read error: 60 (Operation timed out))
03:55:35 Join mc2739 [0] (n=mc2739@cpe-67-10-234-29.satx.res.rr.com)
04:00
04:02:36 Join HellDragon [0] (i=jd@modemcable178.248-201-24.mc.videotron.ca)
04:03:34 Part n00b81 ("Leaving")
04:20:35 Join z35 [0] (n=z35@ool-45701ce5.dyn.optonline.net)
04:26:00UnhelpfulSUCCESS! gcc 4.0.4 arm-elf-eabi builds e200 rockbox without -mlong-calls, cutting 61KB from RAM usage, 56KB from binary size.
04:28:09Unhelpfulalso, i think it might work without bumping gcc, with one caveat... libgcc build fails for arm-elf-eabi with -mcpu=arm9e
04:28:56 Quit courtc (Read error: 60 (Operation timed out))
04:29:20 Quit krazykit (Read error: 60 (Operation timed out))
04:29:53 Join courtc [0] (n=court@unaffiliated/courtc)
04:32:24 Join krazykit [0] (n=kkit@c-24-218-166-241.hsd1.ma.comcast.net)
04:34:51mc2739Bagder: Re: New builds for the build table? −− Sansa e200v2 - Sim
04:39:57mc2739Zagor: Server refused connection: error duplicate name!
04:49:36 Quit dmb (Read error: 113 (No route to host))
04:55:19***Saving seen data "./dancer.seen"
05:00
05:08:01 Join dmb [0] (n=Dmb@unaffiliated/dmb)
05:14:29 Quit luke_dozen (Read error: 110 (Connection timed out))
05:15:13 Join luke_dozen [0] (n=luke_doz@p54AB5FAA.dip.t-dialin.net)
05:26:04 Nick notByan is now known as Byan (n=notByan@logo.csl.mtu.edu)
05:32:28Unhelpfuldigging a bit further... gcc issues ldrd instructions when building __muldc3 with -mcpu=arm9e, but gas says that ldrd is not valid for this processor. as far as i can tell all of the targets that use -mcpu=arm9e could switch to -mcpu=arm946e-s, which appears to be the actual ARM core in the SoC on these devices.
05:44:06 Quit fdinel ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
05:48:38 Quit Horscht ("Verlassend")
05:54:25Unhelpfulhrm... is there any reason to use cmps on arm instead of cmp? i'm getting a warning about the s modifier being deprecated for cmp.
06:00
06:02:08 Quit luke_dozen (Read error: 110 (Connection timed out))
06:08:09*JdGordon prepares a pretty contentious commit :)
06:08:47LloreanWhich one?
06:09:27JdGordonclip keymap changes.... dont worry.. I'll put it on FS first though
06:09:55JdGordonI was thinking of being an ass and implementing 100% customisable keymaps this evening though :)
06:10:15LloreanYou live on the same continent as me now, remember that. :-P
06:11:28JdGordonI can take you on! :D
06:11:47scorche|shJdGordon: i have seen you =P
06:11:57JdGordonI can take you also!
06:12:01scorche|shmmmhmmm
06:12:09JdGordonI've bulked up since you last saw me :p
06:12:16JdGordon... not really :'(
06:13:16JdGordonWTF? home+left is pageup in the list?
06:13:50 Join dash32 [0] (n=dash32@p54AB4D0D.dip.t-dialin.net)
06:25:04*JdGordon has managed to make it worse!
06:32:40JdGordonLlorean: you have a clip yeah?
06:32:55LloreanYup
06:33:33JdGordonagg, never mind, I thought the pitchscreen combo was home+up not select+up
06:33:43JdGordonI dont think home+up works
06:34:40LloreanCombos are limited?
06:34:52JdGordoni cant tell... code looks like it should work
06:35:02JdGordonkeymap isnt letting me do anything with home+up though
06:35:05JdGordonwhich sucks...
06:39:00JdGordonLlorean: can you check in home+right (pg down) in the lists work please?
06:39:55*JdGordon has a suspision
06:40:06JdGordonnot to be confused with a suspicion
06:40:40LloreanDoesn't seem to do anything
06:40:55JdGordonbut home+left works?
06:40:58LloreanIf I hold "Home" and press right more than once quickly, I actually end up back at the home menu
06:41:35JdGordonI tihnk buttons in the same "column" cant be mapped together
06:41:46LloreanI can do Home+Left if I press them at nearly the same time
06:41:52LloreanIf I hold Home, then wait, then press left, nothing happens
06:41:55LloreanSo I can't go up multiple pages
06:41:57JdGordonwhich means no home+up or home+right
06:42:17Lloreanhome+left doesn't really work either.
06:42:20Byancan't be mapped limited by what?
06:42:27JdGordonhardware
06:42:27Byanrockbox or the hardware?
06:42:28Lloreanyou have to release home, then press it again, for each press of left.
06:42:33Byanhm
06:42:50JdGordonthats a keymap issue though, not hardware.. i tihnk
06:43:08LloreanOther combos don't normally work like that, but okay
06:43:22LloreanI guess it's Home+Left rather than Home(Repeat) + Left
06:44:13JdGordonBUGGER... this makes the keymap much more painful
06:44:32JdGordondidnt want to use select as the main modifier because its all too cramped
06:45:28LloreanYeah, someone should write a letter to sandisk about their inconsiderate hardware design for added features
06:46:19*JdGordon wonders who might have all this time now he's not spamming the forums and mailing lists... :D
06:46:55 Join _lifeless [0] (n=lifeless@188.16.122.181)
06:54:13 Quit dash32 (Read error: 104 (Connection reset by peer))
06:55:21***Saving seen data "./dancer.seen"
06:56:05JdGordondown in the wps is id3 and pitch screens... which should be quick and which should be repeat?
06:57:11JdGordon^ that is why we need partially customisable buttons...
06:57:48 Join r0b- [0] (n=rob@adsl-76-235-168-174.dsl.klmzmi.sbcglobal.net)
06:59:46JdGordonyeah, I reckon this is pretty useable now...
07:00
07:00:25ReKleSSroughly how long should it take to reflash the H120 bootloader?
07:00:37ReKleSSI'm guessing <0.5s is means something went wrong
07:01:15 Join __lifeless [0] (n=lifeless@188.16.122.181)
07:01:19 Quit __lifeless (Read error: 104 (Connection reset by peer))
07:02:24 Join dash32 [0] (n=dash32@p54AB4D0D.dip.t-dialin.net)
07:03:29 Quit dash32 (Remote closed the connection)
07:05:15 Join dash32 [0] (n=dash32@p54AB4D0D.dip.t-dialin.net)
07:06:55 Join kugel [0] (n=kugel@rockbox/developer/kugel)
07:08:36JdGordonFS #10421 is my clip keymap changes... hopeing its not too contentious...
07:10:13kugelJdGordon: maybe r21253 has something todo with the failiure of some combos
07:11:01*kugel loves the fuze keymap, btw, sad thing it can't be that way on the clip
07:11:53kugelUnhelpful: can you share the binaries?
07:12:43kugelUnhelpful: So rockbox just works with eabo, or did you need to make some modifications to it or the compiler or special options?
07:13:57kugeleabi*
07:14:43Unhelpfulkugel: i'm loading it on my e200 shortly. :)
07:15:04Unhelpfuli've not cleared *all* the new warnings yet, either, only the ones in core. :/
07:15:52kugelcould you also make a samsa build, to see if there's other binsize changes (the amses don't have long calls)?
07:16:32kugelif you build for fuze, you could send me the binary also
07:16:44kugelor the zip rather
07:17:59 Quit _lifeless (Read error: 113 (No route to host))
07:22:13Unhelpfulkugel: when it doesn't crash on mp3 playback i'll let you know ;)
07:25:38Unhelpfulcore jpeg appears to be a bit broken :)
07:28:03 Quit kugel (Remote closed the connection)
07:30:26Unhelpfulmost of the core stuff should probably have been fixed anyway... it was all things using int in places of enum themable_icons
07:31:58 Quit dash32 (Remote closed the connection)
07:33:40 Quit amiconn (Nick collision from services.)
07:33:42 Join amiconn_ [50] (n=jens@rockbox/developer/amiconn)
07:33:46 Join pixelma_ [50] (n=pixelma@rockbox/staff/pixelma)
07:33:46 Quit pixelma (Nick collision from services.)
07:33:59Unhelpfulhere's the diff: http://pastie.org/541009
07:34:01 Nick amiconn_ is now known as amiconn (n=jens@rockbox/developer/amiconn)
07:34:03 Nick pixelma_ is now known as pixelma (n=pixelma@rockbox/staff/pixelma)
07:34:52Unhelpfulkeep in mind that the gcc patch, which is downloaded, also needs to be edited... replace the arm9e with either arm9 or arm946e-s for libgcc to build.
07:36:12 Join stoffel [0] (n=quassel@p57B4C624.dip.t-dialin.net)
07:37:32Unhelpfuli should double-check that both the snapshot version of binutils and the eabi arch setting are required. if the binutils alone could do it, it would save us a bit of trouble, i think.
07:59:34amiconnUnhelpful: Congrats :) Those rockbox code fixes should probably be committed independently of the actual eabi switch
08:00
08:00:39amiconnDoes the build work properly apart from core jep? What kind of breakage do you see there?
08:00:42Unhelpfulamiconn: it crashes on mp3 playback, at an address about .icode. vorbis playback starts, but there's no progress. and the cover art was a disaster. :/
08:00:56Unhelpfuls/about/above/
08:03:10Unhelpfuli wouldn't be surprised if there are some code generation oddities... i've no idea at all why it's emitting instructions while building libgcc for arm9e that it then claims are not valid. :/
08:03:38 Part toffe82
08:04:31amiconnWhat gcc version did you use?
08:04:49amiconnPerhaps binutils 2.19 should be used in conjunction with a newer gcc as well
08:05:20amiconnIirc the debian wiki mentions gcc 4.1.1 as the first with "usable"eabi support
08:05:27Unhelpfulit would be nice if we could just have a linker option to generate the long-call "veneers"... they seem quite sensible really, it sets up just as if for a short call to the function, using the veneer address, and then the veneer address has a pc-relative ldr to pc, and the target address of the real function.
08:06:43Unhelpfulis there any specific claim made as to which gcc versions are compatible with which binutils?
08:07:51amiconnI don't know. I do know though that some combinations definitely won't work, at least for coldfire
08:09:03Unhelpfulare either of the components of these "bad" combinations known to work in combination with other versions?
08:09:34 Join daurn [0] (n=daurnima@unaffiliated/daurnimator)
08:16:25 Quit FlynDice (Remote closed the connection)
08:18:10Unhelpfulwith regard to the deprecated cmps... changing those to cmp should be fine, right?
08:18:14 Quit MrDuck (Read error: 113 (No route to host))
08:19:10 Join FlynDice [0] (n=FlynDice@c-24-19-225-90.hsd1.wa.comcast.net)
08:28:18amiconnUnhelpful: The coldfire incompatibility is caused by a change in cpu naming. Binutils >= 2.17 will only work with gcc >= 4.1 and vice versa, at least for some cpus of the m68k/coldfire family
08:30:19Unhelpfulbinutils-2.19.51 + gcc-4.1.2 still gives me the same errors building libgcc. :/
08:30:42*amiconn would try 4.3.x or 4.4.x
08:31:22Unhelpfulconveniently i made sure i had archives for those while i was at a decent connection. :)
08:31:36Unhelpfulbut yes, if i'm going to use a binutils on the bleeding edge...
08:32:14amiconnI wouldn't use bleeding edge binutils, but rather release 2.19 or 2.19.1
08:32:41amiconnThe s in 'cmps' is purely redundant, so these should be changed to 'cmp'
08:33:17amiconnSame for 'cmns', 'tsts' and 'teqs', should we have such
08:33:29amiconnhttp://www.sourceware.org/ml/binutils/2006-05/msg00075.html
08:35:10Unhelpfuli had the relocation errors until i used 2.19.51...
