<hipboi>
the watermark is "For cubieboard only" :)
<hno>
Then it's only the CAN controller missing of the features people ask for I think.
<hno>
plus DRAM controller, but we have that somewhat under control.
<hno>
and USB OTG controller.... keep forgetting that one. But we have all registers, working code and it's origin have been identified so it's not so situaiton is not so bad there.
shineworld has joined #linux-sunxi
<oliv3r>
oh that's awesome
<oliv3r>
nand controller docs sorta too still, I think ther where still things missing?
rellla has joined #linux-sunxi
<oliv3r>
and A20 docs :D
<oliv3r>
i wonder how we connect the Descrambler to a (soft) cam + card (reader)
<oliv3r>
TSC STatus Register is really interesting
<oliv3r>
31:0 / / /
<oliv3r>
hipboi: do you know if the 'Touch Screen Temperature Sensor' actually works? I tried poked some of the registers from the 'temperature debug' bit from the code, but I get zero result. The register is 'removed' from the datasheet aswell. Is it possible that it simply does not work?
<rm>
> "You can even boot on a device with AXP without AXP support, right?"
<rm>
just a minor typo somewhere in there
<oliv3r>
er
<oliv3r>
i belive you can boot/use an A?? without an AXP
<oliv3r>
or PMU rather
<oliv3r>
the PMU is optional, though strongly recommended
<rm>
mkay, just the phrase structure made me think that "can't even boot" was intended
<rm>
if you in fact CAN, then I'd expect something like "You can boot on a device with AXP even without AXP support"
unioah1 has left #linux-sunxi [#linux-sunxi]
bsdfox has quit [Ping timeout: 252 seconds]
hansg has joined #linux-sunxi
eebrah has joined #linux-sunxi
eebrah has quit [Quit: Leaving]
bsdfox has joined #linux-sunxi
bsdfox has quit [Changing host]
bsdfox has joined #linux-sunxi
bsdfox has quit [Ping timeout: 248 seconds]
bsdfox has joined #linux-sunxi
bsdfox has quit [Changing host]
bsdfox has joined #linux-sunxi
eebrah has joined #linux-sunxi
_BJfreeman has joined #linux-sunxi
_BJfreeman is now known as BJfreeman
n01_ has joined #linux-sunxi
<mnemoc>
hi, don't know if he mentioned here alrady, but master lkcl got the GPL A20/nand code from allwinner
<oliv3r>
oh sweet
<oliv3r>
cedarX?
<mnemoc>
i just asking him about it ;-)
<mnemoc>
been ffmpeg based, it's lgpl code and so...
<oliv3r>
(l)GPL based; I think if you use any of the optional parts, that are GPLed, the entire thing becomes GPL based
<oliv3r>
can we know for certain, that they didn't do that? :p
<oliv3r>
what if ffmpeg_huffman thingers IS gpl licensed
<mnemoc>
been in a same .so with (l)gpl code turns every bit into (l)gpl
<oliv3r>
in any case, libcedarX is a hodgepodge of all sorts of code
<oliv3r>
tons of copy pasted works
<oliv3r>
i wonder what its real legal status is :p
<mnemoc>
the "l" exception allows you to dynamiclly use that .so from closed code, but been in the same .so/.a the (l)gpl "disease" immediately spreads across every bit of it
<oliv3r>
yeah but they didn't use the .so
<oliv3r>
they copy/pasted code from ffmpeg
<mnemoc>
oliv3r: what is libve? .so
<oliv3r>
ok i don't follow then :)
<mnemoc>
this .so/.a got code ill with gpl disease, therefor the whole blob is gpl
<mnemoc>
it's not a joke when they call it a cancer
<mnemoc>
the gpl license has that effect
<oliv3r>
i call it blessing :p
<mnemoc>
:p
<oliv3r>
anyway, yeah, they used ffmpeg bits copy/pasted. I think the ffmpeg license states, that if you 'enable certain GPL features, the ffmpeg .so becomes gpl'
<oliv3r>
but I don't know what license the huffman functions use that where copy/pasted
<oliv3r>
if it's lgpl or pure gpl
<oliv3r>
anyway, if you include said code in your blob, the blob becomes (l)gpl and thus should be open
<mnemoc>
for our purpose it doesn't matter. as the libve blob includes ffpmeg code, the result is (l)gpl code
<oliv3r>
if they would have linked to it, it would have been 'ok'?
<oliv3r>
ah, yes
<oliv3r>
so it's a violating blob
<mnemoc>
yes
<mnemoc>
every bit of it
<oliv3r>
i herby formally request source for the blob!
<mnemoc>
anyone how received libve directly from allwinner can demand for it's source code
<mnemoc>
not my case, but many here do
<rellla>
mnemoc: libvecore.so, yes
<oliv3r>
mnemoc: there's been discussions going on about that on gpl-violations ml
<oliv3r>
and i think you are wrong
<oliv3r>
I have received the .so (via sunxi for example), but AW is still repsonsible for the source code
<oliv3r>
they SHOULD have added the source/written offer with the blob
<mnemoc>
gpl-violations is a waste of time. but if luke managed to find someone who cares about compliance, the right door is now open
<oliv3r>
and sunxi should have forwarded said written offer/source
<mnemoc>
oliv3r: you can only sue who directly gave you the blob
<oliv3r>
it wasn't about allwinner at all
<mnemoc>
oliv3r: and so it becomes an impossible to trace chain
<oliv3r>
android kernel 'hackerboys' who violate the kernel left and right
<mnemoc>
but rellla and others have received libve (with lgpl cells) directly from allwinnertech.com
<oliv3r>
lkcl linked an interesting article a few days ago, huge read; but it said that Allwinner socs are only used for 20% in chinese market, 80% rest of the world
<oliv3r>
since chinese buy ipads/samsungs
<mnemoc>
VERY few
<oliv3r>
aye, so in theory, rella can sue allwinner
<hipboi>
that's true
<jinzo>
oliv3r, afaik chinese don't buy ipads
<oliv3r>
hipboi: hi hipboi
<hipboi>
chinese buy ipad
<mnemoc>
the cost of those devices is FAR FAR away from the income of the chinese masses
<jinzo>
hipboi, realy?
<oliv3r>
jinzo: i dunno, it was a long read; but these low budget tablets, are being sold to a lot of poor countries; chinese want 'high quality' tablets, e.g. samsung/ipads
tzafrir has quit [Ping timeout: 264 seconds]
<rellla>
oliv3r: but only in theory ;)
<jinzo>
oliv3r, samsung is a different beast all together
<jinzo>
but the last reports I read, apple didn't have any meaningfull success in china
<jinzo>
but what do I know.
<hipboi>
iphone is everywhere in China
<jinzo>
but looks like the sales doubled since last year so.. :D
<mnemoc>
even a minor % of chinese or indian population is still a huge number of people
<jinzo>
hipboi, interesting - the perception of us, not in china, is that your smartphones are far superior (with TV for eample)
<mnemoc>
hipboi: real iphones or knochoffs ?
<oliv3r>
those who can afford, want expensive stuff; those who can't, don't buy cheap tablets either :p
<jinzo>
and why would someone buy an iPhone when I can't wait to get my hands on a Xiaomi or Oppo...
<mnemoc>
my wife and daughter's friend really envy their jiayu g3 :p
<mnemoc>
friends*
<hipboi>
real ones, mostly common is iphone4/s
<oliv3r>
Xiaomi (i think) is slowly coming to our country aswell, with 250-300 Euro phones
<oliv3r>
nice mid-range price, with high-end phones
<hipboi>
yes
<hipboi>
xiaomi is very successful
<oliv3r>
hipboi: quick Q; can you confirm/deny, that the temperature sensor embedded in the touchscreen controller is actually non-functional?
