<nove>
if the _user_ don't support us, how can we be motivated
<nove>
because whatever we do it is not enough, they ignore us
<nove>
(and i bet that this will happen), when finally the video engine mem2mem driver arrive to a working state, they (the _user_) will keep ignoring this work, because the preferred-media-player-of-the-day will still not "work"
<premoboss>
nove, i don't understand exactly who you are complain.
<vpeter>
Who is the _user_ and what should it do?
<nove>
i am complaining to nobody, because only nobody care about what i(we) say
<premoboss>
2 option about _user_: A. the *real* end-user (who buy the device at the market) B. the buyer of the IP to place inside S.O.C. (ALLWINNER in this case)
<nove>
the _user_ should support its one words, if the _user_ wants to have proper software then it should support whoever is working to make proper software a reality
<jackdaniel>
++ last statement
<premoboss>
nove, "support" what means? in case _user_means "end user", shat in your opinion he should do? donate money? press on SOC procucer to open the source? I donst understand what you mean.
<premoboss>
shat=what
<jackdaniel>
premoboss: use software, which strives to be compliant with proper implementation
<jackdaniel>
not one-feature-more hacky software depending on another hacks
<premoboss>
do you mean the user have to use what is available without complain becauser "there is not this feature, there is not that feature, etc"?
<jackdaniel>
no, i mean people tend to prefer eye-candy and feature-rich over stable and carefully improving software
<jackdaniel>
yet demanding stability ;-)
<premoboss>
jackdaniel, that kind of people are >95%. The tipical apple-oriented people, just to say the type.
<premoboss>
so that subnormal people are the normality. and we are the "abnormal".
<jackdaniel>
when it comes to the bridges, nobody gives a fuck about 95% desire to have bridge built before sunday. it is build when it is given to use when it's stable enough. unfortunately building software seems so easy at first glance
deasy has joined #linux-sunxi
<premoboss>
the solution is dont care what people ask and go on on your proper way to realize software as you like to do. if people will use or not, up tho them. satisfation came first fro doing a good jow: if used ok, if used by more people, better.
interrobangd has quit [Quit: Leaving]
<hramrach>
bbrezill1: ok, I will they the other branch
<hramrach>
thanks
<hramrach>
actually, apple-oriented people prefer shininess over anything, especially features
<hramrach>
it's not that Apple products don't have unique features .. and 'features'. They also lack many common ones
<hramrach>
also when there was no civil engineering worth noting people just built bridges before sunday 95% of the time. And 90% of the time they fell. whatever
<premoboss>
i used "apple-people" just to give a raw examples, preferred to avoid for politeness the right words: idiots.
<hramrach>
we now know how to build bridges and learned to demand non-falling bridges
<hramrach>
but in absence of non-falling software for most tasks people don't demand it
<hramrach>
when you had to deal with falling bridges to get anywhere you just prayed or picked non-stormy days for travelling. when you have to deal with falling software to have anything done on a PC you learn to backup, restart, reboot, whatever
<nove>
open source didn't succeed because of one-man-job, but by doing "community" sharing of the hard things, even if the _user_ doesn't know how to do software programing, still the _user_ can help and support to create an ambient "a community" in which the ones that have the "knowledge and skill" feel comfortable and work recognized
<hramrach>
"community" is about two things. developers going out of their way to make the software usable for somebody else on some other machine than their own. and users going out of their way to report and/or fix the issues in the software
Andy-D has quit [Ping timeout: 240 seconds]
<premoboss>
community is made by developer and testers. _user_ are just a huge amount of person that don't care noting about comunity but only use what comunity create withoud gine any kind of positive feedback. imo.
<hramrach>
there is not really developer and user distinction. you might be developer of one thing and user of another, you might be no developer but write manuals, do translations, adapt to local conventions, draw graphics, etc
<vpeter>
nove: Maybe you should write a comment to disqus.com for the _user_. Maybe then it will see things differently.
Andy-D has joined #linux-sunxi
<vector80>
Just managed to have an RTSP server. Board is an A20, and I tested on Hi253, OV5640 and Logitech c510 webcam.. It is just a simple pipe by using FFMpeg and VLC. However, I have two big trouble, one is extremely high bitrate on the output ( almost 2-5Mbps), and I can't go beyond 640x480... Did anybody have any experience with the CSI on sunxi ?
<vector80>
I also wish to write this on the wiki, if you don't mind ?
<nove>
only there? not also in every other of the multiples forums? i would prefer to spend my time doing more useful things, instead of having to run after every _user_
<nove>
vector80: its a wiki, fell free to write
<vector80>
thank you nove.
<vector80>
I wish somebody help me to test my troubles with Cedrus, because I couldn't understand why there is such high bitrate
<nove>
vector80: simple answer, the software is not ready
<nove>
vector80: correction, that is not "cedrus" and not what "cedrus" aims to be
<vector80>
yes nove, I know it is PoC, but I really wonder who is working on it
<vector80>
I am reading forums, trying to understand what people did
<vector80>
But other people also saw these issues
<vector80>
but nobody replied
<vector80>
I am also ready to pay for it, and I am ready to make it opensource again for community use
<vector80>
But since it is soo hard to find the developers, how can I pay them :)
<nove>
vector80: the person that wrote those software, never came here to speak to us or to work together, and only choose to use the name "cedrus"
<vector80>
If this is the case, how developers complain about community interest, or "looking for a driving source to make this hard work" ?
<vector80>
I am saying, stating that I am ready to pay for it. If that would help sunxi community to grow, I am ready to help in paying, documentation, and everything I can.
