Log from #neuros IRC channel

[01:34]

hey all

[02:20]

morning all :)

[02:54]

* crweb is teaching basic C++ today

[03:54]

anders_: ping

[03:56]

<anders_>

nerochiaro: pong

[03:58]

anders_: can you explain why it's not kosher to keep the medialib on NFS ? we are using sqlite on NFS in the OSD for other things and it works just fine. why won't it for xmms2 ?

[03:58]

<anders_>

locking issues.

[03:59]

<anders_>

filelocks may not work reliably over nfs.

[03:59]

you do any special locking in xmms2 on that DB ?

[04:00]

<anders_>

No, just the locking that is done with sql transactions.

[04:00]

crweb: so it might work for our apps because we don't use transactions ?

[04:04]

<crweb>

we don't use transactions

[04:05]

<crweb>

well, my understanding of transactions

[04:06]

ok, so i guess that why our apps work and xmms2 doesn't with nfs

[04:06]

<crweb>

wait

[04:06]

<crweb>

we do transactions..

[04:07]

<crweb>

anders_: sqlite says that its only a problem if 2 processes access the db. We haven't had any problems with nfs and sqlite.

[04:08]

that was my understanding by reading sqlite FAQ too

[04:09]

<anders_>

xmms2 probably works in most cases aswell. Thats why it is a config variable.

[04:10]

anders_: i tried it and i can add stuff to playlists, but when starting playback it complain it can't retrieve the URL of for the entry and then " FAIL: xmms_ringbuf_wait_used: assertion `len <= ringbuf->buffer_size_usable' failed" and crash

[04:10]

looks like it fails on reading from the DB, or wrote bad data to it

[04:11]

<anders_>

Oh, thats interesting.

[04:12]

<crweb>

nerochiaro: if you take that db file and read it with sqlite hows it looks?

[04:12]

<crweb>

err.. how does it look?

[04:12]

crweb: didn't try. in fact i jsut deleted it. making it again and trying

[04:22]

crweb: it's corrupt

[04:22]

Unable to open database "/srv/neuros/fs/nfs-debug-extended-area/.osd-extended/data/xmms2_medialib.db": file is encrypted or is not a database

[04:22]

anders_: crweb: do you want me to send it to you ?

[04:23]

<anders_>

Sure.

[04:25]

sent

[04:26]

<anders_>

Opens fine here..

[04:26]

what commend do you use ?

[04:26]

<anders_>

(did you use sqlite3 commandline tool, and not sqlite2?)

[04:27]

aw dammit, it installed the 2

[04:27]

<crweb>

you want just plain sqlite

[04:27]

<anders_>

The medialib looks fine to me.

[04:27]

<crweb>

2 was a special version

[04:28]

crweb: apt-get install sqlite does install the 2, appearently. anyway, not an issue. i installed the 3 and it works ok

[04:28]

<crweb>

huh.. mine installs 3

[04:28]

anders_: so it's xmms2 that fails to read it correctly ?

[04:28]

<crweb>

ooo, my mistake. nvrmind

[04:29]

<anders_>

nerochiaro: Could you provide full log of xmms2d -v?

[04:29]

anders_: also on the insert of files ? or just playback failure ?

[04:29]

<anders_>

nerochiaro: Just playback failure is a good start.

[04:39]

anders_: pastebin.ca is this what you needed ?

[04:40]

<anders_>

nerochiaro: Doesn't look medialib related to me.

[04:40]

<anders_>

It opens the correct file nicely.

[04:41]

hmm, let me try with different files then

[04:41]

but i could play FLAC files from the same media when the DB was not on NFS

[04:49]

anders_: i tried with an mp3 on another media, and it gave me the same error

[04:51]

<anders_>

You haven't changed anything with the output plugin/output buffer size?

[04:51]

let me check, i may just be *that* stupid

[04:52]

anders_: what's the default output.buffersize ? 4096 ?

[04:53]

<anders_>

32k IIRC

[04:53]

then that's what i broke. my mistake

[04:53]

let me try with that fixed

[04:56]

anders_: embarassingly enough, that fixed it

[04:56]

they play now

[04:57]

<anders_>

nerochiaro: Great!

[05:00]

anders_: hopefully it will work fine on NFS then

[05:00]

anders_: another question i have is with the output buffer (which was the reason why i was messing wiht it)

[05:01]

anders_: can we detect underruns in that buffer and force xmms2 to refill the buffer up to some limit (or to full buffer) before resuming playback after the underrun ?