08:37:22 Join einhirn [0] (n=Miranda@p5DCC1963.dip0.t-ipconnect.de)
08:38:01 Quit safetydan ("Leaving.")
08:42:49 Join Grahack [0] (n=chri@stc92-1-82-227-106-100.fbx.proxad.net)
08:43:14amiconnhmmm
08:53:00 Join robin0800 [0] (n=robin080@cpc3-brig8-0-0-cust436.brig.cable.ntl.com)
08:53:42Unhelpfuli'll try 2.19.1 with 4.4.0. if i get relocation errors i'll have to try the snapshot again. i found mailing list messages regarding newly-committed work to generate veneers in some previously-unsupported cases, so i'm guessing that 2.19.1 won't work... but we'll see.
08:54:09 Join Rob2222 [0] (n=Miranda@p4FDCC312.dip.t-dialin.net)
08:55:20amiconnHmm, there's only 4.4.0 in the 4.4.x line now? In that case I'd probably try 4.3.x (x == maximum available)
08:55:22***Saving seen data "./dancer.seen"
08:55:54*amiconn trusts x.y.0 versions of gcc even less than gcc in general :\\
08:56:07 Quit einhirn (Read error: 104 (Connection reset by peer))
08:57:36*amiconn expects at least some problems in rockbox code to show up on newer gcc
08:58:06amiconnProbably not many in generic code which is also built for the sims, more in target specific code
09:00
09:03:45Unhelpfuli'd expect generic code to have been shaken out pretty well, given how fresh the gcc shipping with many linux distributions tends to be.
09:09:27 Join MrDuck [0] (n=kachna@r3g248.net.upc.cz)
09:10:05 Quit robin0800 ("Leaving")
09:10:08 Join petur [50] (n=petur@rockbox/developer/petur)
09:10:24 Join robin0800 [0] (n=robin080@cpc3-brig8-0-0-cust436.brig.cable.ntl.com)
09:16:14 Quit robin0800 ("Leaving")
09:17:37 Join robin0800 [0] (n=robin080@cpc3-brig8-0-0-cust436.brig.cable.ntl.com)
09:18:35 Join flydutch [0] (n=flydutch@host87-202-dynamic.15-87-r.retail.telecomitalia.it)
09:18:43 Quit robin0800 (Client Quit)
09:18:45 Quit Rob2223 (Read error: 110 (Connection timed out))
09:19:03 Join robin0800 [0] (n=robin080@cpc3-brig8-0-0-cust436.brig.cable.ntl.com)
09:24:52Unhelpfulamiconn: oh! i suspect i know the cause of the jpeg trouble. possibly the other things, as well. EABI specifies different structure packing. i bet the size of struct uint8_rgb changed... it's {uchar;uchar;uchar} so it could very well be unpadded now.
09:24:53 Join Thundercloud [0] (i=thunderc@persistence.flat.devzero.co.uk)
09:25:56Unhelpfulhrm: /home/chshrcat/rockbox-crossdev-test/arm-elf-eabi/lib/gcc/arm-elf-eabi/4.4.0/../../../../arm-elf-eabi/bin/ld: error: no memory region specified for loadable section `.ARM.extab'
09:26:31Unhelpfulthat could actually account for crashes in pretty much anything that mixes C and asm if structures are involved, too, i suppose?
09:27:02amiconnIt depends on what assumptions the asm makes about struct alignment
09:27:26amiconnRegarding arm.extab - that means the linker script(s) need adjustment
09:28:23Unhelpfulwe need to add it somewhere?
09:29:12amiconnMaybe we can just drop it
09:29:53Unhelpfulhow would i go about doing that?
09:30:05 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
09:30:22amiconn".ARM.extab names a section that contains exception unwinding information."
09:30:28Unhelpfulactually, i think i see how :)
09:30:33amiconnI don't know whether we need this, but probably we don't
09:31:49Unhelpfulhrm, getting rid of .ARM.exidx makes link angry... unresolved symbols from libgcc now. :/
09:37:23*Unhelpful scratches his head @ /home/chshrcat/rockbox-crossdev-test/arm-elf-eabi/lib/gcc/arm-elf-eabi/4.4.0/../../../../arm-elf-eabi/bin/ld: .ARM has both ordered [`.ARM.exidx' in /home/chshrcat/rockbox-crossdev-test/arm-elf-eabi/lib/gcc/arm-elf-eabi/4.4.0/arm7tdmi/libgcc.a(bpabi.o)] and unordered [`.ARM.extab' in /home/chshrcat/rockbox-crossdev-test/arm-elf-eabi/lib/gcc/arm-elf-eabi/4.4.0/arm7tdmi/libgcc.a(_udivdi3.o)] sections
09:39:58 Quit robin0800 ("Leaving")
09:40:00Unhelpfuloh, and reloc errors again. i think perhaps i'd best try binutils-2.19.51 again, maybe with gcc-4.3
09:40:15 Join robin0800 [0] (n=robin080@cpc3-brig8-0-0-cust436.brig.cable.ntl.com)
09:43:17ReKleSSYAAAY I unbricked my iriver
09:43:27ReKleSSabout 4 full days of work
09:44:03ReKleSS...bugger, with the BDM interface soldered on I can't put the CF card in
09:48:08 Quit Thundercloud (Remote closed the connection)
09:52:58 Quit kadoban (Remote closed the connection)
09:54:51markunReKleSS: congrats man
09:58:33 Join kadoban [0] (n=mud@cpe-24-93-17-195.rochester.res.rr.com)
10:00
10:03:18 Quit scorche (Read error: 104 (Connection reset by peer))
10:04:22 Join scorche [50] (n=scorche@rockbox/administrator/scorche)
10:04:37 Join DarkDefender [0] (n=rob@78-69-30-229-no36.tbcn.telia.com)
10:25:06tmztReKleSS: BDM?
10:25:44scorchetmzt: http://en.wikipedia.org/wiki/Background_Debug_Mode_interface
10:26:12tmztthanks
10:26:19tmztsomething like jtag?
10:26:37scorchesure
10:27:23 Join J-23_ [0] (n=zelazko@unix.net.pl)
10:28:42 Nick J-23_ is now known as J-23 (n=zelazko@unix.net.pl)
10:35:45 Join mc2739_ [0] (n=mc2739@cpe-67-10-234-29.satx.res.rr.com)
10:36:30 Quit J-23 ("wszed³em")
10:38:39 Join J-23 [0] (n=zelazko@unix.net.pl)
10:43:20 Join efyx_ [0] (n=efyx@lap34-1-82-224-140-171.fbx.proxad.net)
10:44:17 Nick fxb__ is now known as fxb (n=felixbru@h1252615.stratoserver.net)
10:44:17 Quit trisiak (Read error: 110 (Connection timed out))
10:50:20 Quit mc2739 (Read error: 110 (Connection timed out))
10:52:34 Quit stoffel (Remote closed the connection)
10:53:32 Join mc2739 [0] (n=mc2739@cpe-67-10-234-29.satx.res.rr.com)
10:55:26***Saving seen data "./dancer.seen"
10:56:18 Part Grahack
11:00
11:06:09 Join kugel [0] (n=kugel@rockbox/developer/kugel)
11:08:05 Quit mc2739_ (Read error: 110 (Connection timed out))
11:10:19 Join mcuelenaere [0] (n=mcuelena@78-21-191-122.access.telenet.be)
11:25:44 Nick fxb is now known as fxb__ (n=felixbru@h1252615.stratoserver.net)
11:31:35 Join mt [0] (n=MTee@41.233.150.185)
11:39:31 Quit kugel (Read error: 110 (Connection timed out))
11:52:56UnhelpfulTorne: no luck with that configure option, nor with adding -fno-<anything to do with exceptions> while compiling RB.
11:53:12Tornesend me your libgcc.a and i'll see what's different
11:56:01Tornealso, which input objects have an exidx/extab section anyway?
12:00
12:02:10Unhelpfulis there an easy way to list just the sections?
12:03:34tmztobjdump should be able to
12:03:43tmzt-d or -x I think
12:06:54Unhelpfulalmost all of rockbox has them. :/
12:10:50 Join einhirn [0] (n=Miranda@bsod.rz.tu-clausthal.de)
12:14:11Unhelpful+ lines are object files with, - are ones without: http://pastie.org/541170
12:19:05TorneUnhelpful: that looks like "all the objects containing C functions" to me
12:19:09Unhelpfulpretty much all of those files reference __aeabi_unwind_cpp_pr0
12:19:20Unhelpfulpretty much, yes. :)
12:19:31Tornewhich is what you expect from compiler-generated unwindings :)
12:20:06Tornei am totally puzzled on why it's generating them at all
12:22:29Tornewith my toolchain, a trivial C file gets an exception table if compiled with g++ but not if compiled with gcc
12:23:08Tornewhich is what you'd expect: g++ has to generate one for every function, unless it can prove it's a leaf which doesn't throw (or only calls functions which ultimately call leaves which don't throw, in the same object file)
12:25:02 Join n1s [0] (n=n1s@rockbox/developer/n1s)
12:26:39TorneUnhelpful: what happens if you compile something entirely trivial and unrelated to rockbox, like http://pastie.org/541177 with that gcc?
12:26:48Torneis there exception info in the object produced?
12:27:32 Join kugel [0] (n=kugel@rockbox/developer/kugel)
12:29:34 Quit J-23 ("wszed³em")
12:30:48Unhelpfulno... all normal stuff.
12:31:19Torneright... so what about args/etc?
12:31:51Tornethere must be *something* making it decide to generate extab for every bloody thing :)
12:33:32 Join trisiak [0] (n=tree@chello089078243195.chello.pl)
12:38:39Unhelpfulwell, for one, i note that when i tried to kill exceptions, i used -funwind-tables rather than -fno-unwind-tables
12:38:53Unhelpful...not that fixing that fixed rockbox
12:38:59 Join J-23 [0] (n=zelazko@unix.net.pl)
12:39:27TorneUnhelpful: gcc should always have those things off by default anyway
12:39:30Unhelpfuladding our GCCOPTS doesn't create them...
12:39:40Torneso it must be something in our code, then :)
12:39:54Tornetake a nice small file from rockbox that produces exception tables and #if 0 loads of it out
12:39:58Tornetry and narrow it down :)
12:40:31Unhelpfuli doubt we have any "small" files if you consider includes...
12:40:43Torneif 0 those out as well
12:40:56Tornehack and slash job until some subset happens to compile :)
12:42:00Torneif you send me what you changed in rockboxdev.sh i'll have a go later on
12:42:52 Join atb [0] (n=atb@69.90.52.123)
12:48:20Unhelpfulwait a sec, removing that errant -funwind-tables killed them from the files... but i still get the link errors related to get_eit_entry and friends.
12:48:40Torneright, that's a more directly diagnosable problem
12:48:59Torneas long as gcc isn't emitting any into things it compiles, then it's just the runtime libraries trying to be C++ friendly
12:49:48Torneobjdump -x libgcc.a into pastebin? :)
12:49:51Unhelpfulin fact, none of our objects contain any now, so what gives?
12:50:17Tornean object in libgcc probably does, or refers to the unwinding functions directly
12:51:34Unhelpfulit'd have to be an object that *we* refer to, though, wouldn't it?
12:51:42Torneindirectly, yes
12:51:58Tornesomething referred to by something referred to by something we refer to :)
12:52:15Tornejust paste me the whole of libgcc and i can probably spot it ;)
12:53:07Unhelpfulweird, i can't paste into pastebin...
