Log from #neuros IRC channel

<gremlin[it]>

ping crweb

<gremlin[it]>

hi all

crweb: which means i approved the binaries too

<crweb>

anti-pong

hi gremlin[it]

<gremlin[it]>

hi nerochiaro ...

<gremlin[it]>

hi crweb

<gremlin[it]>

i still have trouble ...

<gremlin[it]>

but ...

[02:38]

<crweb>

gremlin[it]: you didn't hear it from me. but the big man on top is becoming very interested in your OE work.

[02:38]

<gremlin[it]>

if after login into neuros i chroot on mtdblock4 and execute /etc/init.d/rcS all seem to work ...

<gremlin[it]>

so i suppose there is just something i do in the wrong order or something i miss ... but i don't understand what ...

<crweb>

what is the problem?

<crweb>

it won't run on startup?

<gremlin[it]>

exactly

<crweb>

this has to do with inittab

<gremlin[it]>

if i execute by hand

<crweb>

# Run the System Initialization Script

<crweb>

::sysinit:/etc/init.d/rcS

<gremlin[it]>

mainsetup start

<gremlin[it]>

yes i know ...

<crweb>

/etc/inittab

crweb: Gao ?

<crweb>

nerochiaro: yes

crweb: didn't know. that's cool

gremlin[it]: see what i told you ?

;)

<crweb>

nerochiaro: i pushed very hard today for a non neuros-bsp for new stuff

crweb: you can count me as an ally there

<crweb>

i don't know enough info for a full sale though

<gremlin[it]>

there is a doc that explain the sequence of software to start ?

<crweb>

so i imagine you know who is in channel to pick up some gremlin[it]'s info

[02:42]

<gremlin[it]>

nerochiaro, probably no ...

[02:42]

<crweb>

gremlin[it]: its standard linux startup. Busybox scans /etc/inittab at boot and start init

[02:42]

<gremlin[it]>

crweb, happy to know someone interest my work :P

<gremlin[it]>

yes i notice is about standard ... but i lack something

<gremlin[it]>

for example who start nanox ?

<crweb>

gremlin[it]: the init script system

<crweb>

/etc/init.d/rcS will call the startup script system /etc/init.d/rc

<crweb>

first with runlevel S

<crweb>

for startup, which runs all scripts in /etc/rcS.d

<crweb>

then it will call 5 (run level 5) which runs all scripts in /etc/rc5.d

<gremlin[it]>

there is a nanox script in /etc/init.d but no in /etc/rcS.d or /etc/rc5.d

<crweb>

oh, that would be deskmon

<crweb>

sorry

gremlin[it]: if you have time, there's a message in mailing list right now called "Build system setup" where you should really jump in and advertise your work on OE

<gremlin[it]>

deskmon script in /etc/init.d or deskmon binary application ?

gremlin[it]: are you on the mailing list, right ?

<crweb>

gremlin[it]: the binary starts nano-x and /usr/local/bin/desktop

<gremlin[it]>

problably no nerochiaro :(

<crweb>

gremlin[it]: its a monitor, it keeps them running at all times

<crweb>

so "crash" will restart them

<gremlin[it]>

ok maybe i understand now ...

[02:46]

<crweb>

they can be started with init.d/nano-X and init.d/desktop though

[02:46]

<gremlin[it]>

i moved some binary from /usr/bin to /usr/local/bin ... and if path is hardcoded all fail :S

gremlin[it]: it would be really nice if you could subscribe to the ML and post a message about the OE work. it will help push it forwards. groups.google.com

<crweb>

all our binary goes to /usr/local/bin

gremlin[it]: this is the specific post: preview.tinyurl.com

<crweb>

deskmon is made specificially for our setup

[02:48]

<gremlin[it]>

ok i'll try to copy all in the right place ...

[02:49]

<crweb>

gremlin[it]: you will also want to sync with svn for etc/init.d/rc

[02:49]

<crweb>

gremlin[it]: i just fixed large problem (syn in like 15 minutes, nerochiaro still has to ok patch)

[02:49]

<gremlin[it]>

any other suggestions about plugins and/or "special" library ?

[02:50]

<crweb>

nothing that i can remember off hand

[02:51]

<crweb>

i wrote all that startup script stuff like a year and a half ago

[02:52]

crweb: i did

<crweb>

so did i

<crweb>

might take a while for USA to sync

15 minutes IIRC

<crweb>

see how about that.

<crweb>

right even with mistake :)

maybe 10, i forgot

[02:54]

crweb: writing up big message about how/if to handle the format of the ext area

[02:55]

<crweb>

good. cause with OE, a lot will be changing

[02:55]

<crweb>

you'll want to format

[02:56]

<crweb>

(much like desktop version stuff)

[02:56]

<mp0>

nerochiaro, I see what you mean. All the fun stuff is packaged in .ko binaries.