[05:02]

anders_: because i have this issue that sometimes there are underruns and xmms2 fills it up a bit, but start depleting it again immediately, so it enters some kind of stuttering loop (underrun, small refill, underrun, small refill, etc)

[05:03]

<anders_>

Maybe you could add a xmms_output_wait_buffered function which calls xmms_ringbuf_wait_used on the ringbuf..

[05:05]

wait_used is a function that makes you wait ringbuf to reach a certain fill size ?

[05:05]

<anders_>

Yes. (defined in src/xmms/ringbuf.c)

[05:06]

when would i call output_wait_buffered ?

[05:07]

<anders_>

nerochiaro: When you output plugin does xmms_output_read and detects it didn't fill the buffer properly?

[05:07]

<anders_>

nerochiaro: Or rather before next iteration.

[05:09]

that makes sense. another thing i was thinking was about doing a full buffer fill when a new track is loaded ? wouldn't that be easier ? (i know it breaks gapless, but it's kind of acceptable)

[05:14]

anders_: ?

[05:14]

<anders_>

Not sure it would be easier.

[05:15]

<anders_>

And it doesn't really have to break gapless either.

[05:15]

anders_: because it may be we can do the fill before the previous track is finished ?

[05:15]

<anders_>

Yes.

[05:17]

<anders_>

Maybe you could make output.c:song_changed to do xmms_ringbuf_wait_used..

[05:17]

anders_: i think i'll look into that

[05:18]

<crweb>

I think thats what we should do

[05:18]

<crweb>

i turn gapless off anyway

[05:18]

<crweb>

lol

[05:18]

crweb: you are statistically insignificant anyway ;)

[05:35]

anders_: can i add new sections to xmms2.conf from outside xmms2 (editing the file manually) and trust xmms2 not to delete them ?

[05:35]

(add them before xmms2 starts)

[05:56]

<anders_>

nerochiaro: Yes, I think so. What are you planning to do?

[05:58]

anders_: when we update our software sometimes we need to force the xmms2 config to change. so i'm thinking on wiping it conditionally based on a version section that is read by an init script

[05:58]

but i don't want xmms2 to manage that version section (by registering it etc)

[05:59]

i just want a script to check it, if it's ok don't wipe, if it's wrong wipe and copy over an updated config that has a newer version

[05:59]

but if xmms2 removed the version section, then it won't work

[06:00]

that's why i asked

[06:00]

<crweb>

nerochiaro: i used a external version file

[06:00]

crweb: for what ?

[06:00]

<crweb>

xmms2.conf.23

[06:00]

crweb: please explain ?

[06:00]

<crweb>

you sure. aparently me explaining things isn't going so well..

[06:00]

crweb: maybe later if we are busy in the other meeting

[06:04]

<crweb>

init script says: VERSION = 34; if /mnt/OSD/xmms2.conf.$VERSION does not exist, rebuild config (copy it)

[06:04]

<crweb>

agh.

[06:07]

<crweb>

init script says: VERSION = 34; if /mnt/OSD/xmms2.conf.$VERSION does not exist, rm /mnt/OSD/xmms2.conf*, cp new config (copy it)

[06:08]

<crweb>

touch /mnt/OSD/xmms2.ocnf.$VERSION

[06:09]

<crweb>

see ldgen

[06:09]

i get the idea

[06:09]

sounds good enough

[07:13]

<crweb>

nerochiaro: did you see new deskmon ?

[07:13]

<crweb>

better yet.. did i send it..

[07:14]

<crweb>

anyway, i removed all paths and made the path specified at cmdline

[07:15]

<crweb>

and, did you see that NWaitingBoxWorkerThread leaks. I think i addressed this in email about the thread just pinging you.

[07:15]

<crweb>

oh

[07:15]

<crweb>

and you replied 8 minutes ago.. nvrmind

[07:16]

ok

[07:16]

didn't see new deskmon

[07:16]

<crweb>

i sent to review

[07:16]

will look

[09:00]

guys does neuros's qtopia support all the fonts in that font folder?

[09:02]

<dab>

hi all

[09:03]

<dab>

newbie help needed on how and which version to get from svn!

[09:04]

<dab>

anybody awake here?

[09:04]

dab: get trunk

[09:04]

dab: or get the one named "torfu" if you prefer the non-QT version

[09:05]

<dab>

ok. i did for d in $(ls) ; do svn up $d ; done

[09:05]

<dab>

is this ok

[09:05]

depends on which one you first checked out. what does "svn info" say when you call it in one of the directories

[09:05]

?