<jinzo>
Oppo opened EU shop previous week to.
<rm>
Jiayu <3
<hipboi>
oliv3r, never heard of that
<oliv3r>
the specs of the Xiaomi phones look really good
<oliv3r>
i tried poking the registers; but while adc works, temperature sensor says 0. but the A10 usermanaul has it entirely removed; so it could be its actually non-functional
<oliv3r>
how is it debatable? bad reputation in china?
<hipboi>
oliv3r, are you sure temp means temperature
<oliv3r>
i think that copy/paste may be 1 page too far
<oliv3r>
but it clearly says 'temperature data'
<jinzo>
I'm trying to get into the chinese state of mind, with so many quality (and cheaper) manufacturers of smartphones - why would you buy an iPhone...
<jinzo>
damn.
<oliv3r>
bah, firefox's pdf javascript renderer still sometimes lags/crashes FF :(
<mnemoc>
jinzo: status, "feel like an American"
<hipboi>
oliv3r, which page
<oliv3r>
273, the register index
<mnemoc>
same reason most people buys Apple products. to show others they are successful
<oliv3r>
offset 0x1c
<oliv3r>
that offset is actually wrong in the datasheet, the code has it properly; but no documentation at all
<jinzo>
mnemoc, but I would hope that only works in America :/ Because I respect someone with a let's say Xiaomi or Oppo way more
<jinzo>
than a iPhone user
<jinzo>
but then again, I dislike people that like to show off.
<oliv3r>
actually, I do find apple products appealing
<oliv3r>
i don't use one due to the software/license
* mnemoc
likes squared phone like jinzo's picture or sony or jiayu g3(s)
<mnemoc>
phones*
<jinzo>
yeah those are good to, I had just opened this one
<oliv3r>
I even appreciate the design of the apps apple uses, even though it actually stupid in a sense. (paper background in the notes.app for example)
<Turl>
jinzo: xiaomi is as closed as iphone
<jinzo>
regarding laptops: Samsung Series 9 for eample...
<jinzo>
I'm looking into learning mandarin, but it's proving to be _very_ challenging. But baby steps.
<jinzo>
The stuff people do on diysmt.com is beyond awesome.
<Turl>
jinzo: google translate doesn't help much there either :p
<Turl>
hipboi: how is A20 doing kernel wise?
<hipboi>
Turl, kernel wise?
<mnemoc>
Turl: lkcl booted his eoma68-a20 with free kernel already
<Turl>
mnemoc: nice
<hipboi>
Turl, i just used translation software, i am not sure what you mean
<hramrach_>
hello
drachensun has quit [Ping timeout: 276 seconds]
<mnemoc>
hipboi: foo-wise = regarding to foo
<Turl>
hipboi: is the kernel 'good enough' or is it too broken / missing pieces?
<hipboi>
we have cubieboard2 with ubuntu desktop ready to ship
<Turl>
great then, that answers my question :)
<hramrach_>
that's nice
<hipboi>
mnemoc, thank you
drachensun has joined #linux-sunxi
<mnemoc>
the a20-sdk (and -dev branch) have a major gpl-violation. the nand driver, but lkcl received the corresponding GPL code
<mnemoc>
bbl
<hramrach_>
mnemoc: what iPod had whic most other players did not - it remebered where you stopped playing. so if you ahd like audiobook, turned off the player in the middle of a chapter and turned it back on it started where you were
<hramrach_>
very hard to find manufacturers that make firmware with such attention to detail
<hramrach_>
and very hard to use firmware from manufacturers wh odon.t even if the hardware is good
<oliv3r>
jinzo: i've started to use 'memrise' to learn mandarin
<oliv3r>
but yeah its hard
<Turl>
hramrach_: most people don't care for such details though, they just turned their mp3 players on and went running or took the bus
<hramrach_>
I do because when I wen off the bus in the middle of the chapter I wanted to stomp on the player next time I wen on the bus :p
<Turl>
personally I hate audiobooks
<Turl>
because my eyesight is not occupied so I cannot concentrate
<hramrach_>
I can't read in a bus so it's audiobook or nothing
<oliv3r>
hipboi: did you see what I ment with the temperature register? Or is it just a bad name and meant something else?
<hipboi>
oliv3r, i looked at the history of the file
<hramrach_>
plus I sometimes don't have choice of format. When it comes as audiobook you listen to it or trash it
<Turl>
hramrach_: I've never seen audiobook only material
<Turl>
the other thing about them is that they're slooooow
<hramrach_>
It probably exists in other form, yes. But local availability is another matter
<hramrach_>
I readl slow too so there is not that much difference
<oliv3r>
i think, they maybe use the ADC for the temp controller
<hipboi>
the guess is the register is used in the older chips
tinti_ has joined #linux-sunxi
<hipboi>
removed in the later chips but the code remains
<oliv3r>
probably
<oliv3r>
maybe sun3i had it? or, its simply a broken feature and they didn't fix it in silicon
<oliv3r>
was just curious if you knew more about it; since it did seem kinda neat, to have a temperature sensor in the SOC itself
<Turl>
it's weird that AW has no temp sensors
<hipboi>
hmm...
<oliv3r>
it does, but either not functional or partially removed/castrated :p
<oliv3r>
it's equally weird, that there's 3 places where the platform is stored (sramc, sid, cedarx etc)
<Turl>
all big name SoCs have for throttling, esp dual and quad coreones
<oliv3r>
and then there's the CPUinfo byte, that's useless
<Turl>
hipboi: does cubieboardii have a boot select jumper?
<hipboi>
no
<oliv3r>
Turl: same board as a`10, 0909 iirc
<oliv3r>
boot select is a31 only for now
<hramrach_>
Turl: interesting
<hipboi>
exactly the same with current pcb
<Turl>
ah, nevermind, a31 is the one with the different boot
<oliv3r>
most of that info up on BROM page
<Turl>
not a20
<hipboi>
a20 boot is the same as a10
<hipboi>
after we use the 5000 pcbs
<hipboi>
there will be another revision
<oliv3r>
with gmac!
<oliv3r>
i read; very cool
<oliv3r>
will you bring out more pins in the next rev?
<hipboi>
gmac is on another product
<hipboi>
cubie mini server
<oliv3r>
ah :(
<oliv3r>
bit, i read you wanted to do 2 GB cubie 2
<hipboi>
yes
<oliv3r>
olimex however said, that 2gb wasn't possible on a1* due to bugged memory controller
<oliv3r>
is the memory controller fixed in a20 and can it REALLY do 2gb?
<hipboi>
2g ram, gmac, sata, wifi/bt, vga
<Turl>
hipboi: does sata work on current a20 kernel?
<hipboi>
2g ram is possible
<hipboi>
yes, sata works, ported from a10
<oliv3r>
well yeah, 2gb was possible for a10; but olimex never got it to work
<hramrach_>
but they did not really try very hard
<hramrach_>
or so it seems from their blog
<oliv3r>
possible
<hramrach_>
oliv3r: can you comfirm that making the android config as modular as the normal linux config does not break something on andriod?
tzafrir has joined #linux-sunxi
<hramrach_>
or anyone running android
<oliv3r>
hramrach_: que?
<Turl>
hramrach_: remember that android doesn't have module autoloading btw
<hramrach_>
yes, that's probably why so much stuff is built-in
<Turl>
and the system partition is limited in size, so the least cruft modules you can ship, the better
<oliv3r>
how did we get to this subject?
<Turl>
oliv3r: 07:44 hramrach_> oliv3r: can you comfirm that making the android config as modular as the normal linux config does not break something on andriod?
<mnemoc>
hipboi: Tsvetan refused to make a A20/2GB/gmii variant of his a10-olinuxino-micro. what's your plan regarding the a20 "cubiebox"?