12:55:09Unhelpfulugh. half a MB. this might not make it up over my cell :)
12:55:17Torneeep
12:55:21Tornealso, apt-get install pastebinit :)
12:55:30***Saving seen data "./dancer.seen"
12:55:45 Quit flydutch (Read error: 104 (Connection reset by peer))
12:55:55Unhelpful"We did say the maximum allowable file size is 150,000 bytes. You sent us a file that is too big. Sorry, it is being ignored." :)
12:56:05Torneaw
12:56:24Tornei guess otherwise you could transfer big binaries with base64 :)
12:57:16Unhelpfulconveniently it's highly compressible: http://filebin.ca/equvwp/libgcc.dump.bz2
12:57:28Unhelpfulactually, seeing as filebin accepts anything <50MB...
12:57:33Tornewell if you were just going to send a binary you could upload the library
12:57:39Tornebut this is fine :)
12:58:35 Join flydutch [0] (n=flydutch@host87-202-dynamic.15-87-r.retail.telecomitalia.it)
12:59:58Unhelpfulseeing as you insist: http://filebin.ca/qoxowz/libgcc.a
13:00
13:00:27Torneinteresting
13:01:00Torneit looks suspiciously like my pykern binary just doesn't happen to include any of the objects that have exception info
13:01:07Tornewhich seems odd given that that includes "division"
13:02:09CIA-71New commit by zagor (r21738): Fixed build score sort bug. ...
13:02:38TorneUnhelpful: did you try explicitly discarding the exidx and extab sections?
13:02:48Unhelpfulyes.
13:03:05Tornehm, did it definately get their relocs as well?
13:03:52Unhelpfuli've no idea how to make sure?
13:04:41Unhelpfuli just added a *(.ARM.extab) to our existing /DISCARD/
13:04:53Tornealso exidx?
13:04:57Torneyou might need a wildcard on the end
13:05:12 Quit mc2739 ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060215]")
13:05:22Torne*(.ARM.extab) *(.ARM.extab.*) *(.ARM.exidx) *(.ARM.exidx.*)
13:05:37*Torne experiments building stuff with division
13:06:00Unhelpfulyup, just tried that.
13:07:34*Torne shakes fist at gcc for optimising away his divisions :)
13:08:11 Join dfkt [0] (i=dfkt@unaffiliated/dfkt)
13:11:26n1sisn't there -O0 for that?
13:11:49Torneswitching the operands is quicker. :)
13:11:51 Join stoffel [0] (n=quassel@p57B4C624.dip.t-dialin.net)
13:11:58Torneanyway. i can do integer division without bringing in exception info either
13:12:24Tornea partial link against libgcc gives me implementations of __aeabi_idiv and also __aeabi_idiv0 which don't have exception tables.
13:15:46Torneyeah, it's _divdi3 that's bringing it in for you
13:16:27 Quit dmb (Read error: 104 (Connection reset by peer))
13:17:00Tornethere must be some way to tell it to throw them away :(
13:22:47TorneUnhelpful: which linker script did you add the discards in? :)
13:25:14Tornediscarding the sections really should work.
13:25:44Tornethe linux kernel ld script does it, for exactly the ame reasons
13:27:38 Join mc2739 [0] (n=mc2739@cpe-67-10-234-29.satx.res.rr.com)
13:27:52 Quit stoffel (Remote closed the connection)
13:28:10 Quit mc2739 (Client Quit)
13:28:33 Join stoffel [0] (n=quassel@p57B4C624.dip.t-dialin.net)
13:30:30CIA-71New commit by mcuelenaere (r21739): Lua: ...
13:31:31kugel2.5min per toolchain
13:31:39kugeloops
13:36:59mcuelenaeresome config-*.h seem to say that HAVE_LCD_SLEEP_SETTING is optional, but if I don't define it I get an undeclared LCD_SLEEP_TIMEOUT error .. ?
13:37:50mcuelenaereah nvm, I found HAVE_LCD_SLEEP_SETTING
13:39:48kugelmcuelenaere: if you don't define _SETTING you need to define the hardcoded timeout
13:40:18mcuelenaerekugel: yes, I just figured that out :) However, it sounds more logical to me that if you don't define _SLEEP_TIMEOUT, it uses the setting
13:40:56kugelmcuelenaere: if your display isn't readable without backlight (and if waking up doesn't take too long), you could also ignore _SLEEP and put all power saving into lcd_enable()
13:41:39mcuelenaerekugel: ah, that's how it's currently done; however I don't see a call to lcd_enable() when the backlight turns off ?
13:41:47mcuelenaere(I added some logf's)
13:43:17kugelmcuelenaere: that call should be in _backlight_off() of your lcd driver
13:43:40mcuelenaerekugel: ah; and don't you mean the backlight driver?
13:43:46kugelI haven't made it target independant like lcd_sleep() yet because I recall the F/X is still doing it wrong somehow and I don't want to break it
13:44:06mcuelenaereok
13:44:43kugelbut the difference between lcd_enable() and lcd_sleep with _TIMEOUT == 0 is neglible
13:48:43UnhelpfulTorne: firmware/target/arm/sansa/app.lds. pretty sure it's the right one as make bin re-preprocesses whenever i modify it. but my system with arm-elf-eabi is packed up for the day
13:49:55TorneUnhelpful: yah. i might have a go late r:)
13:50:59Unhelpfulyou need a snapshot binutils... as far as i can find those aren't on gnu.org? i posted a diff earlier with most of what you should need, although i'm now using a later gcc.
13:52:06Tornethat sounds like a hilarious nightmare. :)
13:53:18CIA-71New commit by kugel (r21740): Fix a few comments in gwps.c.
13:53:45Unhelpfulwell, this exception nightmare started when i built a later gcc. if you use an arm-elf-eabi gcc with binutils <=2.19.1 you still get the reloc errors because it apparently *doesn't* generate stubs for most function calls
13:55:15Tornefun
13:55:43Tornetbh there may be some intrinsic value in moving to eabi anyway, as long as it proves not to make anything slower/bigger ;)
13:56:36Unhelpfulit might be worth padding uint8_rgb to 4 bytes... pixel-fits-in-a-register could come in handy. :)
13:57:15Torneindeed, someone will want to look at all the things whose size have changed and see whether it's really a win or not
13:59:00Unhelpfuland somebody *must* look at all of the things that pass C structs to asm functions. i'm pretty sure that's why core jpeg was hosed, and i suspect that's how the mp3 codec ended up at an address somewhere past it's .icode section.
13:59:29Tornefun
14:00
14:12:24kugelUnhelpful: It would also safe load/store instructions, wouldn't it? 1 ldr vs 3 ldrb
14:12:28kugelsave*
14:17:29 Part linuxstb ("Leaving")
14:18:19Unhelpfulkugel: unless you need to inspect the component values, yes. everything that uses uint8_rgb to my knowledge *does* unpack the components.
14:19:03 Join GreatBeaver [0] (n=chatzill@c-71-59-18-236.hsd1.ga.comcast.net)
14:19:06GreatBeaverhi
14:19:21GreatBeaveri have been searching far and wide for an alternative to using h120 as a transport
14:19:51GreatBeaverdoes anyone know if there is a 500gb external multimedia harddrive that can replace the h120 as transport?
14:20:02GreatBeaverwith coax or usb or optical
14:22:32CIA-71New commit by alle (r21741): Fix a typo
14:27:37 Join LambdaCalculus37 [0] (i=44a0430d@rockbox/staff/LambdaCalculus37)
14:33:14kugelHow did I miss that one !?
14:35:20 Join mc2739 [0] (n=mc2739@cpe-67-10-234-29.satx.res.rr.com)
14:36:03mc2739kugel: another typo on the same line −− surpress should be suppress
14:36:31kugelhm
14:36:51kugelI recall looking in a dict for that word when i wrote it, don't know what happened
14:39:49mc2739Zagor: Server connection stalled. Exiting! −− this happens just before I get the duplicate name error.
14:40:40 Quit n1s (Read error: 110 (Connection timed out))
14:41:29 Quit mc2739 ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060215]")
14:41:45GreatBeaverdoes anyone know of a harddrive that can be used as a transport for audio?
14:42:01LambdaCalculus37Any Rockbox DAP?
14:42:16GreatBeaveri am using an h120 optical out
14:42:20GreatBeaverbut i want more space
14:42:29LambdaCalculus37You can upgrade the hard drive in the H120 easily.
14:42:40GreatBeaveri am waiting for amazon to have the 80gb mk8025gal in stock
14:42:56GreatBeaverbut im researching 500gb-1tb harddrive with transport capability
14:43:17GreatBeaveris there anything like it without the crazy price of music servers?
14:43:27LambdaCalculus37Sorry, but no 1.8" hard drive has even reached that capacity yet.
14:43:53GreatBeaverjust a few months ago this came out
14:43:55GreatBeaverhttp://sdd.toshiba.com/main.aspx?Path=StorageSolutions/1.8-inchHardDiskDrives/MK8025GAL/MK8025GALSpecifications
14:44:44GreatBeaveri have the adapter but the mk8025gal is rare
14:45:06LambdaCalculus37And I'm going to also bet rather expensive.
14:45:29LambdaCalculus37You can also perform a CompactFlash mod on your H120.
14:45:48GreatBeaver64gb cf is very expensive
14:45:54GreatBeaver80gb hdd is $100
14:49:14GreatBeaveri think i might just get a laptop known for its transport quality
14:49:34GreatBeaverthe rarity of harddrive based transports and their price is crazy
14:50:06*GodEater wonders how you define "transport quality"
14:51:54CIA-71New commit by funman (r21742): Sansa AMS SD driver: fix error checking in µSD insertion ...
14:53:09GreatBeaverless jitter
14:53:20GreatBeavergood software
14:54:39GodEaterI'm still not understanding
14:55:32***Saving seen data "./dancer.seen"
14:55:34GreatBeaveran audiophile transport
14:55:44GreatBeaverwith low emi/rfi
14:56:32TorneGodEater: he means transport as in outputting an audio signal, not as in moving mp3s around
14:57:06GodEaterbut he asked for a harddrive. Harddrives don't ouput an audio signal.
14:57:13*GodEater is clearly confused
14:58:10GreatBeaveri mean a multimedia harddrive
14:58:12Tornea harddisk transport is a box with a hard drive and an audio out and enough of a processor to play audio off the drive
14:58:19Tornei.e. an mp3 player, but less portable :)
14:58:31GreatBeaverdo you know of any good ones Torne ?
14:58:34TorneNope
14:58:35GodEaternever even heard of the things
14:58:42GreatBeaverim searching everywhere and no luck im about to just search for notebooks
14:58:53TorneGodEater: they're a replacement for carting huge boxes of CDs around to DJ with, or hwatever :)
14:59:42Torneanyway, no we don't know, it seems, and this is offtopic: this channel is for rockbox development and support
14:59:53Torneany rockbox player will do what you want, but may not have enough storage: website lists them all
15:00
15:00:01Torneother than that we are unlikely to know
15:00:20GreatBeavermaybe rockbox should rockbox a multimedia harddrive ^^
15:00:35GodEaterI'm not sure what percentage of rockbox people that hang out here are audiophiles anyway
15:00:40GodEaterI'm certainly not ;)
15:00:45Torneindeed :)
15:01:50TorneGreatBeaver: anyone's welcome to try a port like that if they have the hardware for it :)
15:02:08Tornewe don't write ports because someone else wants them; we help people do it themselves :)
15:02:20GreatBeaveri have to find a harddrive with a display interface
15:02:25GreatBeavercant find any
15:03:01 Join teru [0] (n=teru@KD059133112132.ppp.dion.ne.jp)
15:03:15GreatBeaveris it possible to rockbox a computer?
15:03:25GreatBeaverlike no operating system just rockbox
15:03:29Torneno.
15:04:34 Quit bmbl (Read error: 110 (Connection timed out))
15:04:35Tornenobody has done a PC port; it's assumed that such a thing would not be particularly useful
15:05:00GodEaterwhat an utter waste of a computer it would be.