[09:06]

<dab>

hmm. i downloaded the tarball dated 26th march

[09:06]

<dab>

then i built that one

[09:06]

<dab>

but now i already gave that svn up

[09:07]

<dab>

and its in progress.

[09:07]

<dab>

its downloading qtopia

[09:07]

<dab>

can i run svn info now or should i wait till it finishes?

[09:08]

you can run it from another shell, no problem

[09:09]

but i think you have trunk

[09:09]

since you're downloading QT

[09:09]

so if trunk is ok for you, then the svn up command was ok too

[09:10]

<dab>

ok.

[09:10]

<dab>

Path: .URL: svn://svn.neurostechnology.com/svn/neuros-bsp/tags/OSD_VER_3.33-2.05-00.837_080307Repository Root: svn://svn.neurostechnology.com/svn/neuros-bspRepository UUID: ed9ee5b5-7908-0410-92af-aedb66848885Revision: 1606Node Kind: directorySchedule: normalLast Changed Author: clinLast Changed Rev: 1091Last Changed Date: 2007-06-29 05:45:06 -0500 (Fri, 29 Jun 2007)

[09:11]

wow, it's a tag

[09:11]

<dab>

so its ok?

[09:11]

no, if you want trunk, you should do an svn switch to trunk

[09:12]

svn://svn.neurostechnology.com/svn/neuros-bsp/

[09:12]

<MattJ>

dab: The build-helper script has an update command I think

[09:12]

<dab>

:-) i really dont know much... learning all the way... following the guides

[09:12]

MattJ: that updates only within the same branch/tag

[09:13]

dab: that is the way to go :)

[09:13]

<dab>

it would be better if you guys help me out by telling the exact commands to give

[09:13]

<dab>

should I stop that svn up script?

[09:13]

let it go

[09:13]

it won't hurt

[09:13]

<dab>

ok.

[09:14]

you will have less stuff to download when you do the switch anyway

[09:14]

<dab>

ok.

[09:14]

<MattJ>

nerochiaro: I meant instead of having to do for d in $(ls) ; do svn up $d ; done

[09:14]

MattJ: that's true

[09:14]

<dab>

but that qtopia/dist is large!

[09:14]

dab: svn switch svn://svn.neurostechnology.com/svn/$d/trunk $d

[09:14]

dab: and repeat it in a for loop

[09:15]

but wait for that up to finish

[09:15]

it's large but you need it all

[09:15]

patience

[09:15]

<dab>

:-)

[09:16]

<dab>

thanks... its only a 512kbps dsl connection

[09:16]

<dab>

not like what you guys have in the western world

[09:16]

it's the server that's slow, not your connection

[09:16]

<dab>

ok...

[09:16]

well, both

[09:16]

<dab>

its a good news to me.... that my connection is faster than the server!

[09:17]

well not so good for the server, but anyway :)

[09:17]

<dab>

Hi greyback!

[09:17]

dab: hey, how is netboot going?

[09:18]

<dab>

grayback: just fantastic

[09:18]

<dab>

I even built a tarball downloaded and ran it

[09:18]

<dab>

now nero is helping me out for svn thing to get updated.

[09:19]

Nice work! Am looking forward to seeing what you produce :)

[09:19]

<dab>

:-) I hope not to disappoint you.

[09:19]

I doubt that somehow

[09:20]

<dab>

:-D

[09:20]

<dab>

I have some embedded experience... not with emb. linux though

[09:20]

<dab>

did some DSP things too...

[09:21]

<dab>

again not on the TI ones....

[09:21]

Are you thinking of working on the DSP?

[09:22]

<dab>

:-D

[09:22]

<dab>

you got to do what you got to do...

[09:22]

Good luck, but I don't envy you!

[09:22]

That sort of stuff is way over my head

[09:23]

<dab>

:-) lets see how far i go...

[09:24]

dab: get in touch with may1937 if he's still around or Darchon, if you want to do DSP stuff

[09:24]

they worked on that, Darchon especially

[09:24]

hi dab

[09:24]

there you go

[09:24]

<dab>

Hi!!

[09:25]

hwat is it you want to know?

[09:25]

what*

[09:25]

<dab>

I know that question has been asked before... but cant help it

[09:25]

<dab>

I want raw frames captured via the analog inputs

[09:25]

ah, alright

[09:25]

that's possible

[09:26]

<dab>

whao!

[09:26]

but requires alot of work

[09:26]

<dab>

hmm...