<mnemoc>
hipboi: olimex will only do a10/512/mii and a20/1GB/mii
<hramrach_>
oliv3r: it was you talkin about the configs on the ML, rihgt?
<hipboi>
mnemoc, our hardware design is done
<mnemoc>
-v
<hipboi>
mnemoc, now wait the pcb back
<hramrach_>
how limited is the system partition?
<oliv3r>
Turl: yeah where did that come from? :p
<oliv3r>
aka; what did I do
<hramrach_>
and how do you pick what to include?
<mnemoc>
hipboi: can you tell a bit about it's specs?
<hipboi>
mnemoc, exactly the cubiebox you want
<oliv3r>
hipboi: \o/
<Turl>
hipboi: with IR? :P
<oliv3r>
how many pins are ont he headers? what's the header spacing?
<hramrach_>
there are people coming all the time saying they connected this weird device to this board and it does not work /o\
<oliv3r>
Turl: and with both SVHS + Composite out :p and maybe RGB out too :p (VGA is rgb + sync iirc)
<hipboi>
Turl, with ir, nowadays, smart phones have ir
<oliv3r>
hramrach_: yeah but why did you mention this now, about me?
<oliv3r>
nowadays? even dumb phones used to have IR :)
<Turl>
hipboi: indeed, samsung added it to some phones
<hipboi>
oliv3r, the header will be 2.54mm
<oliv3r>
more pins?
<hramrach_>
because you were asking why the firewall options are built-in
<hipboi>
oliv3r, less
<oliv3r>
hramrach_: ahh, yeah of course! the review :p
<oliv3r>
hipboi: oh :(
<Turl>
oliv3r: bigger pin, less qty
<oliv3r>
yeah but we want more pins brought out :p
<hipboi>
oliv3r, mainly for those who want to put in a case with 2.5 hd
<hramrach_>
hedgehog board
<oliv3r>
true, that's the first thing I tried, to match the screwhols ontop of a 2.5 HD
<Turl>
hipboi: how big will it be, compared to a mele?
<oliv3r>
did you flip the sata connector? so that i's facing sideways instead of upwards?
<mnemoc>
Turl: he said cubiebox, not cubieboard. that implies larger PCB
<hipboi>
hramrach_, a toothpick saw a hedgehog, he shouted bus!
<oliv3r>
hramrach_: well, I first wanted to do a full review, but then i noticed you pretty much included nearly everything. But yeah, some things may require to be build in _for android_ if auto loading doesn't work right
<hipboi>
Turl, smaller than mele
<oliv3r>
poor hipboi, over run with questions :D
<hipboi>
a little bigger than 2.5hd
<Turl>
oliv3r: :p
<mnemoc>
open schematics? :p
<Turl>
hipboi: :)
<mnemoc>
layout?
<hipboi>
open schematics
<mnemoc>
*g* /me wants to spoil the surprise :p
<hramrach_>
hipboi: it would be nice if the layout was such that you could mount it on top of a 2.5" hdd. Matching holes, matching SATA port position
<mnemoc>
iirc the goal was VESA mounting
<jinzo>
hramrach_, afaik the board is basicly the same as cubieboard (1) - not sure tho.
<oliv3r>
'server' formfactor, cubieserver; matches a 2.5 HDD, with SATA ports facing the same sides (flat)
<oliv3r>
'mediabox' formfactor, cubiebox; tons of audio/video/dvb connectivity (both IR, both PWM's brought out)
<oliv3r>
some gpio's for buttons
<mnemoc>
oliv3r: more variants, higher cost
<mnemoc>
specially if you use different PCBs
<oliv3r>
and the devboard, with tons and tons of pins brought out :)
<hramrach_>
yes, you can easily make 'mediaserver' form factor ;-)
<oliv3r>
mediabox = vesa formfactor :p
<oliv3r>
well there's 2 formfactors allready in mind anyway?
<oliv3r>
cubieboardII (new rev)
<oliv3r>
and cubiebox
<oliv3r>
so for 'version 3'
<hramrach_>
you can add an adaptor from 2.5" disk holes to VESA holes, especially if there is a case
<oliv3r>
those are formfactors to concider, server, media (vesa) developer
<mnemoc>
oliv3r: he said the design for the PCB of the cubiebox-a20 is done, and waiting for the factory
bsdfox has quit [Ping timeout: 260 seconds]
<oliv3r>
well yeah, i ment for 'next rev'
<mnemoc>
too later for new suggestions :|
<oliv3r>
think ahead!
<hramrach_>
there is 75mm and 100mm VESA anyway so adaptors are likely going to be required
<hipboi>
i think VESA can be archived on the case
<mnemoc>
hramrach_: 12:55:28 < hipboi> mnemoc, oops, VESA is forgotten
<mnemoc>
hipboi: indeed
<hramrach_>
you want VESA mounting for encased boards anyway so case it the place for VESA
<mnemoc>
hipboi: want about headers alignment? will it match a 0.1" grid?
<hipboi>
mnemoc, yes, 2.54mm
<hramrach_>
that sounds good
<oliv3r>
yeah i read someone who had issues with grid allignment
<hramrach_>
the 2mm headers are space efficient but way much harder to get connectors for them
<mnemoc>
hipboi: grid alignment is like sticking the *full* board in a breadboard and getting headers properly inserted
<hipboi>
mnemoc, not considered that..
<hramrach_>
breadboard ..
<hipboi>
but maybe it happens to be
<hramrach_>
it always makes me think of bakery
<mnemoc>
:)
<mnemoc>
hipboi: it's one of those details that makes EEs doing extensions happy
<oliv3r>
yeah, matching it to breadboards makes a lot easier; even though it's all 2.54 mm; I much better liked the cubieboard 1.0 headers; but hard to find parts for
<mnemoc>
hipboi: but I guess you'll sell a custom breadboard too ;-) a10's looks sexy. and adding VGA was a great idea
<mnemoc>
hipboi: LiPo connector?
<hipboi>
mnemoc, yes considered
<mnemoc>
\o/
<hipboi>
mnemoc, even consider you use the mini server on the go
<mnemoc>
very very cool
<oliv3r>
so you have a lipo and the server keeps 'running'
<mnemoc>
even a humble LiPo battery can do a very nice UPS for an arm device
<oliv3r>
well it'll protect you from short power failures
<oliv3r>
which is awesome i think
<mnemoc>
and the AXP give you that feature for free
<oliv3r>
hipboi: btw, a13 user manual (page 263) also mentions temperature measurement; i'll play with it some more, maybe it requires more to be enabled
<oliv3r>
mnemoc: exactly
<mnemoc>
hipboi: one last feature question, battery for the RTC?
<hipboi>
mnemoc, considered
<mnemoc>
:)
* mnemoc
eager to see the detailed layout and schematics :p
<Turl>
hipboi: NAND? :)
<mnemoc>
eMMC!
* mnemoc
shuts up
<oliv3r>
mnemoc: who needs eMMC :p
<Turl>
mnemoc++
<hipboi>
Turl, not NAND, and not eMMC
<Turl>
but it's expensive
<mnemoc>
hipboi: dual uSD?
<Turl>
hipboi: dual SD slot!
<Turl>
:d
<Turl>
:D*
<hipboi>
it's uSD interface
<mnemoc>
only one?
<hipboi>
sdio interface
<hipboi>
but the nand chip form factor
<mnemoc>
separated system and user storage is kind of important. even if both are SD cards
<hipboi>
it's easy to install system if you boot from an external uSD card
<hramrach_>
hipboi: so no jtag/uart over sdio?