15:05:09GreatBeaverit could bypass the windows kmixer with certainty though
15:05:19GreatBeaverthat resamples music to 48khz
15:05:26Torneif you don't like windows's sound system, don't run windows
15:05:31GodEaterwindows is a dirty word here :)
15:05:38GreatBeaveryes, we should rockbox windows
15:05:45Tornethere are other operating systems with other sound systems
15:06:19GreatBeaveryes but im boycotting apple
15:06:31GodEaterthere are other OSes besides OSX too
15:06:50GodEaterthere are a plethora of them in fact
15:06:52TorneLinux? DOS? The BeOS derivatives? The BSDs? etc :)
15:06:58GodEaterQNX
15:07:00GreatBeaverbut do they start up as fast as rockbox?
15:07:13courtcQNX does.
15:07:20Tornethey start up faster than rockbox, given that rockbox doesn't start at all on PC hardware
15:07:28GodEaterhehe
15:07:31Torneand thus its startup time would have to include "port rockbox to PC"
15:07:35Tornewhich would be months :)
15:07:44courtcMonth shmonths.
15:08:02Tornecourtc: depends if you want it to work on any pc other than the author's, really :)
15:08:17Tornegetting an OS to run on *one* PC is not too difficult, getting it to run on a majority of PCs is a lot more work ;)
15:08:21GreatBeaveri think i'll just hardwire my h120 to a 1tb harddrive
15:08:40TorneGreatBeaver: if you're happy with the sound hardware on the h120 then that sounds like a perfectly sensible plan
15:08:51Torneassuming we can support that big a drive which i have no idea
15:08:51GreatBeaveri only use the optical output on the h120
15:09:14GreatBeavermaybe the external harddrive needs to be plugged to a power supply
15:09:29courtcNah, it's easy, as long as you don't care about functionality, which would be given, since you are porting rockbox to a pc.
15:09:54GreatBeaverWHY IS THIS WORLD SO CRUEL TO AUDIO PEOPLE
15:09:58Tornecourtc: there's several dozen popular sound chipsets to support, for a start :)
15:10:06Tornecourtc: without which rockbox would be *particularly* useless
15:10:11courtcWho said anything about sound?
15:10:53 Join domonoky [0] (n=Domonoky@rockbox/developer/domonoky)
15:11:23 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
15:11:24courtcfork the project, and call it NotSoRockBox.
15:13:41GreatBeaveri think i will make my own laptop and rockbox it
15:13:55GreatBeaveri only trust rockbox software
15:14:05GreatBeaveri dont trust any operating system to handle the audio data
15:15:08courtcHrm... wouldn't be much of a laptop, why don't you just buy a DAP?
15:15:09bubsysee, guys? Fanboys DO exist on our planet.
15:15:23Tornecourtc: because he wants a 1TB hard drive :)
15:15:34Torneand they don't make those in 1.8" :)
15:15:46courtcSure they do... I bet.
15:16:18courtcCommission one from toshiba, they'd be happy to oblige... for the right price.
15:16:26bubsyPrice. Exactly
15:18:34courtcJust get a DAP with wifi; make rockbox work with wifi/said DAP; precache playlists.
15:19:32rasherGreatBeaver: Not to rain on your parade, but Rockbox resamples everything to 44.1kHz
15:20:06GreatBeaver44.1khz is what i want
15:20:13GreatBeaveri dont listen to 96khz
15:20:40ReKleSSwould it be worthwhile writing up my H120 unbrick procedure?
15:21:00ReKleSSit involves soldering wires to the motherboard and wiring it to a microcontroller...
15:21:03TorneReKleSS: if it's not already on the wiki, then ys, put it there
15:21:08GreatBeavernm im just going to use 80gb hdd on the h120
15:21:13Tornewell, if you can be bothered ;)
15:21:50GreatBeaveryou guys made an awesome software for h120
15:21:58GreatBeaveri will just buy more h120's if i need more space
15:23:05 Join BdN3504 [0] (n=5ce227df@gateway/web/cgi-irc/labb.contactor.se/x-001b70fc5385bdd7)
15:23:39GreatBeaveris any dap company going to pay rockbox for their software?
15:23:48GreatBeaveri bet rockbox is 100x better than the ipod software
15:23:53GreatBeaverapple should pay you guys for it
15:26:03 Join Blue_Dude [0] (n=chatzill@adsl-235-206-197.mco.bellsouth.net)
15:26:14BdN3504is marc guay around? i just saw that the bootloader task has changed. Was the name only updated or did you also change the bootloaders? or have only some bootloaders been updated? is the sansa bootloader new? then i'll test...
15:26:25domonokyit should also be possible to make a 1,8 -> 3,5 adapter and connect a 2TB Harddrive to a rockboxed DAP.. Of course thats not very transportable :-)
15:26:46GreatBeaveroooooo thats what i want ^^
15:27:42GreatBeaveri dont think an h120 has enough juice to power on that harddrive though
15:27:49GodEaterany DAP witha 1TB drive attached is at least going to need a custom build of Rockbox with LBA48 compiled in.
15:27:51Blue_DudeGreatBeaver: I've often wondered about a Rockbox port to a multimedia appliance, but there's nothing on the drawing board that I know about. You might try looking at something like a Mac Mini and go the HTPC route.
15:28:23GreatBeaveryeah but i will have to research if people think it is good quality optical output first
15:28:24domonokyGodEater: yes, but thats only a define. no real work needed... :-)
15:28:46GreatBeaveralso i have something against apple, i dont like how they boycott flac
15:29:01Blue_DudeGreatBeaver: define "good quality".
15:29:09ReKleSShrm... should the svn bootloader be ... maybe not safe, but at least somewhat tested?
15:29:17GreatBeaverequivalent to expensive audiophile cd transports
15:29:24GodEaterlots of people don't implement FLAC in their DAP firmware.
15:29:27GodEaterit's not just Apple
15:29:43GreatBeaveryeah which means im boycotting them too :D
15:30:05 Join n00b81 [0] (n=taylor@unaffiliated/n00b81)
15:31:55Blue_DudeGreatBeaver: expensive "audiophile" CD transports are often triumphs of marketing over engineering. Unless you have an extremely discriminating system (and if you're worried about the expense of HDD capacity, you don't have one) you just can't tell the difference between an "audiophile" transport and a $49 Walmart special CD player. Digital is digital.
15:32:15 Quit n00b81 (Client Quit)
15:32:40GreatBeaverya i have a pretty expensive headphone system though
15:32:54GreatBeaver$500 dac, $1000 amp, $500 headphones
15:33:09Blue_DudeGreatBeaver: I'd be far more concerned about playable formats vs. tachnical stats that can not audibly discerned.
15:33:14Blue_Dudetech
15:33:19GreatBeaveryes formats is important to me
15:33:27LambdaCalculus37GreatBeaver: That's nice and all, but can you please keep it on topic here?
15:33:32Blue_DudeSorry.
15:34:13 Join stettberger [0] (n=stettber@peer.zerties.org)
15:34:22GreatBeavermaybe in the future i'll buy a mac mini and rockbox it
15:34:26Blue_DudeWith all that said, *can* Rockbox be ported to an appliance?
15:34:27GreatBeaveri'll let u guys know how it goes
15:34:43GodEaterGreatBeaver: how will you do that? Rockbox doesn't run natively on PC hardware of any sort.
15:36:08 Join n00b81 [0] (n=taylor@unaffiliated/n00b81)
15:36:31stettbergeris here anybody, who uses rockbox on an sansa e200v2?
15:36:58GodEaterstettberger: you're better off just asking your actual question.
15:37:55 Quit stoffel ("No Ping reply in 90 seconds.")
15:38:16 Join stoffel [0] (n=quassel@p57B4C624.dip.t-dialin.net)
15:38:22stettbergeri want to try the actual development, but with the longer thread in the forum (72 pages atm) i don't get an ideo how to do so
15:38:50stettbergeris there some condensed information?
15:39:01kugelSansaAMS wikipage
15:39:20LambdaCalculus37And the source code is also a good place to look at what's been done so far.
15:39:32kugelAnd the official testing builds forum, where we published installation tools and methods
15:40:02LambdaCalculus37And here. If you have a specific question, ask away.
15:40:23stettbergerah, ok is see thank you, i think that will be a good starting point
15:41:20 Join evilnick [0] (i=0c140464@gateway/web/freenode/x-131b8d4f6bc52f0c)
15:45:53 Quit Blue_Dude ("ChatZilla 0.9.85 [Firefox 3.0.11/2009060215]")
15:46:32 Quit Zarggg (Read error: 104 (Connection reset by peer))
15:47:04 Quit stoffel (Remote closed the connection)
15:47:31 Join stoffel [0] (n=quassel@p57B4C624.dip.t-dialin.net)
15:51:36stettbergeryeahaa, rockbox is really working on it :-) thx to all developers
15:57:15LambdaCalculus37stettberger: Even at this early stage, quite a lot works on the AMS Sansas.
15:57:29LambdaCalculus37stettberger: If you see any places where improvments can be made, feel free to chip in.
15:57:50LambdaCalculus37We always welcome volunteers to the project... that's how it works. :)
15:58:05stettbergerits written in c or?
15:58:24LambdaCalculus37stettberger: Mostly C, with some assembler for speed-specific bits.
15:58:24rasherstettberger: Almost entirely in C
15:59:03stettbergernice :-)
16:00
16:00:20LambdaCalculus37stettberger: So feel free to checkout the source from SVN and dig in.
16:02:38stettbergerthere is only one rockbox svn where the code for all models is in?
16:03:28LambdaCalculus37Yep.
16:03:40bubsyall models?
16:03:46LambdaCalculus37If you have svn installed, run svn co svn://svn.rockbox.org/rockbox/trunk rockbox
16:03:50LambdaCalculus37bubsy: All models.
16:04:03bubsyeven Creative Zen Vision:M?
16:04:05bubsyNo.
16:04:12stettbergerifdef-hell?
16:04:15LambdaCalculus37bubsy: There's code for the ZVM in the trunk.
16:04:26bubsywithout a proper HDD initalizer
16:04:29LambdaCalculus37But it's not a complete port by any means.
16:04:30rasherstettberger: There's some, but it's better than it has been
16:06:39CIA-71New commit by teru (r21743): Correct return value of function get_ucs, position of next character, in viewer plugin (FS #9387, patch by Yoshihisa Uchida, small modification by ...
16:22:38 Quit BdN3504 ("CGI:IRC (EOF)")
16:23:14 Join Zarggg [0] (n=zarggg@65.78.69.194)
16:30:53 Join BdN3504 [0] (n=5ce227df@gateway/web/cgi-irc/labb.contactor.se/x-65ad84c60b5cabb4)
16:31:19BdN3504somebody give me a quick link to the latest sansapatcher please.
16:33:08BdN3504nvm
16:33:09BdN3504http://download.rockbox.org/bootloader/sandisk-sansa/sansapatcher/
16:34:18 Quit robin0800 ("Leaving")
16:34:18 Quit BdN3504 (Client Quit)
16:34:41 Join robin0800 [0] (n=robin080@81.98.157.181)
16:36:56 Quit courtc (Read error: 104 (Connection reset by peer))
16:37:00 Join courtc [0] (n=court@unaffiliated/courtc)
16:37:25 Quit timc (Remote closed the connection)
16:37:34 Quit stoffel ("No Ping reply in 90 seconds.")