[09:26]

as you'll have to write your own "driver" for the DSP

[09:26]

<dab>

where should I start?

[09:26]

<dab>

ok..

[09:26]

and a small RTOS

[09:26]

etc. etc.

[09:26]

<dab>

ok.

[09:27]

basically you start from scratch if you want that

[09:27]

as the current modules are closed API

[09:27]

<dab>

ok.... but any starting points?

[09:27]

<dab>

Do I have to sign an NDA with TI guys!!!

[09:27]

* Darchon nods

[09:28]

<dab>

I dont think they will entertain me.

[09:28]

atleast... I think the video input is connected to the DSP

[09:28]

dab: the NDA is an easy matter, you sign it with Neuros, not with TI

[09:28]

<dab>

Ya... the block diagrams shows so

[09:28]

<dab>

thats a good news.

[09:29]

<dab>

will neuros entertain me?

[09:29]

darchon.net <- link to my GSoC2007 project

[09:29]

<dab>

ok. I am on it.

[09:29]

I have started on translating the assembly to C

[09:30]

as there's finally a free compiler for the DSP

[09:30]

dab: we're very entertaining people ;)

[09:30]

dab: what do you mean with "entertain" ?

[09:30]

the next obstacle is really the lack of a debugger....

[09:30]

<dab>

I thought there is a JTAG on the board!

[09:30]

my project only supports "printf" type debuggin :P if you know what I mean

[09:30]

<dab>

but to debugger?

[09:31]

<dab>

*to = no

[09:31]

nope

[09:31]

no debugger

[09:31]

<dab>

hmm....

[09:31]

I had to debug my stuff using memory dumps

[09:31]

<dab>

its like working on 8052 then

[09:31]

and my debug library

[09:31]

<dab>

ok. thats a great job you have done

[09:31]

which basically sends a message to the ARM linux module, which prints in in the syslog

[09:32]

<dab>

ok...

[09:32]

and sometimes it misses one

[09:32]

so yeah... debugging is crap

[09:32]

<dab>

is the compiler that TI giving is of some use?

[09:32]

yeah

[09:33]

it compiles

[09:33]

<dab>

hehehe

[09:33]

and links

[09:33]

but that's all it does really

[09:33]

it's not like you have any other choice, there's no gcc for it

[09:33]

<dab>

I was told earlier that the documentation is not available

[09:33]

dab: only in NDA

[09:33]

<dab>

ok.

[09:34]

but if you're interested in this DSP stuff

[09:34]

I suggest applying for GSoC2008

[09:34]

<dab>

Darchon: you have really did a fantastic job (I was looking at your site)

[09:34]

for the OSD 2.0

[09:34]

I can give you some pointers on your application

[09:34]

<dab>

nero: how can I proceed with the NDA?

[09:35]

cuz yeah, OSD 2.0 has that TI Codec thingie

[09:35]

<dab>

Darchon: sure ....

[09:35]

which gives you direct access to the raw frames (I hope)

[09:35]

dab: send an email to Joe and CC me, explain what you need to do and we will send you the NDA doc

[09:35]

<dab>

Are OSD 2.0 is out in beta?

[09:35]

adn!

[09:35]

and!*

[09:35]

you get TI's code composer studio from neuros

[09:36]

which does have a debugger :)

[09:36]

<dab>

ok...

[09:36]

Darchon: not sure that applies anymore

[09:37]

ah..

[09:37]

Darchon: but doesn't hurt to ask

[09:37]

<dab>

hmm.... so the open things are not all that open...

[09:37]

nerochiaro: it said so on the wiki

[09:37]

dab: the DSP is not open, we never said that :)

[09:37]

<dab>

I know.

[09:37]

"The student awarded this project will be loaned an OSD 2 prototype (obviously) and a complete suite of hardware and software tools for porting the codec, including TI's proprietary development environment, Code Composer Suite."

[09:38]

Darchon: yup, but that was A) for SoC and B) before they released the open compiler

[09:38]

yeah, I know

[09:38]

<dab>

so does that means I cant buy a beta unit?

[09:39]

I was saying that if Dab applied for that project, he might get a real debugger and some really cool hardware (the OSD2.0)

[09:39]

<dab>

I dont want to do anything else but the project that I have in mind...

[09:40]

<dab>

A CCTV camera connected to neuros

[09:40]

<dab>

and when it detects motion, it records/sms/mail and what not.

[09:40]

dab: osd20 or current ?