[02:56]

crweb: i know your opinion, but we already saw it's more complex than that.

[02:57]

mp0: well, not all, but some of it yes. all the stuff you see in neuros-bsp/kernel/kos_64

[02:58]

<mp0>

what's the difference between kos and kos_64?

[02:58]

kos is actually obsolete. should be deleted really

[03:04]

<crweb>

gremlin[it]: so inittab is working correctly?

[03:04]

<mp0>

That'll make things a little tricky.

[03:04]

<gremlin[it]>

nerochiaro, done ... i reply also about packaging system

gremlin[it]: thanks !

<gremlin[it]>

yes .. the OE inittab work fine :) ... i copy someof the /etc/init.d/ stript from cramfs and worked ...

<crweb>

gremlin[it]: did you complete cooler convert to autoconf?

<gremlin[it]>

nerochiaro, i confirm that with OE number generator work :)

[03:06]

<gremlin[it]>

crweb, no ... i try to follow the faster way to have something barely working to show at fosdem :s

mp0: i see it can really be a problem. if you want, we can probably let you sign an NDA to see the code of these closed modules, and contribute the CC modifications to them.but they have to stay closed :(

<gremlin[it]>

nerochiaro, also using kernel from flash (so original one) ...

mp0: or you can write open versions of them, but they need to be compatible with the ones we use now

<crweb>

gremlin[it]: do you know what made number generator work?

gremlin[it]: sorry, i didn't understand

[03:10]

<gremlin[it]>

crweb, no i didn't investigate

[03:11]

<gremlin[it]>

nerochiaro, i try to boot with the kernel present in the flash of OSD ... that is the original neuros's one

<crweb>

gremlin[it]: any way i can get this working version from you so i can investigate and enable in current firmware?

<crweb>

just the rootfs would be ok

<mp0>

nerochiaro, I'll try the second option first... we'll see how little I have to rewrite to get what I need.

gremlin[it]: what is the problem with that ?

[03:14]

mp0: my point regarding that option was that if you only write it for your own use, then ok. but if you want to make it work with the rest of the system (so that neuros media framework is able to read video input) then you need to write it compatible with it

[03:14]

<mp0>

yes

[03:14]

mp0: but i'll tell you one thing: if you make it read the CC data in a re-usable way, we can take your code to read CC and add it to the neuros driver

<gremlin[it]>

crweb, mid term : as soon I'll publish some OE setup that work :)

<mp0>

okay

mp0: anyway, feel free to ask for info if you need help

<gremlin[it]>

crweb, short-term : i send you my OE image (less then 10MBytes) and you can try booting from CF as i do

<mp0>

alright, thank you for the info.

<crweb>

gremlin[it]: i got it to work. nvrmind

[03:16]

mp0: you're welcome

[03:17]

<gremlin[it]>

crweb, if u give me your email i can put the image on my web site

[03:18]

<crweb>

gremlin[it]: ok, tbruno at neuros dot us

[03:18]

<crweb>

gremlin[it]: i found the missing link already

[03:18]

* crweb is enabling random now

[03:19]

crweb: where does it fill entropy from ?

<crweb>

nerochiaro: I'mnot sure. seems to work now though

<crweb>

nerochiaro: before it would just be 0

would be nice to know, though

<crweb>

we can at least enable it

sure

random or urandom ?

<crweb>

right now, you can't even access it

<crweb>

id idn't try urandom. I'll make it now

urandom should be the one used by ssh, IIRC

<crweb>

its working

<crweb>

plenty of output i think

<crweb>

before it was NULL

cool

[03:22]

<crweb>

maybe we were trying old device nodes or something..

[03:25]

<crweb>

could possibly make someone happy

[03:26]

<gremlin[it]>

heheh :)

<crweb>

nerochiaro: how long do you let your tea bags in the drink?

crweb: depends on kind of tea

crweb: some need less than a minute, some up to three

<crweb>

so.. 45 mintues will do...

[03:28]

<crweb>

heh

[03:28]

crweb: sure, it will taste bitter as vengeance, though

[03:28]

<crweb>

aye..

[03:29]

it is also said you destroy most of its theine content if you do that (never experimentally confirmed it though)

[03:29]

<crweb>

i had jsut forgot about it

[03:29]

<crweb>

i usually go for 2-5

[03:30]

<crweb>

and yeah.. not good

[03:30]

i know, it happens to me a lot too. i usually just make another

[03:31]

you made me crave for a cup of Gao's tea. i'm going to setup the water.