<hipboi>
hramrach_, on board jtag
<hramrach_>
nice
<hipboi>
dinner time, mummy is calling
<hipboi>
bbl
<Turl>
hipboi: looks like emmc
<mnemoc>
<3
<hramrach_>
still the attractive part of uSD slot is that you can easily replace the card when it wears off
<oliv3r>
bbl hipboi enjoy your meal
<oliv3r>
downside of having emmc (or tSD) is that it doesn't have wearlevl controller, its exported as block device, so you can potentially accidentally very easily install some journaled FS and it'll all die fast
tinti has quit [Ping timeout: 245 seconds]
<Turl>
oliv3r: the page says the chip handles wear leveling
<oliv3r>
the page is in chinese :(
<hramrach_>
SD cards tend to have some wear levelling
<Turl>
I've had good experiences with tuxfamily hosting, I'd give them a try
<oliv3r>
oh! tom mailed that cubieboardii is shipping today!
<oliv3r>
Turl: how do you load the data into the ram!
<ssvb>
even ignoring the non-free infrastructure aspect, google groups web interface (when used as a ML archive) is horribly slow and difficult to use
<libv>
ssvb: luckily, noone except for maybe mnemoc really is forced to use it :)
<libv>
the joys of being the projects administrator :)
<mnemoc>
*g*
<Turl>
oliv3r: tftp
<oliv3r>
while I fully agree with his reasoning; 'if they would like to move to free-software-hosted infrastructure' is kinda a bitchslap. Weren't we 'refused' earlier?
<oliv3r>
Turl: my PHY is broken! :(
<oliv3r>
Turl: but yeah, how do you do tftp without mmc/nand?
<Turl>
oliv3r: uboot can tftp
<oliv3r>
Turl: the BROM doesn't do tftp
<oliv3r>
Turl: how do you load uboot without nand/mmc :p
<ssvb>
libv: when I need to fish out a link to an old post in the ML to give it to somebody, I feel the pain
<hramrach_>
you can send copies to mail-archive or such
<libv>
ssvb: can gmame or whatever not be used on it?
<ssvb>
hramrach_: yes, gmane now mirrors the ML archive since some time ago
<hramrach_>
nice
<ssvb>
but unfortunately not the older posts
<oliv3r>
so I wonder who's lkcl 'free ML infrastructure' would host us, maybe now, but earlier we had to 'sort ourselves out'
<hramrach_>
hmm, LP does not seem to have mailing lists. sucks
<ssvb>
mnemoc: as the linux-sunxi boss, probably it would be a good idea if you contacted gmane folks yourself? :)
<oliv3r>
i connect to gmane via my newsreader :p
<Turl>
oliv3r: I just tag the emails and read in thunderbird :p
<oliv3r>
mnemoc: !! see, your the linux-sunxi boss; TOLD YOU!
<mnemoc>
o_O
<mnemoc>
i'm noone's boss
<oliv3r>
Turl: linux-sunxi I am subscribed; eoma86 i read via gmane
<oliv3r>
sometimes :p
<oliv3r>
mnemoc: fearless hero-leader
<Turl>
mnemoc: you were just given the benevolent dictator title, take it! :p
<mnemoc>
doh
<hramrach_>
Turl: that's where I started. It does not show the email I am looking for or I cannot find it so I use the search and then I get to the no-scroll search results
<mnemoc>
but I can't export the mbox... my mail is hosted by G the "linux-sunxi" label isn't as clean as it needs to be for properly exporting a ML
<mnemoc>
but any early subscriber with not-gmail-hosted mail can do it
<n01>
please, if you change ml address notify all the group using the current ml :)
<oliv3r>
not me, also i removed a whole lot
<hipboi>
shineworld, 34 boards in total
<mnemoc>
n01: if we move the ML we can inject all current subscribers
<Turl>
mnemoc: you can make a filter based on some header that indicates it's linux-sunxi list and apply it to old email to make a new tag
<oliv3r>
hipboi: that's reasonable; if all those 34 post patches to the ML; we should be progressing much better :)
<n01>
ok :)
<hramrach_>
except it maches private replies too
<mnemoc>
Turl: I can't, but I'm not the only with access
<shineworld>
hipboi: thanks for sharing
<hramrach_>
there is a way - download from G using imap and match locally
<hramrach_>
then you have as precis matching as possible
<mnemoc>
Turl: my problem isn't technical. it's time.
<mnemoc>
but it's simpler to solve by someone using a normal imap mail account than by someone who's imap is a lame mapping of tags and gmail threads
<Turl>
mnemoc: well, I use gmail too, and wasn't subscribed since the early start iirc
<hramrach_>
I found my machine does not have enough memory to mirror gmail using imap :/
<Turl>
hramrach_: huh?
<mnemoc>
there is a very nice lua library for scripting imap
<hramrach_>
started offlineimap and after some time I got oom killer running loos
<rz2k>
regarding drama at our ML/armnetbook ML, another move will kill linux-sunxi completely
<Turl>
oliv3r: consider :p
<rz2k>
user-side kill, obviously
<rz2k>
we barely managed to convince people to stop using github issues
<Turl>
I agree that moving with no benefit is on-purpose pain
<Turl>
oliv3r: well, he's saying that if you unbind the device and then use that function stuff will explode basically
<Turl>
and that your driver doesn't support multiple SIDs
<mnemoc>
rz2k: and expect specific-product-oriented lists help users of other products/boards/devices?
<mnemoc>
arm-netbook is about eoma, cubieboard is about cubieboard, olimex is about olinuxino. all of which are $-oriented
<oliv3r>
rz2k: theoretically, the move could be made without anybody noticing, if users can be moved over
<oliv3r>
Turl: well supporting multiple sid's is silly, but i don't mind working that in; but unbinding the devicetree, how does that make the function blow up?
<rz2k>
mnemoc: I didnt want to say that openly, but I am agree 100%
<oliv3r>
i mean, if we remove the create_bin_sysfs() thing; the driver will still work
<oliv3r>
Turl: so i don't understand the how (i know, my lack of knowledge, I'll admit that immediatly)
<rz2k>
it also disgusts me to see when you come to people for help, recieve help. and then publicly trashtalk them.
<rz2k>
because the people are "non-free"
<oliv3r>
making the pointer NULL;i'll agree quickly with that and change it. so on probe, we check for if (p_sid != NULL) {go};
* jinzo
looks to find said drama.
<oliv3r>
i think lkcl is/was a little upset that linux-sunxi brought to live 'without him knowning'. I can understand why he feels that way, he put a lot of work and effort into rhombus-tech /A10; then again, this i just a community which started up around A10, which didn't even have to know about RT if you think abou it
<oliv3r>
jinzo: arm-netbook ML
<jinzo>
Yes I+m subscribed but I missed it.
<Turl>
oliv3r: yeah, looks like nitpicking mostly when you consider you only have 1 SID on the SOC
<oliv3r>
also, I find that it wasn't so nice to say 'oh no, i don't read that ML, and haven't since ever because it's non-free hosted'. While the reasoning is very fine, he could have said, months ago 'hey, what you guys do is cool, but want your list hosted somewhere with a free backend?'
<Turl>
oliv3r: on unbind the memory would be unmapped I hope, so your pointer would beinvalid and you need to NULL it and possibly add a check
<oliv3r>
Turl: yeah, but I don't mind building in fixes, if I understand how :p
<Turl>
oliv3r: you're using devm_* so I think it unmaps on remove
<focus>
4 GPL'd KiCAD boards released for A10 CPU system development
<oliv3r>
Turl: well i thought that exiting the module would do a lot of unmapping cleanup; i thought that was the purpouse of platform_get
<oliv3r>
Turl: yeah, but I understood that using devm_ makes that everything gets cleaned up nicely on unload
<Turl>
oliv3r: yeah, memory mapping wise etc
<oliv3r>
Turl: so what is left bound?