[09:40]

<dab>

well, I thought that it would be possible on the current

[09:41]

<dab>

so I bought 2 of them

[09:41]

Darchon: ah, wait, i though you were looking at soc 2007, sorry

[09:41]

that's 2008, then everything in there is still valid

[09:41]

my mistake

[09:41]

dab: it's possible, you just need the official documentation

[09:42]

it's just not as easy as it would've been if the video input was connected directly to the ARM

[09:42]

but if the ARM's fast enough, simply using the DSP DMA controller to copy video data to SDRAM, would do the trick

[09:43]

and then you can just do all the fancy stuff on the ARM

[09:43]

the DSP wouldn't even need an OS running

[09:44]

<dab>

yeah. I might start using a USB video capture device

[09:44]

it just waits for the video input and copies, notifies the arm. and keeps repeating that cycle

[09:44]

dab: on osd1 don't try that

[09:44]

dab: osd1 USB sucks

[09:44]

<dab>

:-D

[09:45]

<dab>

so I gather its not fast enough for full screen full speed SD frame capture in YUV

[09:46]

not sure about the speed, but the USB driver is broken with regards to some transfer modes. if you device needs them, you're screwed.

[09:46]

<dab>

Darchon: how far your project has gone...

[09:46]

<dab>

ok.

[09:46]

i can copy from internal DSP memory to SDRAM

[09:46]

using circular buffer

[09:47]

but the buffer is kinda small

[09:47]

Darchon: video frames or test data ?

[09:47]

only (2^16)-1 bytes

[09:47]

any data that's in the DSP memory

[09:48]

I've only copied strings so far, that's true

[09:48]

<dab>

wait... thats 64k only!

[09:48]

but it's done using the DMA controller

[09:48]

so I don't see why it should have problems with other types of data

[09:48]

<dab>

and here goes my all video data in the water...

[09:49]

well, it just means you have to copy a few times

[09:49]

<dab>

ok...

[09:49]

<dab>

I can also cut down my requirements to grayscale images.

[09:49]

<dab>

may be SIF resolution image

[09:50]

<dab>

If thats not fast enough... but I just cant stop of dreaming of running my code in the DSP!

[09:51]

no, you just shouldn't make any mistakes whilst coding

[09:51]

so you won't need a debugger ;)

[09:51]

<dab>

:-)

[09:51]

I'm not saying it's impossible without... it's just harder...

[09:51]

<dab>

May be your libraries will help me out...

[09:52]

<dab>

if TI is not helping...

[09:52]

ow... wait

[09:52]

dab: you get only TI's docs

[09:52]

my kernel buffers are only 16k

[09:52]

dab: TI won't talk to you

[09:52]

<dab>

yeah... I dont buy 1 million chips annually.

[09:52]

I forgot I had problems allocating bigger consecutive memory space

[09:53]

* Darchon notes to dab that the DSP has no MMU

[09:53]

<dab>

hmm...

[09:53]

so if you want a bigger buffer the copying mechanism has to be updated

[09:53]

so I will swith to another buffer location

[09:53]

I = it*

[09:55]

<dab>

ok... So... currently it is not possible to have a frame in memory and have a callback... is it true?

[09:56]

nope

[09:56]

next month I hope to have more time

[09:56]

as I only have 1 project at school then

[09:57]

maybe I can update my project then

[09:57]

<dab>

is there any, by chance, any codec which has the source open available for this DSP?

[09:57]

sort of...

[09:58]

but I wonder how legal it is...

[09:58]

and it's audio codec only

[09:58]

Darchon: which one ?

[09:58]

<dab>

any crappy codec..

[09:59]

Vorbis if I remember correctly

[09:59]

the one we used on Neuros2 ?

[09:59]

<dab>

that will have access to the raw frames... and If we modify it to do the motion detection instead of encoding then we are done.

[09:59]

no...

[09:59]

that may be useful to him, maybe

[09:59]

Darchon: what other ?

[09:59]

I'm talking about ArchOpen

[10:00]

they have C code in their distribution

[10:00]

for the DM320

[10:00]

and they link to the TI BIOS stuff

[10:00]

so yeah... I don't see how that works... as it's an open source project

[10:01]

<dab>

ok.

[10:01]

<dab>

thanks guys I have to leave now.

[10:01]

www.archopen.org

[10:01]

<dab>

Will join you again tomarrow.

[10:01]

check that link out

[10:02]

<dab>

thanks once again for the enlightment.

[10:02]

it has C-code for the DSP