[03:36]

<crweb>

i want some swoag tea

[03:36]

<gremlin[it]>

nerochiaro, i hear that if tea is infused for too long time it can be bad for health, is true?

[03:40]

<crweb>

wooo.. sugar doesn't even help the 45 min brew

[03:40]

gremlin[it]: maybe. from what i read, it gets high content of "tannin" substances if you leave it too long. in huge quantities tannins can be not so good for health. but i'm not a (bio)chemistry scientist ;)

[03:41]

gremlin[it]: en.wikipedia.org for the bad effects

[03:43]

<crweb>

With the exception of tea, long-term and/or excessive use of herbs containing high concentrations of tannins is not recommended.

[03:44]

<crweb>

i'm asuming this is because no trace tannin acid is found in teas

[03:45]

right, didn't read that part. interesting

[03:45]

long bitter brews are safe then ;) except for the vile taste, of course

[04:23]

<crweb>

nerochiaro: say you had a piece of code that "Has to output data"

[04:24]

<crweb>

and the data goes to stdout

[04:24]

<crweb>

and it printf(data) and then never exists

[04:24]

<crweb>

it only output 2 lines, which does not cause the buffer to flush

[04:25]

<crweb>

the App needs to phsyically flush its buffer before it goes into its infinite loop, yes? It is no one elses job to do so?

[04:25]

app does it

[04:26]

<crweb>

ok. aparently something went down between two devs and one says its not "Apps" job and "App" should not be changed.

[04:26]

<crweb>

but, i took care of it.

[04:30]

ok

[04:30]

<crweb>

something to do with led blinking during update or something

[04:39]

crweb: i don't know about that, sorry

[04:40]

<crweb>

well, incase you get blind sided ;)

[04:41]

ha! blindsided, me ? pfff

[04:41]

;)

[04:57]

gremlin[it]: still around ?

[04:58]

gremlin[it]: can you receive PM messages ?

[05:06]

<gremlin[it]>

here i'm ...

[05:06]

<gremlin[it]>

what's PM message ?

[05:07]

<gremlin[it]>

nerochiaro,

[05:24]

gremlin[it]: private message, in IRC

[05:24]

<gremlin[it]>

ah yes yes

[06:34]

<gremlin[it]>

crweb, where u live ?

[06:34]

<crweb>

I'm 6 hours south of Chicago IL, usa

[06:35]

<gremlin[it]>

6h south ! ... same level of philadelphia ? about

[06:36]

<crweb>

yeah probably pretty close

<crweb>

I'm about an hour from the southern border of Illinois

are you close to St.Louis, right ?

<crweb>

about 90 miles south east

<crweb>

no wait

<crweb>

thats parents

<crweb>

about 120 miles south east (but more south)

<crweb>

Carbondale IL

google maps says 105, and it's in a very roundabout way

<crweb>

eh, give or take

[06:39]

<gremlin[it]>

ohhh south ... was a my dream to visit St.Louis ... before katrina ... i like soul music !

[06:39]

<crweb>

katrina was about 15 hours south of St. Louis

[06:40]

wasn't it new orleans ?

[06:40]

<crweb>

New Orleans

[06:40]

<crweb>

yeah

[06:41]

<crweb>

st. louis is home of the gateway arch, and that one famous person in a plane that flew around the world or something

[06:42]

<gremlin[it]>

hahah my fault right !!!!

[06:43]

<gremlin[it]>

spirit of St.Louis ... probably first direct fly over Atlantic ocean ...

[06:43]

<crweb>

yeah thats it

[06:45]

<crweb>

new orleans is fully functioning now i hear

[06:47]

<gremlin[it]>

ah good ... i saw something on TV recently but said isn't as before ... still a lot of empty blocks ...

[07:07]

crweb: still there ?

[07:07]

<crweb>

yes

[07:07]

regarding that thing of not creating a window more than once, what about using deleteonclose and a QT guarded pointer ?

<crweb>

guarded pointers use templates. I agree with what he did

but would that work ?

the guarded pointer

<crweb>

yes

[07:09]

so why not use that ?

[07:09]

<crweb>

avoid templates at all costs. they are the c++ performance drop

[07:09]

so let's drop all QList

<crweb>

well

maps, hashes

hmm ?

<crweb>

theres to many features that depend on lists

<crweb>

thats actually worth it

[07:11]

<crweb>

this is trivial work, and even if used guarded pointer, the functions will look exactly the same

[07:11]

<crweb>

oh wait.. you mean use guarded pointer.. if ! deleted, show

[07:11]

<crweb>

instead of setting up a connection ?