<Turl>
oliv3r: but it won't NULL your pointer for you
<oliv3r>
i do the cleanup of the sysfs entry
<oliv3r>
ah ok
<oliv3r>
fair nuff
<Turl>
oliv3r: and sunxi:sid_read_byte is not static, so it can be used on other kernel places
<oliv3r>
ok; if you don't mind helping me understand one more thing, i'm a userspace dev mostly so I think in the terms of, 1 memory space per process
<Turl>
and therefore may hit the bad pointer
<oliv3r>
Turl: no, it is now, but it can be made unstatic and EXPORT_SYMBOLed, i took that into account
<oliv3r>
is it non static? ohh that was dumb
<Turl>
at least on that patch it's not static
<oliv3r>
i had it exported for a while, tried to set the serial to SID: but decided that was dumb, but forgot to add static
<oliv3r>
(again)
<oliv3r>
well i want it to work 'always'
<oliv3r>
as long as the driver is loaded of course
<Turl>
the null check won't hurt I guess in any case
<oliv3r>
so if the module is unloaded, is that function still available to the kernel?
<Turl>
no
<rz2k>
oliv3r: I think what makes him upset is not having linux-sunxi under his control, I am sad to say this, but thats the only reason.
<Turl>
but it is if it's unbinded
<jinzo>
that thread is pure gold.
<oliv3r>
Turl: so how do you unbind it, but keep the driver loaded? (I'll add the null check absolutly)
<oliv3r>
rz2k: purely technically speaking, it's good that its seperated. What if eoma continues with Freescale's IMX cpu, and not cares for the A* for many years. What's the point of having it there?
<Turl>
oliv3r: $ find /sys -name unbind
<oliv3r>
but even so, lkcl is cool and i like him
<Turl>
you can echo a device id or stuff there to unbind iirc
<oliv3r>
Turl: ok, so that 'unmounts'? or runs .remove()?
<rz2k>
yep, i agree, hope this crap will sort itself out
<rz2k>
because, as I said, another move will be absolute disaster
<jinzo>
it's nonsense.
<oliv3r>
well, if we do use a new server/service; i think only dev@linux-sunxi.org as a start; move our regular devs over, and slowly the rest
* rz2k
goes back to adding tons of debug messages to mtd-sunxi-nand
<oliv3r>
rz2k: good luck :(
<rz2k>
thanks, i will need it
<Turl>
oliv3r: it runs remove and triggers the devm stuff, but you can bind it again later on I think
<jinzo>
not worth it imo, because realy - if we look at it as luke does, I don't think we can use the internet (I doubt it anyone can guarantee you the source to all the routers/other networking)
<Turl>
oliv3r: ask mripard_ for the fine details :P
<oliv3r>
Turl: he'll just yell at me again :(
<Turl>
jinzo: have a look at the gnu link I dropped there
<oliv3r>
Turl: but ok, if it calls remove, then i understand, so the function should check the validity of the pointer
<jinzo>
Turl, I did, and I know, and lukes reasoning baffles me.
<oliv3r>
Turl: ok then what I was going to ask before, that pointer, is that what I check on .probe() if i want to know if this driver is allready loaded?
<Turl>
oliv3r: btw, please add a small changelog when you send a new version of the series, makes finding changes easier
<jinzo>
I don't think he is capable of completing any projects in reasonable time if he's picking fight like that
<oliv3r>
'Changes from'
<Turl>
ah, it's on the topic letter
<Turl>
s/topic/cover/
* Turl
should actually read :)
<oliv3r>
jinzo: it's a silly fight to pick. While I agree with his reasoning and he does have a point, it's not affecting 'freedom' per GPL. Because then, he cannot use anything. CLothes? made witha machine running non-gpl software, shipped with non-gpl UPS etc. Car? Runs non-GPL software, train? ... etc
<oliv3r>
Turl: ok so, if i check the pointer in _probe(); if it's null, go; if its not null, driver allready loaded?
<Turl>
oliv3r: your driver doesn't handle > SID, so yeah, sth like it
<jinzo>
oliv3r, +1
<Turl>
sigh numlock
<Turl>
>1* :p
<Turl>
I'm not sure what's the right error to return then though
<Turl>
possibly ENODEV, but I'm plain guessing
<Turl>
mripard_: ^
<oliv3r>
jinzo: so while I'm a 100% freedom guy, and I think RMS is a hero and he is right in every way
<oliv3r>
You can't live your life 100% that way, not even RMS can.
<Turl>
RMS makes a difference between some kinds of firmwares and actual software though
<oliv3r>
Turl: does mripard_ i2c driver handle multple version? I think so, since there's 3-4 I2C devices, right?
<oliv3r>
Turl: that line is getting thinner and thinner though
<oliv3r>
it's somewhat 'old' reasoning, but yeah
<oliv3r>
he says, some devices require firmware to be loaded, and it doesn't affect your freedom
<Turl>
he uses kind of duck typing for it
<oliv3r>
but that's not entirely true anymore. A lot of devices are now micro controllers, that run 'regular' software. And some of these devices are quite invasive. think of modems in cell pohnes, that handle everything from GPS to nand storage
hipboi has left #linux-sunxi ["Leaving"]
hipboi has joined #linux-sunxi
<oliv3r>
so then it's very much a problem with freedom
<oliv3r>
personally, I think someone wrote 'the 5 levels of freedom' forgot the details
<oliv3r>
Turl: duck typing?
<oliv3r>
basically what it said was, there's things that are important to be free, and things not at all. think of very simple dumb microwaves with a few lines of microntroller code
<drachensun>
I think he decided the mailing list was too quiet and so wanted to say something that got everyone all stirred up
<oliv3r>
and while it could be potentially dangerous (bug that blows up your house) it's extremly unlikly
<oliv3r>
drachensun: lol
<drachensun>
like we he excused me of doing something illegal by shipping to the EU then shut up when I defend myself from his lies
<drachensun>
that accusation started a 20+ post thread
<oliv3r>
but in the future, after most software IS gpled and free; other firmwares will be the next target to tackle. we simple don't have the resources at this moment. but look at the maker movement, the open hardware movment, open firmware is going to happen, as much as open source software is happening
<oliv3r>
drachensun: in the end, we're all just humans though ;)
<Turl>
oliv3r: "
<drachensun>
well if I falsely accuse someone of a crime and then find out I'm wrong I apologize, I think he was just trolling to get people to talk like he is doing now
<Turl>
When I see a bird that walks like a duck and swims like a duck and quacks like a duck, I call that bird a duck."