16:37:56 Join stoffel [0] (n=quassel@p57B4C624.dip.t-dialin.net)
16:38:00 Join dash32 [0] (n=dash32@p54AB4D0D.dip.t-dialin.net)
16:47:47 Quit ReKleSS (Read error: 110 (Connection timed out))
16:49:42 Join BryanJacobs [0] (n=bryanjac@128.151.67.243)
16:55:35***Saving seen data "./dancer.seen"
17:00
17:01:06 Join toffe82 [0] (n=chatzill@74.0.180.178)
17:03:35 Quit teru ("Quit")
17:05:59 Nick fxb__ is now known as fxb (n=felixbru@h1252615.stratoserver.net)
17:08:51 Quit Sajber^ (Read error: 104 (Connection reset by peer))
17:13:54 Join domonoky1 [0] (n=Domonoky@g229247165.adsl.alicedsl.de)
17:15:39 Quit mt (Read error: 104 (Connection reset by peer))
17:16:28obokugel: thanks for your pointers yesterday. I've added those into the lcd_update routine, and fixed a missing udelay in my SPI function, but still no luck displaying anything
17:17:29kugelobo: lcd_update is mainly "flushing" the rockbox framebuffer into the gram.
17:18:01kugelI would try to that manually (just writing a red pixel or so), to see if the the adresses are correct, at least
17:18:25oboyup, it's coded at the moment to dump the entire thing over. I've tried that, no effect
17:20:19 Join mt [0] (n=MTee@41.233.150.185)
17:25:09 Quit petur ("beer time")
17:29:08 Join Horscht [0] (n=Horscht2@xbmc/user/horscht)
17:32:06CIA-71New commit by kugel (r21744): Rearrange things a bit for less #ifdefs and less duplication.
17:32:09 Quit domonoky (Read error: 113 (No route to host))
17:35:37 Quit flydutch ("/* empty */")
17:40:29kugelobo: what are you trying to display?`
17:41:09 Quit BryanJacobs ("Java user signed off")
17:41:14 Join BryanJacobs [0] (n=bryanjac@e33.cs.rochester.edu)
17:41:20kugeldoes lcd_clear_display() (followed by lcd_update()), that should get away with the sandisk logo
17:44:42CIA-71New commit by kugel (r21745): Adding last minutes comments to explain things is only cool if you close it also (aka fix yellow).
17:45:41obokugel: anything but the sansa logo would be good :) Shouldn't just writing a fixed value be even more simple? And since that's not working, I'm not sure what I'm missing...
17:45:58kugelthat's what clear_display is doing
17:46:42kugelmaybe you didn't get the window positions right yet, and you actually write outside visible area
17:47:41obobut a hardcoded value has the same effect? Plus that bypasses reading from main RAM (although I think it's mapped properly, I'm not certain)
17:48:00FlynDicekugel: Comment on line 389 needs to be closed
17:48:08kugelFlynDice: svn up
17:48:33FlynDicejust too fast for me....
17:49:18oboI'm just trying to go for the most simple, likely to work case to begin with - to try and rule out as many other potential issues as possible
17:50:06kugelobo: yes, I understand that. Have you tried to write single pixels or the whole display so far?
17:50:30obowhole display at the moment.
17:51:15 Quit dash32 ("Verlassend")
17:51:16obokugel: http://www.rockbox.org/twiki/bin/viewfile/Main/GSoCSansaView?rev=2;filename=lcd-view.c - but with a hardcoded value instead of *addr in lcd_update
17:52:22kugelobo: You doest seem correct (from what I see what other drivers are doing)
17:53:05kugelThat doesn't*
17:53:09obowhich bit?
17:53:43kugelother drivers are seting the window position before the transfer, not within
17:54:15kugel0x22 is the gram register I assume?
17:55:16oboyes, 0x22 is GRAM. 0x20 is horiz GRAM address, 0x21 is vert
17:55:22kugelI really can't tell you much as I have no idea of this driver (less that you even), but other drivers are working differently
17:55:40kugelI think you set those only once at the beginning
17:55:47kugelthe hardware will advance the internal pointer
17:56:22oboin which case, with the init routine setting them both to 0x0, they shouldn't need setting at all?
17:56:48kugelyou need to set them before writing to the GRAM of course
17:57:24oboOkay, I'll try that.
17:57:45kugelif they're really only used in the init function then they're probably not that what you're thinking they are
17:57:54oboI wish the datasheet had more details on the SPI mode...
17:58:16kugelwell, in fact, if you only use fullscreen updates, you might only need to set them once
17:58:34kugelbut you also don't set the end position, that confuses me a bit
18:00
18:00:06obodifferent registers provide the start and end positions during init
18:02:45*JdGordon points clip owners to FS #10421
18:23:22 Join mt_ [0] (n=MTee@41.233.144.236)
18:23:39 Join n1s [0] (n=n1s@rockbox/developer/n1s)
18:27:27CIA-71New commit by kugel (r21746): Correct another small typo in a comment.
18:33:57 Join JdGordon| [0] (i=ad75f7a5@gateway/web/freenode/x-b996193561a0d76b)
18:34:18 Quit robin0800 ("Leaving")
18:35:52 Quit MrDuck (Read error: 110 (Connection timed out))
18:38:04 Join barrywardell [0] (n=barrywar@86-45-11-60-dynamic.b-ras2.prp.dublin.eircom.net)
18:39:41 Quit mt (Read error: 110 (Connection timed out))
18:52:14CIA-71New commit by gevaerts (r21747): Also bump version in trunk
18:53:00 Join mrkiko [0] (n=mrkiko@host232-107-dynamic.24-79-r.retail.telecomitalia.it)
18:53:25kugelis it released?
18:53:48gevaertsnot yet, no. I just don't want to forget this...
18:55:03JdGordon|you mean you just want more build runs... :)
18:55:21CIA-71New commit by rob (r21748): Make the TCC NAND driver use the (virtual) disk activity LED.
18:55:36***Saving seen data "./dancer.seen"
19:00
19:01:04 Quit mt_ (Read error: 101 (Network is unreachable))
19:01:19 Quit barrywardell ()
19:01:27 Join pixthor [0] (n=48c1aacc@gateway/web/cgi-irc/labb.contactor.se/x-2878c4a867cd85f3)
19:02:35amiconnhmpf
19:02:56 Quit pixthor (Client Quit)
19:03:02JdGordon|well said!
19:03:03 Join pixthor [0] (n=48c1aacc@gateway/web/cgi-irc/labb.contactor.se/x-d49d50f260ed4897)
19:03:07 Join fdinel [0] (n=Miranda@modemcable204.232-203-24.mc.videotron.ca)
19:03:58 Quit thegeek (Read error: 54 (Connection reset by peer))
19:04:05amiconnThe lost connection detection in the client works, but it is often refused on reconnect due to duplicate name
19:04:17amiconn...and neither Zagor nor Bagder are around :\
19:04:46pixthorHey.
19:05:08pixthorHey everyone.
19:06:07 Quit JdGordon| ("Page closed")
19:09:14 Quit pixthor ("CGI:IRC")
19:13:39 Join mt [0] (n=MTee@41.233.136.213)
19:25:39bertrikI think firmware/drivers/i2c.c misses a mutex_init
19:26:21bertrikbut this looks like very old code, so I'm surprised nobody ran into problems with it yet
19:26:41 Join jgarvey [0] (n=jgarvey@cpe-098-026-065-013.nc.res.rr.com)
19:30:32 Join JdGordon| [0] (n=Miranda@nat/microsoft/x-2076cf0324777e70)
19:32:15GreatBeaveri purchased a 80gb 1 platter 5mm harddrive for the iriver h120 today :D
19:32:38GreatBeaveronly 1 person has said he has done it correctly, isanggon on rockbox forums
19:32:53JdGordon|re the duplicate client problem... what would happen if when that happens, the server pings the client it thinks is connected? as long as the clients dont respond to pings unless it is connected then the server would know the connection was dropped, kill that client and accept the new one...
19:33:27GreatBeaveri have a rockbox question, what is the easiest way to shuffle through all the music on my mp3 player?
19:33:38LambdaCalculus37Turn Shuffle on?
19:34:55bertrikwhat USB controller (compatible with what other target usb) was in the s5l8700 again?
19:35:18gevaertsthe tcc one
19:35:28gevaertssee usb-tcc.c
19:35:42bertrikah, maybe something to work on this weekend :P
19:35:48GreatBeaverLambdaCalculus37: will that shuffle through multiple folders?
19:36:06GreatBeaverthe way i access music is to use folder directories, not the playlist thing
19:36:09*gevaerts points GreatBeaver to our fine manual
19:36:11GreatBeaverwill i have to use playlist?
19:36:19gevaertsyou *always* use a playlist
19:36:24*LambdaCalculus37 also points GreatBeaver to the manual... the Manual Knows All
19:36:56bertrikgevaerts, we'll have to drive both the "usb function" part and the "usb phy control" part, right?
19:37:28 Quit r0b- (Read error: 110 (Connection timed out))
19:37:48 Join r0b- [0] (n=rob@adsl-76-235-168-174.dsl.klmzmi.sbcglobal.net)
19:38:04gevaertspossibly. I'm not entirely sure. From what I understand some controllers handle the phy bit for you
19:39:49GreatBeavermy mp3 player takes a few seconds to enter folder directories, is there a way to prevent it from doing that until i click an album for it to play?
19:39:50*bertrik spots rockbox code indented with 3 tabs
19:39:54bertrik*3 spaces
19:40:11AlexPGreatBeaver: Turn dircache on
19:40:20GreatBeaverok
19:40:40AlexPGreatBeaver: And for shuffling all the music, create a playlist of all the music, then turn shuffle on (or shuffle the music)
19:40:47GreatBeaverok
19:41:14GreatBeaveri keep changing music on my mp3 player so i dont like having to create playlists so much
19:41:43AlexPTurn on recursivly insert directories, then if all your music is in a folder (e.g. Music) then open the context menu on it, then select add to playlist (or add shuffled)
19:42:16GreatBeaverok
19:42:35AlexPWhenever you play anything in Rockbox you are using a playlist - select a file in a directory and it creates a playlist of that directory for you
19:42:39 Join linuxstb_ [0] (n=linuxstb@rockbox/developer/linuxstb)
19:43:10 Quit linuxstb ("Leaving")
19:43:19 Nick linuxstb_ is now known as linuxstb (n=linuxstb@rockbox/developer/linuxstb)
19:43:59bertrikThe "erase dynamic playlist?" message always scares me for a second
19:45:46gevaertsBagder: is rbclient.pl expected to do basically nothing after getting a build on arm?
19:46:06gevaertsI added the cpu model to the list of 32 bit cpus
19:46:35Bagderthe server won't care
19:46:38CIA-71New commit by bertrik (r21749): Add missing mutex_init to i2c driver
19:47:08gevaertsno, but all I get from it is a list of "Got <some-build>" lines, and nothing happens
19:47:24Bagderweird
19:47:39gevaertsEvery now and then it gets another build, but that's it
19:48:18rasherBagder: Maybe some experimentation should be done with the build order - with the current "easiest to slowest clients" order, my laptop builds for about 200 seconds, with the other order, it managed a single build of.. 350 seconds I think?
19:48:32rasherThat is, when counting builds that actually succeed
19:48:42 Join _zic [0] (n=user@91-165-236-113.rev.libertysurf.net)
19:49:17Bagderyes, the order is indeed subject for further tweaking and testing
19:49:57*gevaerts spots something
19:49:59gevaerts"and 0 cores"
19:50:22Bagderthat's probably why then
19:50:53gevaerts/proc/cpuinfo is a bit different
19:51:38gevaertsServer connection stalled. Exiting!
19:52:18rasherHm, the cpuinfo on my nslu2 has processor with a capital p
19:52:40gevaertsyes, same here. Also BogoMIPS
19:53:13*rasher gives gevaerts two is
19:54:37rasherIt could try to read /sys/system/cpu first, to get a more reliable number of cores
19:54:47 Join barrywardell [0] (n=barrywar@86-45-11-60-dynamic.b-ras2.prp.dublin.eircom.net)
19:54:50 Quit barrywardell (Remote closed the connection)
19:55:24*pixelma wonders if the "sansavista" theme on the c200 themes.rockbox.org is violating any copyrights
19:55:36 Quit linuxstb ("Leaving")
19:55:55 Join webguest91 [0] (n=1824bd5b@gateway/web/cgi-irc/labb.contactor.se/x-e7054cbf0518b93d)
19:57:08webguest91I have a Sansa e280v2 and I was wondering if I install the bootloader (and rockbox itself) will all of the patches (for disk corruption etc.) be included in the installation or will I have to put them in myself?