[07:12]

with a guarded pointer, you don't have to connect to the event when the window goes down so you can NULL the pointer

[07:12]

that's actually good, if possible

[07:12]

<crweb>

i don't like it. but it would simplify the class logic

[07:13]

<crweb>

i really like having a open handler, and close handler for every window created.

[07:13]

<crweb>

in most cases you have to end up taking action on close() of a window anyway

when you don't, the guarded pointer avoids a lot of useless code

<crweb>

ok

plus, frankly, i'm not really all that convinced that templates == automatic perf drop

<crweb>

well, thats not what we told him to do

<crweb>

err sorry

<crweb>

mis read

[07:15]

<crweb>

and yes, it is

[07:15]

<crweb>

thats pretty common knowledge

[07:16]

<crweb>

though.. now that i think about it that could have been compile time

[07:16]

pointers to reliable articles that demonstrate this in scientific way ?

otherwise, it's witch lore

;)

<crweb>

well, hold on a sec. i have a few

<crweb>

they are in embedded c++ books

[07:21]

<crweb>

The features of C++ that are typically to expensive for embedded systems are templates, exceptions, and runtime type identification. All three of these negatively impact code size, and exceptions and runtime type identification also increase execution time.

[07:21]

<crweb>

page 261, Limiting the Impact of C++

[07:22]

<crweb>

Programming Embedded Systems with C and GNU Development Tools, O'Reilly, Michael Barr & Anthony Massa, 0-596-00983-6

[07:24]

<crweb>

www.embedded.com

[07:24]

<crweb>

* Templates are omitted.

[07:24]

<crweb>

embedded c++ compiler

[07:25]

code size, well ok. that's probably right.

[07:25]

<crweb>

code size does effect runtime exec

[07:26]

<crweb>

when it takes 2x steps to do a task

[07:26]

<crweb>

add that to the overhead of a guarded pointer

[07:27]

sure, but it's not like you're checking them in a tight loop. it's checkeing them once in a while. plus you can always check the real pointer directly, when you don't need the guard

[07:28]

all i'm saying it "should be avoided at all costs" was a bit brutal of a statement

[07:28]

<crweb>

ok. obviously we have to use templates for apsects like lists, iostream, etc

[07:29]

<crweb>

with our already loaded system, i don't want to add any more overhead, thats for sure

[07:29]

<crweb>

if he had done it more correctly, we could have eliminated the if all together

[07:31]

for example ?

[07:32]

<crweb>

connected it to a slot that run when THAT window was closed

[07:32]

<crweb>

no need to check, the only way it was run was for that window

[07:33]

i don't understand

[07:33]

<crweb>

he connected the destory to a generic destroy function

[07:34]

<crweb>

then he had to check to see which window was destroyed

[07:34]

<crweb>

if he had just connected to, mywindowXDestroyed(), he wouldn't have had to check to see which window was destroyed

oh, ok, but he still would've had to connect it

<crweb>

yes

which is an overhead just like the guard is. so really, no major loss

<crweb>

but the guard adds way more code

<crweb>

we just dont "see it"

we don't see the code to connect signals and other QT stuff either

[07:37]

<crweb>

yeah you do. its in the moc file

[07:37]

<crweb>

its just function pointers

[07:37]

you have a point

[07:41]

but anyway, for this case, it think we won't lose much with the guard

[07:41]

<crweb>

but you open a can of worms I close ;)

[07:41]

<crweb>

err closed

[07:42]

<crweb>

its ok. i agree with you completely. I just really don't want to start using templates unless we need to.

[07:42]

<crweb>

and as i said. most of the time you have to connect a closeWindow slot anyway. or you should rethink why you are opening a new window

[07:44]

sure

[07:44]

let's just prevent over-use, and we should be fine

[07:45]

<crweb>

what i def dont' want to see is guarded pointer AND connection

[07:45]

<crweb>

we'll have to watch out for that

[07:45]

ok

[08:32]

* MattJ wonders if he should try AZ again today

MattJ: everyday is a new day for AZ :P

<MattJ>

So what ejding today? 2j? :)

<MattJ>

*ending

<MattJ>

It complained the package was corrupt, yesterday

[08:42]

add one letter

[08:42]

to yesterday

[08:42]

it shouldn't have complained. something was wrong, definitely. please try again today

[08:43]

and if corrupt again, please let me know

[08:43]

<MattJ>

Ok

[10:01]

<MattJ>

"Unable to upgrade" "Update package corrupted!"

[10:01]

where are you putting the package ?

[10:01]

card ?

[10:02]

<MattJ>

CF card

[10:02]

isntalling AZ to CF card from CF card ? seems wrong

do you have other cards ?

<MattJ>

To quote the wiki: "Load it onto a USB, SD or CF card, and plug that card into your Neuros OSD."

<MattJ>

I don't have an SD card free