<oliv3r>
Turl: that wasn't in reference to my mripard i2c question, right? :p
<oliv3r>
Turl: cause if it was, i am more so confused :p
<Turl>
oliv3r: no, in ref. to the firmware thing
<Turl>
oliv3r: if it operates like pure hw and is usable as pure hw then it must be hardware :p
<oliv3r>
ah; well a lot of devices (that require 'real' firmware) isn't hardware anymore I suppose
<oliv3r>
as I said, most is a microcontroller running (quite a lot) of software
<Turl>
if your thing needs you to load a 3MB blob it isn't
<oliv3r>
I guess 20 or so years ago, when he made up that statement
<Turl>
it you just power it and it works, then I consider it firmware
<oliv3r>
Turl: in theory, 3MB could still be a lot of static 'data' like look up tables etc (think encoders/decoders)
\\Mr_C\\ has quit []
<Turl>
oliv3r: yeah but the thing is not the size but that you need to load it imo
<Turl>
if it were burned into a ROM on the chip I wouldn't care that much if it's free or not
<oliv3r>
well to be fair, i have little closed blobs i think
<oliv3r>
high-end video cards are one
<oliv3r>
broadcom encoders are other
<oliv3r>
crystal-HD
<oliv3r>
but think of raspberry-pi 'firmware' :)
<Turl>
yeah that's obviously closed
<oliv3r>
but far from what you'd concider 'firmware' however
<Turl>
that's a full proprietary OS on itself
<Turl>
that runs on another core :p
\\Mr_C\\ has joined #linux-sunxi
<Turl>
and that, I bet, shares the same memory
<oliv3r>
yep
<oliv3r>
it does
<mnemoc>
anything needing more than a couple of kilobytes needs to be updatable, because it will have bugs
<oliv3r>
oh i agree
<oliv3r>
absolutly
<oliv3r>
but I also think that it should be open source :)
<Turl>
so technically it could be spying on you freedom-wise, like iirc, nexus s modem and so on
<Turl>
replicant had a series of posts on that
<mnemoc>
oliv3r: that's where free software as religion collides with the free software as good engineering practice
<mnemoc>
for the FSF a read-only "firmware" of several MBs is "fine"
<hramrach_>
but it's still bound to be bug-ridden junk
<mnemoc>
that's why the idea of "whatever is not updatable is firmware" is f*ed up
eebrah has quit [Remote host closed the connection]
<mnemoc>
it's kind of odd to see how ideas (even atheism) can become a fundamentalist idea
<mnemoc>
s/idea/religion/2
<oliv3r>
well, as an engineer, having source publically developped (via github or whatnot) is the only future way
<Turl>
woo python virtualenvs ftw, I hated moving my python symlink to py2 or 3 depending on what I was doing :P
<mnemoc>
all software has bugs. to solve those bugs software *needs* to be updatable. software is better developed is public, and all software *should* be free
<mnemoc>
Turl: and that's why python is much nicer to use than ruby/RoR :p
<Turl>
mnemoc: it is said that Chuck Norris' software doesn't
<mnemoc>
:p
<Turl>
mnemoc: RoR has sth similar iirc
<oliv3r>
mnemoc: perfect :p
<hramrach_>
speaking of weird programmig environments
<hramrach_>
anyone has some idea if tools for converting x86 biaries into java bytecode are available?
<Turl>
why would anyone do that? :P
<hramrach_>
to make legacy application cross-latform ;-)
<mnemoc>
hramrach_: google has a sandboxed thing for native bins
<hramrach_>
it's been actually done commercially I suspect
<hramrach_>
just wondering if there is a tool available for conversion somewhere
<mnemoc>
hramrach_: and it has a list of "related projects" at the bottom
<Turl>
some-some isn't useful if you actually intend to run it
<hramrach_>
why?
<hramrach_>
it's what quemu does, anyway
<hramrach_>
just the target is the native assembly
<Turl>
because you cannot run a program if you just translated half of the instructions? :)
<hramrach_>
you translate the rest when you get to them
<oliv3r>
if it where something like 'take your x86 compiled C source; then this 'jit' compiles it to bytecode
<oliv3r>
then i could understand, but taking x86 asm and wrap it in bytecode; sounds ... horrible :)
<hramrach_>
x86 compiled C source is x86 assembly, basically
tinti has joined #linux-sunxi
<Turl>
hramrach_: what's wrong with qemu-system-x86 btw?
<Turl>
you can run qemu-system-x86 ./myx86binary and it'll run on arm
<hramrach_>
sure
<hramrach_>
I expect that if hte translation to jvm was done right you get the often used bits of code optimized
<Turl>
in my experience java isn't very 'optimized'
<hramrach_>
with qemu it's static some-some pattern translation as I understand it
<hramrach_>
java is not optimized for shellutils
<hramrach_>
but for long running stuff it can give decent performance
<mnemoc>
JITs are cool
<hramrach_>
not only it JITs, ot can optimize at the bytecode level too
<Turl>
hramrach_: the only long running thing I used in java (JDownloader) didn't precisely shine for its performance either
<hramrach_>
in what respect?
<Turl>
it was horribly slow and ate a crapton of memory compared to a native app with the same functionality
<jinzo>
may I ask which native app did you use?
<hramrach_>
it probably uses an over-engineered Java GUI which makes it appear slow
<mnemoc>
openoffice and that beast IDE from ibm are impossible to use on my humble laptop
<jinzo>
as I'm forced to use JDownloader for some stuff, but find it quite good feature wise.
<hramrach_>
and it uses a craptom of memory because memory is pre-allocated with jvm
<Turl>
mnemoc: OO isn't java is it?
<hramrach_>
most of OO is not java
<Turl>
jinzo: I don't recall now, let me see if I find it
<hramrach_>
it uses some java plugins
<mnemoc>
right. I was mostly thinking in that IDE thing. not OO. my mistake
<jinzo>
that would be great, thanks.
<Turl>
jinzo: console wise, plowshare works fine for most of them
<Turl>
eclipse is also another slow java beast
<shineworld>
I'm used with java programs running in mac/win/lin and are all champions of slowness
<jinzo>
thanks.
<shineworld>
but run in everysystem so...
<Turl>
jinzo: I think it was tucan manager, but it looks pretty dead noawadays
<hramrach_>
the shudo link is interesting. looks like some ancient site reporting progress on the JIT development before it went into JVM mainline
<shineworld>
yeah eclipse is slow and a lot complex system however
sud0x3 has joined #linux-sunxi
<hramrach_>
but it't the JIT from JVM to x(x86, ia64, ..) while I would like JIT or static translation from x86 to JVM
<shineworld>
something like JGC.. where you write java and objain an executable ... remain the framweork and os problem
<shineworld>
GCJ
<hramrach_>
and NaCl is interesting but requires code specifically written for the framework
<hramrach_>
I am not sure what GCJ is supposed to be. Isn't it supposed to be javac replacement which together with gnu classpath wassupposed to replace jdk were it completed and less bug-ridden?
<shineworld>
in past I've tried also pascal way with borland delphi and kylix... one frawework and two builds running in Windows and Linux OSs
<shineworld>
now embarcadero is doing same thing with new delphi, which compile for windows / mac / IOS
<oliv3r>
Turl: didn't you mention some android <--> irssi bridge a few days ago?
<Turl>
oliv3r: irssinotifier
<Turl>
it just alerted me when you wrote that line ;)
<hramrach_>
well, I am looking for way to run existing binaries. When the source is available you can just rebuild
<oliv3r>
btw, do we need to write a bindings.txt file for devicetree entries that are only 1 address and 1 name?
<Turl>
oliv3r: the more docs the merrier :D
<oliv3r>
i'll do that in my v3 patch then :p
<shineworld>
if your target is to cover windows/linux/MAC OS you can use FCL + Lazarus IDE. They are working on porting android/ios
<shineworld>
overall you get a pure x86 code (for first three)
<shineworld>
embarcadero is using same open-source compiler to do cross-compiling from windows to mac os
<hramrach_>
shineworld: That's not what I am looking for. If I can compile it it can run on arm all right. I am looking for solution for existing binaries
<shineworld>
what about any framework dependencies ?
<hramrach_>
also not quite awesome but the best match I guess along with that JS stuff
<hramrach_>
Turl: any idea what is the limit on system size on Android?
<hramrach_>
I built the 'kithensink' kernel and it is 4.3M in size with 258M of modules
mturquette has joined #linux-sunxi
<hramrach_>
on an ext3 filesystem
<Turl>
hramrach_: usually around 200-250M on AW devices
<Turl>
so, that's a big no-no
<hramrach_>
the livesuit cannot change the layout?
<Turl>
yeah but didn't livesuite also have a bug about partition sizes no larger than some size?