19:57:29 Nick fxb is now known as fxb__ (n=felixbru@h1252615.stratoserver.net)
20:00
20:00:40webguest91I have a Sansa e280v2 and I was wondering if I install the bootloader (and rockbox itself) will all of the patches (for disk corruption etc.) be included in the installation or will I have to put them in myself?
20:04:12bertrikwebguest91, what patches do you mean exactly?
20:04:19 Join CaptainKwel [0] (i=2669ecc2@gateway/web/freenode/x-c16ad4487b4d809e)
20:04:31webguest91Will the patches be included in a new installation of rockbox for the ams models?
20:04:50bertrikthe rockbox zip that you download from the rockbox page is up-to-date with the latest version in svn
20:05:33AlexPAgain, what patches?
20:05:41BagderJdGordon: the server already pings clients very frequently
20:05:53BagderI don't think doing yet another ping to fix dupe clients is going to help much
20:05:53AlexPwebguest91: If you mean patches on open flyspray tasks, then no.
20:06:24BagderI do believe there's a bug in the server that causes the dupe error a bit too often
20:06:45bertrikOh, we don't have zips for download of ams sansas I think
20:07:24JdGordon|well doing an extra ping when a dupe tries to connect hopefully would fix the issue when it cant connect stratight after being disconnected/falling off
20:07:50BagderJdGordon: yes, but it'd add a lot of weird extra code for a case that will be sorted out anyway within ~15 secs
20:08:27JdGordon|the client needs to keep trying then, which it looks like it doesnt do?
20:08:47 Join funman [0] (n=fun@rockbox/developer/funman)
20:08:48Bagderyes, but then again I believe this is a server bug
20:09:31CIA-71New commit by Ubuntuxer (r21750): FS #10418: Change menu button in Sudoku on Fuze, by Nick Tryon
20:09:56 Quit webguest91 ("CGI:IRC (Ping timeout)")
20:10:14 Join webguest01 [0] (n=1824bd5b@gateway/web/cgi-irc/labb.contactor.se/x-401e080b6cd1df21)
20:10:48 Quit BryanJacobs ("Java user signed off")
20:11:31webguest01Sorry, I was disconnected, I was refering to the patches under the section "Requirements for release"
20:12:52AlexPlink?
20:13:01webguest01Sorry, I was disconnected, I was reffering to the patches under "requirements for release" in the ams main page.
20:13:10funmankugel: i don't know in which conditions disk_mount can fail and how SYS_FS_CHANGED event will harm if it failed (in ata-sd-pp.c)
20:13:36kugelI was just saying, I haven't looked at it
20:13:50kugelmy error source was that file :)
20:14:21funmanthe card initialization is a bit different from code in AMS driver
20:14:39webguest01Oh sorry
20:14:46webguest01http://www.rockbox.org/twiki/bin/view/Main/SansaAMS
20:14:46kugelit is?
20:14:48funmanperhaps with a bit of work the sd_thread() could be moved in the common driver
20:14:51gevaertsBagder: in some cases $lastcomm stays 0 on my arm client so it gets kicked off. I've now initialised it to time(), which seems to fix things
20:15:07kugelI mean, without looking *into* the functions, the general init is mostly the same
20:15:33funmanthey are not called from the same places, the init is not done in sd_thread for PP
20:15:43AlexPwebguest01: I don't see any patches linked there
20:15:44Bagdergevaerts: ah, then commit the fix!
20:15:55AlexPwebguest01: I see old ones that have since been committed
20:16:21gevaertsBagder: I would, but I'm not sure if it's really correct
20:17:05 Quit webguest01 ("CGI:IRC")
20:18:20kugelfunman: there's no additional init, that's why
20:18:56kugelthe isr posts to the queue, which does disk_mount (retval not checked for pp)
20:19:54CIA-71New commit by gevaerts (r21751): Fixes to make the client work on sheevaplug
20:20:34CIA-71New commit by gevaerts (r21752): more fixes to make the client work on sheevaplug
20:20:54rasherBagder: I'm increasingly of the opinion that "hand out hard builds first" is in fact smarter
20:21:09rashercheck out jupiter-amiconn's performance in the last round for example
20:21:18rasheror maybe the mid-range builds first
20:21:40amiconnrasher: jupiter has the problem that it's running new and old builds in parallel
20:22:11 Join Zagor [242] (n=bjst@46.35.227.87.static.tab.siw.siwnet.net)
20:22:39 Quit n1s (Read error: 110 (Connection timed out))
20:22:50rasheramiconn: that just means that it rates as slow client, it seems like it should've been able to complete at least one "proper" build
20:22:59 Join saratoga [0] (i=9803c6dd@rockbox/developer/saratoga)
20:23:08rasher2 rounds without errors. Push to production!
20:23:21CIA-71New commit by gevaerts (r21753): One more "bogomips" match made case insensitive for arm
20:24:28*gevaerts thinks that someone more fluent in perl should review these last few commits (that should actually have been a single commit...)
20:25:30Zagorgevaerts: done :)
20:27:36gevaertsIs there a list of remaining blocker issues somewhere?
20:28:39funmankugel: see sd_init_device() , it still exists (the initialization needs to be made for every card)
20:29:25funmanit will be triggered by the call to disk_mount which will make the first I/O operations on the inserted card
20:30:38kugelfunman: that's in the sd thread then
20:31:51funmanyeah, i think moving sd_thread() out of target drivers should be easy
20:32:16*ej0rge wonders if what that guy was trying to ask was how up to date the official test build for c200v2 is
20:42:07GreatBeaverhas anyone replaced the battery in the h120?
20:42:14GreatBeaveri am looking at a guide right now and kind confused
20:42:46*rasher has, but it was years ago
20:42:49bertrikI think I did the wakup_wait/signal wrong for the s5l8700, yet it seems to work ...
20:43:12funmanbertrik: how wrong ?
20:43:20ej0rgeGreatBeaver: I've done a few, but lets move this to #rockbox-community
20:44:18bertrikfunman, I start a ADC conversion, then do wakeup_wait to wait for the interrupt. The interrupt routine does not clear the interrupt, so I'm surprised to not end up in an endless interrupt loop
20:44:56bertrikwell, the interrupt does get cleared in the interrupt controller, but not in the ADC module
20:45:30funmanperhaps there is some latency introducing a delay between 2 interrupts
20:47:39bertrikfunman, the datasheet isn't exactly clear how the interrupts should work
20:47:58 Quit mcuelenaere ()
20:48:16 Join Thundercloud [0] (i=thunderc@persistence.flat.devzero.co.uk)
20:49:09bertrikhm, the samsung s3c6400 datasheet mentions an extra interrupt clear register
20:49:17 Join thegeek [0] (n=nnscript@s243b.studby.ntnu.no)
20:50:33 Quit freqmod (Remote closed the connection)
20:50:39 Join freqmod [0] (i=quasselg@dhcp208-240.ed.ntnu.no)
20:51:00amiconnBagder: The duplicate client error doesn't sort itself out - it causes runclient.sh to exit
20:53:15*amiconn wonders about obo's driver code mix
20:54:50Zagoramiconn: that error is fixed in the new (uncommitted) server
20:55:18oboamiconn: code mix?
20:55:39***Saving seen data "./dancer.seen"
20:56:08rasheramiconn: is your runclient.sh up to date?
20:56:31amiconnlcd_write_cmd and lcd_write_data look like the equivalents in the mini g2 driver (using the pp mono lcd bridge in serial mode)
20:56:52amiconnBut you're doing bit-banged serial transfers
20:57:09amiconnBtw, what lcd controller is this? Is there a public datasheet?
20:58:48oboamiconn: AFAIK it's a ILI9320 - the datasheet is on my GSoC wikipage
20:59:04 Join tessarakt [0] (n=jens@e180072202.adsl.alicedsl.de)
21:00
21:00:39obokugel was correct when he said that the GRAM address normally auto-increments - it does on this controller (so that part of lcd_update isn't needed)
21:04:09amiconnHmm, a 320x240 colour lcd connected via SPI? Somehow I don't believe this...
21:04:55 Join MrDuck [0] (n=kachna@r4ax178.net.upc.cz)
21:06:23oboamiconn: well there is code in there for talking to it via SPI, which the lcd init routine (at least) uses...
21:06:42pixelmarasher: shouldn't runclient.sh update "itself"?
21:07:04rasherpixelma: it doesn't, I'm not convinced it needs to
21:07:08rasherit has barely changed at all
21:07:16kugelthe perl file is updated, not the .sh file
21:07:26pixelmawhen did it change last time?
21:07:40kugelI think I haven't changed it since devcon
21:08:32rasherYou probably should. Probably
21:08:57rasherThe "exit on fatal errors" was added on june 23rd
21:09:35amiconnMaximum SPI clock is 10MHz. That means transferring a full 320x240x16 takes 122ms, i.e. a maximum full-screen update frequency of ~8fps
21:09:41CIA-71New commit by rob (r21754): D2: Update the battery discharge curve to observed values, and add a crude runtime estimation (this is based on playback from SD card, other usage ...
21:09:50amiconnThe View does video, so there must be an alternative transfer method
21:10:53kugelhm
21:10:57kugelno build round?
21:11:06oboamiconn: okay, makes sense. Urgh, back to the disassembly then
21:11:07amiconnPerhaps there is some gpio to select transfer method
21:11:27pixelmathen mine should even be newer, I remember updating both when one of the both caused troubles when adding sdl to the archlist
21:11:35amiconnSPI would be a good start, because it's simple, and you could work from there
21:12:31amiconnpixelma, rasher: runclient.sh can't auto-update because it contains those client specific adjustments
21:13:06rasheramiconn: it could refer to a standard configfile (in theory)
21:13:41rasher(with rbclient -config=client.cfg or such)
21:13:50 Quit evilnick ("Page closed")
21:14:14funmanobo: do you have a way to run modified OF code ? if the lcd_init function has no effect (put a return instruction at the start) and LCD still works this is the wrong function
21:14:54amiconnMy copies of runclient.sh are from 25 June, and they do contain exit-on-fatal-error
21:15:06kugelfunman: it could be initialised again in the firmware
21:15:07 Join evilnick [0] (i=0c140464@gateway/web/freenode/x-fb52fd1466f2a915)
21:15:20*JdGordon| probalby also needs the newest runclient.sh
21:15:44obofunman: I could build & sign a modified OF bootloader I guess?
21:15:47rasheramiconn: which would make sense, since it was added on the 23rd
21:16:19funmanobo: only if you can recover from a bad bootloader ?
21:16:35 Join n1s [0] (n=n1s@rockbox/developer/n1s)
21:16:37kugelobo: have you tried to find the lcd_update function?
21:16:37oboI'd only replace the main firmware, not the bootloader
21:16:53obokugel: tried? yes, but not actually found
21:17:01kugelhow did you search for it?
21:17:19kugelon the samsas you could find it rather easily by searching LCD_WIDTH-1 or LCD_HEIGHT-1
21:17:49 Quit einhirn ("Miranda IM! Smaller, Faster, Easier. http://miranda-im.org")
21:18:08oboso far all the LCD functions have been grouped together - also grouped together as a collection of offsets near the end of the file
21:19:42oboI've found a single occurance (so far) of LCD_WIDTH next to LCD_HEIGHT
21:20:10kugelhave you tried searching for 239 and 319?
21:22:34obothey appear in the lcd_init code - but it does nice things like r3=0xFF, add r3 0x40 to create them
21:23:41kugeltheir compiler seems to suck :p
21:24:10funmankugel: it is a limitation of arm instruction set
21:24:23kugelwhat?