<hramrach_>
you can cut a lot on the modules but still there are going to be new drivers all the time so some initial pre-determined partition size is never going to last forever
_BJfreeman has joined #linux-sunxi
<hramrach_>
I never flashed anythign with livesiut tbh
<Turl>
hramrach_: the thing is, needing custom modules for android is rare
<hramrach_>
so long as the initial build includes enough stuff
<Turl>
because android devices are pretty much fixed function - run apps, use the wifi on the device
<Turl>
you don't usually plug webcams or usb ethernet to a tablet
<hramrach_>
you dont't?
_BJfreeman is now known as BJfreeman
<Turl>
nice, hansg wrote a new axp152 driver
<hramrach_>
well, with lack of module autoload it's going to be difficult
<hramrach_>
yes, sounds like it
<shineworld>
I do :)
<hramrach_>
I have devices with axp 209 and axp 199 ;-)
<hramrach_>
there you go
<shineworld>
I mean I plug a lot of devices on my android device depending by use
<hramrach_>
but none of them work because android
<hramrach_>
1) does not have drivers
<Turl>
looks like the driver is pretty clean, I bet it'd be pretty easy to mainline it
<hramrach_>
2) cannot load them
<hramrach_>
nice, at least a13 can work with mainline then
<hramrach_>
can the android gadget be reasonably used for networking with an unsuspecting PC
<hramrach_>
provided it workes
<Turl>
well, you don't need an axp driver to boot mainline
<Turl>
hramrach_: yeah, it has rndis
<Turl>
never used it on sunxi though
<hramrach_>
still does not solve the problem of connecting to an actual ethernet
<hramrach_>
but makes usb-usb kind of pointless
<hramrach_>
bbl
paulk-desktop has joined #linux-sunxi
ganbold_ has quit [Remote host closed the connection]
<Turl>
mnemoc: what gpl compliant code did you get from luke? :)
<mnemoc>
nand
<mnemoc>
but haven't looked at it yet
shineworld has quit [Ping timeout: 264 seconds]
shineworld has joined #linux-sunxi
<Turl>
mnemoc: the mail made it sound like boot0/1
<mnemoc>
there is nand code for boot1, u-boot and linux. but nand
hipboi has quit [Quit: Leaving]
shineworld has quit [Quit: Leaving]
awafaa has joined #linux-sunxi
<nove>
this is nice, h264 also appears to replay ok
<nove>
now to write a minimal blob player, because vlc is not ok
BJfreeman has quit [Quit: had a good time]
<rz2k>
nove: you are working on cedarx?
<nove>
rz2k, yes to RE it
<rz2k>
great! good luck with that. are you going to tie it to some existing api or just experimenting?
<nove>
ssvb, i did nothing new, only today that i first made h264 traces
<nove>
is so much complex than vp8
<ssvb>
nove: if you can replay the traces, then it's surely doable
<ssvb>
can you try tracing and documenting something more simple like mpeg1/mpeg2?
<nove>
i did 1 mpeg1, same level off complexity than vp8
<ssvb>
while not really very useful for the end users, these codecs are well supported in various video decoding frameworks
<ssvb>
is mpeg1 replayable?
<nove>
i will try now
leowt has joined #linux-sunxi
<leowt>
is there any A20 android SDK?
<nove>
ssvb, yes it works
<nove>
ssvb, currently this "works" is by visually check the mmap areas dump for something that looks a decoded frame
<ssvb>
would it be difficult to track the actual buffer pointer?
<wingrime>
nove: can you implemen image dumping in YCrCb color space ?
<wingrime>
On some events
<wingrime>
for example reading status register
<wingrime>
nove: And I recomend you focus on mpeg
<wingrime>
nove: I documented jpeg decoding on cedar, some must help with PoC of if
<ssvb>
nove: how are you passing the input data to the decoder? are they using some pointers to memory blocks with the data, or writing this data to FIFO HW registers?
<nove>
wingrime, yes eventually i will make that
<ssvb>
wingrime: the information about jpeg on the wiki page is still not really usable
<wingrime>
nove: Please read my results on page
<nove>
ssvb, pointers to memory blocks
<wingrime>
ssvb: enought to make PoC
<ssvb>
nove:
<nove>
ssvb, the frame bitstream is clear seeing
leowt has quit [Quit: leowt]
<wingrime>
nove: mpeg decoder have 4 buffer reserved (registers that constain pointers on RE page)
<ssvb>
nove: is the frame bitstream passed through the hardware registers? or do you see something like writing a pointer of the memory block and its size to some HW registers?
<nove>
ssvb, mpeg/h264 uses fifos , but vp8 desn't
<wingrime>
ssvb: you defenetly right mpeg decoder use registers that constain pointers
<nove>
ssvb, the register pointers way
<ssvb>
nove: ok, then the output buffer probably should be also configured in a similar way
<nove>
ssvb, it is
<ssvb>
ok, sounds good, then probably it would be also easy to use it instead of scanning the dumps
<wingrime>
ssvb: [MPEG_BASE+0x28] constain pointer to Stream buffer begin ,[MPEG_BASE+0x2c] constain relative offset form buffer begin (current frame)
<wingrime>
oliv3r: are you still have intension make jpeg decoder?
<ssvb>
wingrime: regarding the jpeg information, it is really insufficient for anyone trying to make the jpeg decoder (unless the decoder developer is ready to do reverse engineering for the missing details)
<rz2k>
wingrime: i just missed the window, that was not about cedarx :)
<wingrime>
ssvb: realy , i fully described full process for single jpeg
<wingrime>
ssvb: what extractly not explained?
<ssvb>
wingrime: for example, you don't explain how to convert the DQT data into the format, which needs to be written to HW registers
<ssvb>
wingrime: it is not clear that 'IQ table' is the same as DQT based on your description, and people don't like to make guesses with this kind of information
<ssvb>
wingrime: is it so difficult to make the description really unambiguous and easy to use?
<wingrime>
ssvb: DQT is JPEG tag, IQ is generan name
<ssvb>
please use the fucking terms from the JPEG spec
<wingrime>
ssvb: I can't use Jpeg-specific trem here
<wingrime>
ssvb: Or I have make some into with terms?
<ssvb>
well, I would prefer to read the documentation, written by somebody who can :)
<ssvb>
I'm saying that your general overview is not a real documentaion yet
<wingrime>
ssvb: that page not only for jpeg stuff, I write pege for whole cedar here , and a defently use same terms for all stuff
<wingrime>
ssvb: for mpegs this have name IQ (Inverse Quantization)
<wingrime>
ssvb: and I defently sure that IQ are used in others decoders here
<ssvb>
and all of them having identical bitstream format for it?
<nove>
mpeg2 also replays
<wingrime>
ssvb: only jpeg-decoder need that (may be code for mjpeg are not fresh and use old method) others mpeg need two values here minmum level and normal level
<wingrime>
ssvb: only jpeg-decoder need send this matrix
<ssvb>
it's about the same as saying that something is using 'unicode' format, without explaining the fine details about how exactly it is stored (UTF-8, UTF-16, UTF-32, ...)
<wingrime>
ssvb: 32 bit integers two matrix (8x8) in zig-zag order
<ssvb>
forget about zigzag, it is irrelevant and misleading here
<wingrime>
ssvb: are you about input steam ?
<ssvb>
and your claim about 32 bit integers seems to be wrong, based on the trace logs that you have shown
<wingrime>
ssvb: waaaat ??