21:24:37funmanimmediate values are 8 bits with a shift
21:25:04funman0x13F is not representable this way
21:25:22kugelI thought they were 12 bits (i.e. 4096)
21:25:31funman0x140 == 0x14 << 4 can be
21:25:45kugel4095*
21:27:04kugelfunman: http://www.peter-cockerell.net/aalp/html/frames.html
21:28:24funmani don't know which instructions use the 12 bits for immediate values
21:28:50funmanmov use 8 bits + 4 bits for shifting (rotating right by an even amount)
21:29:11 Quit LambdaCalculus37 ()
21:29:20funmanlook for the "ARM Architecture Reference Manual" it has detailed instructions encoding
21:30:20funmankugel: the immediate values encoding is detailed in chapter 3.2 of your link (Immediate operands)
21:30:28kugeloh, I see. I read that wrong
21:30:52kugelI apparantly stopped after the first paragraph
21:31:13 Quit stoffel (Remote closed the connection)
21:32:48kugelobo: alright, that explains this doing then. But you can search for that too, and still for 239
21:34:31kugelif the update functions check boundaries before feeding the gram, they probably check for 319 and 239 instead (from becase the hw updates row0 upto row319, etc)
21:34:59 Join matsl [0] (n=matsl@host-90-233-242-98.mobileonline.telia.com)
21:35:49obokugel: I only see one instance where that value is assigned (rather than a add or sub)
21:39:59 Quit funman ("free(random());")
21:40:20 Quit JdGordon (Read error: 60 (Operation timed out))
21:41:06 Join JdGordon [0] (n=jonno@rockbox/developer/JdGordon)
21:42:50oboHmm, the LCD connector to the board has 40 lines - enough for a 16/18 bit interface as well as SPI (plus power/backlight)?
21:43:24 Join bmbl [0] (n=Miranda@unaffiliated/bmbl)
21:58:16kugelamiconn: ping
21:58:35kugelor, anyone
21:59:56kugelI've been trying to make the SCALE(x) macro in chopper.c a inline function (because it a) expands its argument 4 times and b) it is often used with expressions instead of plain vars). It gave a nice binsize boost, and likely also some speed bonus but I noticed something werid
22:00
22:00:59CIA-71New commit by zagor (r21755): Implemeted GIMMEMORE command instead of next-to-build queue.
22:01:27gevaertsDeath to unneeded kills! :)
22:01:33BagderZagor: does it just respond and then hand out a build to that command?
22:01:41Zagoryes
22:01:46kugelwhen I had it as normal inline (non-static) it was *less* binsize than making it a static inline.
22:01:47Bagdernice
22:02:25CIA-71New commit by zagor (r21756): New build distribution concept: Clients get one build at start, and then request more with GIMMEMORE command.
22:02:28kugelThe disassembly looked weird. In one case, all 6 arguments of a function are used with SCALE(x). Without the static I saw only 5 instances of this inline function + a stack operation, and with the latter I saw 6 instances (1 for each argument)
22:02:36pixelmaumm, I just noticed that running battery bench for checking runtime in FM mode is probably useless since there won't be any disk spinups... or am I wrong?
22:02:39CIA-71New commit by rob (r21757): D2: Remove unnecessary ide_power functions.
22:02:41kugelthat was all over the place, and I can't explain that :(
22:02:48BagderZagor: so when do they stop asking for more?
22:03:02Bagderah they only ask once?
22:03:10Bagderfor each completed I mean
22:03:19ZagorBagder: yes, they ask for more when they have available resources
22:03:21bertrikpixelma, it _should_ flush its buffer on power-down on when the buffer gets full
22:03:23pixelmakugel: is this macro used for the half width/height version?
22:03:29kugelyes
22:03:31ZagorI've added a little provision for single-core builds too
22:03:50pixelmabertrik: how long can it take entries (how many and what is this in hours)?
22:04:01saratogapixelma: doesn't it spin up eventually to log results?
22:04:06Zagortime for a live test...
22:04:12kugelthe macro version is doubtlessly the worst version, but I can't figure out what it does with static vs non-static
22:04:47pixelmakugel: amiconn had the idea of making it use bitmaps anyways and then you can adapt it better to different screen sizes and it will look better. I even drew images...
22:05:21kugelchopper really doesn't need bitmaps imo
22:05:22*amiconn hrmphs
22:05:51bertrikpixelma, about 16 hours
22:06:11pixelmakugel: I think it will look a lot better - and as I said will be easier to adapt to different screen sizes.
22:06:34pixelmabertrik: what happens on overflow?
22:06:36kugelhow are bitmaps easier to adapt? It means manually converting every bitmap
22:07:05bertrikpixelma, it tries to flush the buffer anyway, if it can't it ignores new values
22:07:16kugelchopper just uses plain foreground color, without any fancy color shades or something.
22:07:39pixelmawell, can you scale to arbitrary sizes? Like 1.23425 times?
22:09:23kugelthat would be like using using bitmaps for the line selectors to make be better adapted to different font sizes
22:09:47pixelmaI don't understand what colour shades have to do with that. The images I drew are 2 chopper "states" each (just the rotor a bit differently)
22:10:24pixelmawhat?
22:10:35kugelanyway, I don't care
22:10:46kugelI'm after the SCALE(x) thing, that makes me curious
22:10:53 Join linuxstb [0] (n=linuxstb@rockbox/developer/linuxstb)
22:11:34pixelmabertrik: damn... how do I find the FM runtime on my M5L out, in case it can't write the log away...
22:13:57kugelgevaerts: yea, I think I made about all bootloader
22:14:09amiconnZagor: Btw, I think that adding timestamps to the log output would be a good idea
22:14:21*Bagder agrees with amiconn
22:14:27Zagorsure
22:15:06kugelZagor: I don't seem to get a new build after getting one killed
22:15:25Zagorah, that's a client bug
22:15:26BagderI see the same here
22:15:43Zagorit doesn't ask for a new when it gets killed. fixing.
22:16:02kugelI'd also find it nice if the commands would be printed more verbosely, is there an option for that?
22:16:05JdGordon|svn: Syntax error in revision argument 'mt'
22:16:09ZagorI don't understand why the bootloaders come on the middle of the round. they should be after sims.
22:16:28rasherJdGordon|: that doesn't sound right
22:16:33JdGordon|thats my output
22:16:39JdGordon|something is FUBAR
22:16:40gevaertsSame error here
22:16:56Zagoryep, something is wrong. exiting server.
22:17:58ZagorI'm probably being punished for removing the @buildids array...
22:18:17Bagderprobably
22:18:39*amiconn got the same svn error in one client, but a different svn error in the other client
22:18:47amiconns/in/on/g
22:18:52*rasher saw no svn errors
22:18:54rasherBut that could be luck
22:19:49kugelI was getting those too
22:20:08pixelmaI also got an svn error and a "Server socket disconnected" after removal of builds
22:20:26Zagorperl is a bit nasty here. if you check for the presence of $hash{key}{value}, suddenly $hash{key} is created...
22:20:46Zagorunless you use "defined"
22:21:02kugel"a bit"? "here"?
22:21:25Zagorthose who don't know perl are doomed to reinvent it :)
22:22:22 Quit amiconn (Remote closed the connection)
22:22:23 Quit pixelma (Remote closed the connection)
22:24:19JdGordon|have we got any logging mechanism for the ipod accessories so users can try debbuging ?
22:24:50kugellogf...
22:25:07kugeli.e. no
22:26:26saratogatheres a patcht hat logs communications over the port on teh tracker I believe
22:26:34 Join guest001 [0] (n=someone@151-99.106-92.cust.bluewin.ch)
22:27:45 Quit guest001 (Client Quit)
22:28:36 Join pixelma [50] (n=pixelma@rockbox/staff/pixelma)
22:28:38 Join amiconn [50] (n=jens@rockbox/developer/amiconn)
22:30:18CIA-71New commit by zagor (r21758): Ask for more after CANCEL.
22:36:44 Join soycamo [0] (n=camo@c-76-115-2-113.hsd1.or.comcast.net)
22:37:54CIA-71New commit by bagder (r21759): Add 'Onda VX777' to the mix
22:37:57soycamoHi. How do I edit or create fonts for Rockbox?
22:39:23soycamoAre they just flat bitmap files?
22:40:26JdGordon|they are .bdf
22:41:03 Quit n00b81 ("Leaving")
22:41:53kugelJdGordon: fnt is bitmap, isn't it?
22:42:41soycamojdgordon| all I see are fnt files which I'm not sure what to open with.
22:44:33Bagdersoycamo: they're converted to fnt from bdf
22:44:42Bagderand bdf is more genericly used for fonts
22:44:48 Quit matsl (Read error: 110 (Connection timed out))
22:45:08Bagderbdf is bitmap as well
22:46:29soycamobadger I can't open it w/ photoshop and Gimp is hella slow on osx
22:46:51Bagderit's a font format
22:46:56Bagderwhy would those tools work with it?
22:47:03saratogakugel: what is your build server that has 6x the bogomips of my quad opteron?
22:47:17CIA-71New commit by zagor (r21760): Reverted to use of @buildids array.
22:47:34soycamobadger good point... though uh... tool suggestion? guide?
22:47:34kugelsaratoga: a octa xeon (>3GHz)
22:47:47saratogawow
22:48:10gevaertssaratoga: it's wasted on bootloaders of all things...
22:48:32Bagdersoycamo: I have no idea. A font editor that can read bdf files?
22:48:44 Join funman [0] (n=fun@rockbox/developer/funman)
22:48:45ZagorBagder: I found the ping bug in the server btw. it detected it fine, it just didn't do anything about it. :)
22:48:54Bagderah, hehe
22:48:58Bagdernice find
22:49:06saratogai saw someone tested the new bootloaders on the H10 finally
22:49:08soycamobadger: found stuff, n/m
22:49:10saratogaare we ready to release them then?
22:49:11gevaertssoycamo: I believe people have used fontforge
22:49:20Bagderand my nick isn't spelled like that
22:49:36*kugel hands saratoga a TAB key
22:49:46kugeluh
22:49:48JdGordon|haha
22:49:49Bagderhahaha
22:49:51AlexPhaha
22:49:54kugelmy TAB hangs too
22:50:06amiconnfontforge is rather crappy imo. I'd suggest gbdfed
22:50:20gevaertssaratoga: everything is tested except H10 20GB.
22:50:39gevaertsej0rge: RockboxTesting says you have a 20GB H10 :)
22:50:55kugelfinally this page is useful :)
22:51:12soycamoThe UI is a little kludgy. Who can I report that to?
22:51:16gevaertskugel: I've used it before. I just don't always say :)
22:51:26Bagdersoycamo: the ui?
22:51:45soycamobadger: yeah. it's not very intuitive imo
22:52:11pixelmasaratoga: I always wonder why you sound so assured in the forums when giving advice. E.g. do you know for sure that the timestretch mode is really accessible (UI wise) on the c200? I have a c200 and would have to try... although the guy didn't sound like he tried very hard to find out how to use it
22:52:19AlexPsoycamo: UI for what?
22:52:22Bagderyou in vague terms speak
22:52:29*Bagder does a yoda impression
22:52:45soycamobadger: the ui for rockbox, ie when I turn on my c250, it is not obvious what I should do
22:53:07AlexP"Who do I report it to" haha
22:53:21*gevaerts looks around for this badger person
22:53:21soycamoAlexP: I thought there might be an Awesome UI Super Team
22:53:22AlexPFeel free to suggest specific improvements
22:53:58soycamoAlexP: okay. Well, there should be a very easy "Music" menu at the top, because that's most important.
22:53:59Bagdergevaerts: he might look for someone to badger!
22:54:10AlexPsoycamo: That goes where?
22:54:14saratogapixelma: if its on the e200 how would it not be on the c200?
22:54:17AlexPsoycamo: File tree or database?
22:54:28gevaertssoycamo: is it? For everyone?
22:54:28AlexPsoycamo: And if filetree, then which directory?
22:54:30rashergevaerts: Hush, that's our line!
22:54:39soycamoalexp: A link I guess? It's hard to explain this in text.
22:54:43pixelmasaratoga: because the c200 has a different keymap?
22:54:45funmankugel: (about SCALE(x) in chopper) perhaps one call was optimized?
22:54:51pixelmathere might be problems?
22:54:52AlexPsoycamo: A link where?
22:55:02soycamoalexp: at the root of Rockbox menu.
22:55:06AlexPsoycamo: Customisable menus are not very highly thought of
22:55:11AlexPsoycamo: To where?
22:55:12ej0rgegevaerts: I do have a 20gb H10
22:55:17kugelfunman: it didn't look like
22:55:19soycamoalexp: this is hard to describe in words, I'm sitting here with my c250
22:55:19ej0rgegevaerts: I plan to test the new bootloader on it tonight
22:55:25gevaertsej0rge: thanks!
22:55:28AlexPsoycamo: i.e. you click on music, what does it show you?
22:55:33AlexPsoycamo: The database?
22:55:35 Join matsl [0] (n=matsl@host-90-233-177-32.mobileonline.telia.com)
22:55:37AlexPsoycamo: The filetree?
22:55:40***Saving seen data "./dancer.seen"
22:55:46ej0rgegevaerts: No problem - I've been leeching off this project for years, so I'm happy to help :)
22:56:04soycamoalexp: the database.
22:56:11AlexPI don't use the database
22:56:17AlexPThat isn't how I get to music
22:56:29AlexPVirtually all the devs don't use the database in fact
22:56:36soycamoalexp: Don't think like a dev
22:56:44pixelmaexcept Slasheri maybe
22:56:51soycamoalexp: think like an idiot user. end user just wants music.
22:56:56AlexPSo for all those users that don't use the database, your music label would be wrong
22:57:07pixelmasoycamo: why shpuld we make the life harder for us?
22:57:08AlexPWe don't force people to use a database
22:57:09soycamoalexp: that could be fixed in settings
22:57:17gevaertssoycamo: also, lots of people use it for non-music audio...
22:57:28soycamopixelma: do you ever have that moment when you want it to *work* without thinking?
22:57:33AlexPIt does
22:57:41gevaertssoycamo: what's so hard about choosing "Database"?
22:57:42soycamopixelma: yes, and I use it for both
22:57:55pixelmasoycamo: you could also read the manual and maybe find the start screen options...
22:57:55kugelgevaerts: how could we optimize what the server is giving a client upin GIMMEMORE?
22:57:58AlexPI think - I want to use the database, or I want to browse files, so I choose "Database" or "Filetree"
22:58:05soycamogevaerts: it is non-obvious. I am approaching this from a user perspective
22:58:12kugelor is that a plain linear list now?
22:58:44*soycamo feels like she is talking to a wall
22:58:49AlexPsoycamo: And making things either wrong or more complicated
22:59:04soycamoalexp: okay, sorry, never mind.
22:59:16AlexPDon't get me wrong, improvements are welcome
22:59:35AlexPBut a big factor of Rockbox is to allow access how people want to what they want
22:59:45soycamoalexp: and I want it to be stupidly simple
22:59:53AlexPRockbox isn't for you then
23:00
23:00:00soycamoalexp: why not?
23:00:08AlexPIt has too many options and possibilities to be stupidly simple
23:00:16pixelmait won't be stupidly simple if it goes to the database automatically and you don't even use it
23:00:17AlexPThat is why we have a rather fine manual
23:00:35pixelmait = database here
23:00:40soycamoalexp: but what if I want to have "zomg easy button" for when I'm on my bicycle, and then use the menus for "srs business" mode
23:00:42n1ssoycamo: i find a database to be extra complication, i know where my files are on the drive i just want to get to them
23:00:57AlexPsoycamo: That just makes it even worse
23:01:04soycamoalexp: why?
23:01:16AlexPsoycamo: Why can't I find this setting? Oh, maybe I'm in the wrong mode
23:01:19soycamon1s: then why have the database then?
23:01:29AlexPDo I have to switch modes to be able to do x?
23:01:39soycamoalexp: I'm just asking for a shortcut to a shuffle mode for when I'm on my bicycle
23:01:40AlexPsoycamo: As some people do use it
23:01:48AlexPsoycamo: Quick screen
23:01:55n1ssoycamo: i think a few people would whine if i removed it :)
23:02:00soycamojesus christ if you all aren't interested in my suggestions fine
23:02:12pixelmaresume button
23:02:27 Part soycamo
23:02:29n1ssoycamo: your suggestion is "make the ui fit my usage pattern"...
23:02:54funmanwhile the policy is "make the ui fit the developers' usage pattern"
23:03:34pixelmaAlexP: the quickscreen is only easily available from the WPS on the c200 unfortunately
23:03:43n1sor at least try to compromise to make if as good asm possible
23:03:56n1sfor as many people as we can
23:04:02AlexPpixelma: Ah, I misread e not c
23:04:20saratogapixelma: its a standard menu option, why would it need a different keymap?
23:05:11pixelmasaratoga: doesn't it need switching the mode in the pitch screen, I don't know whether that works correctly
23:05:43gevaertsWho can build osx binaries for sansapatcher and e200rpatcher?
23:05:59AlexPsaratoga: Where can you set the timestretch rate in the menus?
23:06:13JdGordon|gevaerts: domonoky1 can on my osx box
23:06:17AlexPrather than just on and off
23:06:31JdGordon|I can give you access if you want
23:06:57saratogaAlexP: well he said it was unimplemented so presumably that refers to the menus . . .
23:07:20pixelmasaratoga: I can imagine it does, since pondlife has a c200 too, if I remember correctly. That was just an example though, but maybe it just bothers me a bit since I'm usually very careful when giving advice
23:07:39saratogabut what I meant was that its added to the existing screen so its hard to imagine we wouldn't have had a keymap for it
23:07:45 Quit brndyhite (Remote closed the connection)
23:07:45 Quit archstech (Remote closed the connection)
23:07:45 Quit tucsbgns (Read error: 54 (Connection reset by peer))
23:07:55saratogapixelma: is there some specific complaint you have?
23:09:10amiconnsaratoga: There is a difference in having a keymap and having a keymap where all actions are accessible
23:09:19pixelmasometimes it just sounds too quick too sure to me (but as I said, maybe it's because I'm not)
23:09:19 Quit _zic (Read error: 54 (Connection reset by peer))
23:09:25JdGordon|last chance for people to comment on my clip keymap changes... or ill commit when I get home (3 hours or so)
23:09:53saratogaI admit I do not have a c200 to test on but I am certain we have implemented time stretching on the c200
23:10:11 Join archstech [0] (n=archstec@206.251.250.215)
23:10:31amiconnWe do. The uncertainty lies in whether it is accessible
23:10:36pixelmathat yes, but is it really "usable"? E.g. there is metronome on the c200 too but you can't really use it
23:12:21pixelmaargh, stupid MacOS
23:12:25saratogawell i don't mean to imply that we have no playback bugs, so if hes certainly welcome to report one
23:12:40saratoga"so hes certainly welcome to report one"
23:14:39amiconnThere can be issues with keymaps. Even if an action is mapped that doesn't mean it actually works. It can e.g. be covered by another action in the context - context chaining can hide this pretty well
23:14:51CIA-71New commit by bagder (r21761): mktitlepics.pl is a tool for generating a full set of title bitmaps for the ...
23:15:05amiconnOr there is an electrical or mechanical constraint that causes a combo to be impossible
23:15:22ZagorBagder: excellent!
23:15:25gevaertspixelma: in this case I don't think it changed any keymaps, so if the pitchscreen used to work, the new stretch feature will as well
23:15:39BagderZagor: it'll make them in the same dir right now
23:15:50 Join soycamo [0] (n=camo@c-76-115-2-113.hsd1.or.comcast.net)
23:15:56pixelmagevaerts: yes, but I don't know if the pitchscreen works
23:15:56gevaertsand I think that the pitchscreen is old enough for us to know if it would be broken
23:16:18amiconnBagder: Couldn't we use css for rotating text?
23:16:23Bagderno
23:16:28*amiconn has no real idea whether css allows that
23:16:30Bagderthere's no such thing in css
23:16:44Bagderonly in some very recent version that no browser supports
23:16:53pixelmahaven't seen something like that either
23:17:24Bagderand this is pretty easy anyway so...
23:17:54Zagoris ondavx777 harder or easier or the same as ondavx767?
23:18:01Bagdervirtually the same
23:18:05Bagderafaiu
23:18:10ZagorI'll give it the same score then
23:18:23Bagderright, forgot about that
23:19:02Zagordoes 747 have plugins while 767 doesn't?
23:19:12Zagor747 is listed as our second heaviest build
23:19:22CIA-71New commit by zagor (r21762): Added ondavx777.
23:19:50Bagderoh
23:19:55BagderI don't know the details
23:20:03Bagdermaurus is the man
23:20:18Bagder(too hard to spell nick for me to try without tab completion)
23:20:23pixelmagevaerts: and there was a change in WPS keymaps recently (don't remember testing accessibility of the pitchscreen)
23:20:54gevaertshm, true
23:21:32 Quit DarkDefender ("Leaving")
23:24:03 Quit jgarvey ("Leaving")
23:26:35ZagorBagder: I guess we should merge build and buildmaster. the division is a bit messy now.
23:26:57Bagderright
23:29:51 Quit Thundercloud (Remote closed the connection)
23:32:12kugelpixelma: metronome is the reason I'm asking you to test my PLA rework
23:33:04 Nick soycamo is now known as soycamo|duchando (n=camo@c-76-115-2-113.hsd1.or.comcast.net)
23:36:27CIA-71New commit by zagor (r21763): Added total build round time.
23:37:43 Quit n1s ("Lämnar")
23:38:13funmanwhat is wrong with metronome on c200 ?
23:38:40gevaertsfunman: nearly everything works. Only the button to actually make it make noises doesn't
23:40:04funmanah it's a keymap problem, i thought it was a PCM or DMA hardware problem
23:41:29amiconnSame problem also exists on the Player
23:41:40pixelmaand Ondio
23:41:57amiconnOh?
23:42:01*amiconn didn't know
23:44:28pixelmaZagor: did your commit trigger a build round?
23:44:38Zagorno
23:45:24pixelmaok, already thought something was wrong here (also got a few "Server connection stalled" in the backlog
23:45:56Zagorpixelma: you won't get those anymore. the server disconnects faster than the client now.
23:46:16Bagderwell, they can still happen if the client has lost touch with the server
23:46:37Bagderin a silent manner I mean
23:46:43Zagoryes but the server will disconnect first, which the client will react to with a "cleanup + restart"
23:47:04Zagorah right, if the net is really borked
23:47:05Bagdernot if someone pulled the cable in a router in between, then the client won't notice the disconnect
23:47:20Bagderright
23:48:19obowill the new build system handle standby/hibernation mode?
23:49:07Zagorobo: yes, it handles clients coming and going as they wish
23:50:06obobut when the client powers back up, I'm not sure it reconnects?
23:50:18Bagderit will
23:50:37Bagderif it continues, it'll detect the idle connection, disconnect and reconnect again
23:54:15pixelmasaratoga, gevaerts: timestretch is also accessible on the c200, I really had to look up how to enter the pitchscreen in the manual though (or the code)
23:54:51CIA-71New commit by zagor (r21764): Added timestamp to printouts.
23:56:46 Quit soycamo|duchando (Read error: 110 (Connection timed out))
23:57:23 Quit CaptainKwel ("Page closed")

Previous day | Next day