<ssvb>
please show the fucking logs for initializing this 'IQ' table again
<ssvb>
earlier you have shown a dump of bitstream for the DQT chunk from jpeg, you have also shown the logs of HW register writes
<ssvb>
and they were surely not exactly identical
<ssvb>
while a lot of similarities naturally were there
<wingrime>
ssvb: jpeg in file as I remeber was 16 bit values
<ssvb>
a proper documentation is all about taking the information from the bitstream, and writing it to the right hardware registers in the right way
<wingrime>
ssvb: but it can differ according stangart
<ssvb>
guess what? these differences are kinda important if we want to implement a correctly working decoder
<wingrime>
ssvb: it can be 16 bit or 21 bit values
<wingrime>
sorrt 32 bit
<wingrime>
*sorry
<wingrime>
16 bit or 32 bit , and code for parsing jpeg chunks can be used from libjpeg her (blob use)
<ssvb>
the wiki page currently does not say anything about this
<wingrime>
ssvb: whait a second, I try find code for IQ generation in ffmpeg
<wingrime>
ssvb: wiki page not about have get IQ , wiki page about have send it to cedar
<wingrime>
ssvb: DQT chunck have simple struture
<ssvb>
exactly, I want the wiki page to describe how to write this information to cedar HW registers
<ssvb>
also you mention zigzag there for some reason, does it mean that the values need to be reordered before writing to HW registers?
<ssvb>
if not, then please remove all the misleading references to zigzag
<wingrime>
ssvb: no zigzag order are in jpeg file aleady
<wingrime>
ssvb: but it important to understand that 8x8 matrx in that order
<ssvb>
you can safely assume that the reader of your wiki page is sufficiently familiar with jpeg
<ssvb>
if you add unrelated trivia bits to the places where they do not belongs, they can be easily misinterpreted
<wingrime>
ssvb: ffmpeg have no normal documentation also (((
<ssvb>
ffmpeg has a readable C code, and everyone can get (or at least try to get) the specifications for the relevant codecs
<wingrime>
ssvb: I already say that PoC is required step now
<ssvb>
yes, a proof of concept code would surely help and could be used as a repalcement for the documentation
<ssvb>
you are just not allowed to share the reverse engineered code (the direct conversion of disassembly listing to some C source)
<wingrime>
ssvb: I ask oliv3r write PoC for this reason
<ssvb>
making a proof of concept code based on the traces of HW registers access and guessing the logic behind it is fine
<wingrime>
ssvb: I think REes should not code directly
<ssvb>
well, the problem is that oliv3r probably just can't develop a proof of concept code only using your wiki page and without working with the traces himself
<wingrime>
ssvb: That I have writed in wiki are "fine too" for oliv3r
<wingrime>
ssvb: traces are normal
<ssvb>
it's fine if he also agrees with you
<wingrime>
ssvb: he can ask me here
<wingrime>
ssvb: and I will help
<wingrime>
ssvb: indirect code writing
<wingrime>
ssvb: and offcouse I will impove wiki after we will test bunch jpegs
<wingrime>
ssvb: and get "somehow" workable code
eebrah has joined #linux-sunxi
<ssvb>
ok, you just sounded like the wiki page about decoding jpeg is already complete :)
<wingrime>
ssvb: It need PoC
<wingrime>
and I think I need to talk with ffmpeg coders directly in IRC if they have some channel
<ssvb>
they do
<nove>
wingrime, vlc is not good, it skips frames if even tries to decode
<nove>
wingrime, i started written a minimal player for the blob
<wingrime>
nove: I have noticed this
<ssvb>
nove: it skips B-frames or what?
<wingrime>
ssvb: traces have heavy load so it skip frames for sync
<wingrime>
*tracer
<nove>
runing under valgrind is slow
paulk-desktop has quit [Quit: Ex-Chat]
<ssvb>
does decoding without audio help to solve this?
<nove>
this is with no audio
<wingrime>
nove: and still we not log IRQ
<nove>
it doesn't matter, a mini player will be better, with included valgrind support
<nove>
this with 16x16 and similar 1 frames video files
rellla has joined #linux-sunxi
<nove>
the replay tool then should dump the mmaped areas, and there should be visible the decoded frame
<nove>
wingrime, only tested in a A13
<wingrime>
nove: and action on screen
<wingrime>
?
<wingrime>
*any
<wingrime>
relayed frames are on scereen?
<nove>
wingrime, no
<wingrime>
nove: than we need trace /dev/disp IOCTL too
<nove>
wingrime, for what?
<wingrime>
nove: for trece on screen
<nove>
wingrime, is not needed
<nove>
wingrime, if is for knowing when a frame start and ends another
<wingrime>
nove: it defently after status register check
<nove>
wingrime, the mini player will do that, using valgrind with print directly to the output trace
<wingrime>
nove: I glad we hav you
<nove>
wingrime, ? why for?
<n01_>
mripard_: ping
notmart has quit [Read error: Operation timed out]
notmart has joined #linux-sunxi
notmart has joined #linux-sunxi
notmart has quit [Changing host]
hansg has joined #linux-sunxi
jemk has joined #linux-sunxi
hansg has quit [Quit: Leaving]
KamiKaze_Phoenix has joined #linux-sunxi
<KamiKaze_Phoenix>
hello
<KamiKaze_Phoenix>
how i can change boot arguments in U-boot ?
<n01_>
KamiKaze_Phoenix: uEnv.txt
<KamiKaze_Phoenix>
thk n01_
wingrime has quit [Ping timeout: 264 seconds]
KamiKaze_Phoenix has quit [Quit: leaving]
rellla has quit [Read error: Connection reset by peer]
shineworld has joined #linux-sunxi
<mripard_>
n01_: short pong :)
<mripard_>
oliv3r: Turl: hi, what were these highlights about ?
jemk has quit [Read error: Operation timed out]
mdfe has joined #linux-sunxi
<Turl>
mripard_: hi
<Turl>
mripard_: we were discussing some comments oliv3r got about unbinding his driver and what it meant
eebrah has quit [Ping timeout: 240 seconds]
<Turl>
mripard_: the devm_* stuff will unmap all the things when unbinding right?
eebrah has joined #linux-sunxi
<mripard_>
yep
<mripard_>
it's precisely the point of it :)
<Turl>
mripard_: he wasn't nulling the pointer on .remove() so it'd be invalid, and his other function wasn't static, so it could potentially dereference random stuff
<Turl>
mripard_: btw, david took all the emac stuff - even dt :p
<mripard_>
yeah, I saw that
<mripard_>
it's not very important, I did agree on merging the commits ;)
<Turl>
mripard_: it might conflict a bit when i2c gets in
<mripard_>
we'll see
<Turl>
I had it happen when building sunxi-next :p
<mripard_>
ah, let's wait for wolfram to show some lifesign before speaking about merging i2c :)
<Turl>
haha
<Turl>
mripard_: btw, I pushed a sunxi-clk branch to my repo with mmc stuff the other day if you want to have a quick look
<mripard_>
ok, on bitbucket?
<Turl>
mripard_: yep
<Turl>
mnemoc: btw, bitbucket closed my 'bitbucket is slow' ticket claiming they reworked the commit listing to be done in linear time now :)
<mripard_>
Turl: I'll take a look at it, thanks!
<Turl>
mripard_: if you find anything broken, typoed or otherwise ugly let me know
<Turl>
mripard_: also, I'm open to patches to add the required DT nodes to sun5i :)
<mripard_>
:)
eebrah has quit [Ping timeout: 252 seconds]
<Turl>
mnemoc: it really does feel a bit faster :)
shineworld has quit [Quit: Leaving]
notmart has quit [Quit: notmart terminated!]
tzafrir has quit [Read error: Operation timed out]
nove has quit [Quit: nove]
n01_ has quit [Read error: Operation timed out]
<rz2k>
hno: slapin_n1: got yuq's driver to read the id's right, now to fix the actual reading issue, it detects all blocks as bad right now.
<rz2k>
problem was the chipselect function working first time ok, second time it places wrong data and controller outputs zeros