<vector80>
But we usually come here, try to ask for something, but we can't find developers easily.
<nove>
vector80: better then $$$, would be to ask (allwinner) to support this community(linux-sunxi) every time your company gives $$$ to allwinner
<vector80>
Nove, do you think that I didn't talk this with AW ?
<jackdaniel>
vector80: hire developer to work on sunxi with community
<vector80>
I went to their office in zhuhai many times, tried to talk with them face to face
<vector80>
I am trying to hire also
<vector80>
Pls consider
<vector80>
Lets say, for USB driver code, does anybody need to hire a developer? It just works out-of-the-box
<vector80>
But for CEDAR?
<vector80>
That is the most troublesome part of this SoC
<vector80>
Under android, CEDAR works very very well
<vector80>
I have tested it more than 50K of our products. But for linux, it is not.
<vector80>
Anyway, I am writing here. If anybody is interested in Cedar / Cedrus kernel driver development under linux-sunxi, I am ready to pay for it.
<premoboss>
wat more AW want is money. tell them u move to other soc (i.e. Rockchip). if u are a big custorer, her will do whatever you want to not loose your money.
<vector80>
Unfortunately I am not a "big" customer for them
<vector80>
Their "big" is "huge" for me
<premoboss>
so they will newer give you a shit.
<premoboss>
they are money drived.
<premoboss>
not else.
<vector80>
premoboss: I am talking with AW for about 1.5 year
<premoboss>
any success?
<vector80>
I already know that they will not help on anything
<jackdaniel>
vector80: there are many consulting companies
<vector80>
Of course I could go over RockChip, but pls compare, which one would be a better choice by now for camera encoding ?
<vector80>
jackdaniel: Pls tell me one ?
<jackdaniel>
free-electrons for instance
<jackdaniel>
other one i work for is antmicro
<premoboss>
vector80, of how many soc we are talking about?
<vector80>
I also talked with them. They told me that they don't have time for a closed source library.
<nove>
vector80: i am not here for $$$, so with or not $$$ nothing will change for me
<vector80>
premoboss: For now, lets say only A20, and for next year, lets say A8X
interrobangd has joined #linux-sunxi
<vector80>
nove: I know you are not, I am not talking for you
domidumont has joined #linux-sunxi
<vector80>
nove: I know most of the developers do this hard-work only for community or for themselves.
<vector80>
And I appreciate it
<premoboss>
no, i mean: how much number of soc u are interested to buy? maybe a "second line" company do what you want if youd "big" is "huge" for they.
<nove>
vector80: a proper driver for the video engine could had started already more than one year ago, nobody can forward to task of writing it
<vector80>
premoboss: Sorry misunderstand. I also tried that way. In that case, I also couldn't get a "guaranteed result"
<premoboss>
ok.
<premoboss>
so u are in a "cul-de-sac"
<vector80>
premoboss: I am sorry but I don't know what that means.
interrobangd has quit [Client Quit]
interrobangd has joined #linux-sunxi
<vector80>
nove: Anyway, when I visit this IRC, I feel happy because at least I see such people here.
<vector80>
I am also thankful to all developers for their work until this moment.
<premoboss>
"cul de sac"= you are walking in a "one way direction" street, but you reach the end of the street and there is not way to get out from it.
<hramrach>
for video encoding AW is probably better atm
<hramrach>
it works at least partially
<nove>
vector80: from the information gotten by reverse-engenering , we waited, nobody came forward to the task of writing the driver
<vector80>
However, the weakest / troublesome part of AW chips is the Cedar under linux
<vector80>
hramrach: You are %100 right
<vector80>
I also checked many SoCs for my products, RockChip, Freescale, TI, and Amlogic
<vector80>
For price / performance /stability...
<vector80>
The Aw was the "best" solution for 2013-2014
<hramrach>
rk uses a popular hantro or what it's called engine but the open driver for that is in for of non-compilable code dump
<hramrach>
+m
<vector80>
Anyway. I say happy new year to everyone in linux-sunxi
<hramrach>
happy new year :)
<vector80>
If anybody is interested in doing Cedar reverse engineering,
<vector80>
I am ready to pay for it
<vector80>
after complete, I am ready to make it open-source
<vector80>
just send me an email, or find me here
<vector80>
Need to leave now, sorry
<vector80>
good day.
<hramrach>
the situation is that it's partially reverse-engineered
<premoboss>
vector80, look inside linux-sunxi in "video" pages,
<nove>
vector80: what choice we have to not make this work in a waste, i have been spending the last months learning how to write a v4l mem2mem driver
<premoboss>
maybe there are people that u can contact there.
<hramrach>
so you need to find a developer that is willing to read the docs, implement a codec, test it, fix any issues, etc
<vector80>
Thank you very much for all your answers, and I know you are trying to help
<vector80>
But I just need to leave,
<vector80>
hope to see you in more proper time
<vector80>
have a nice day
<premoboss>
ok see u, vector80
vector80 has quit [Remote host closed the connection]
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
domidumont has quit [Ping timeout: 246 seconds]
slopez has quit [Quit: brb]
slopez has joined #linux-sunxi
ganbold_ has quit [Ping timeout: 255 seconds]
<nove>
and to better explain the current situation of the "video engine v4l2 driver"
<nove>
is the following
deasy has joined #linux-sunxi
<nove>
i came here and started helping reverse-engineering with no intension of to also be the one to write the driver
<nove>
jemk, can speak better for himself, but is similar to me,
<nove>
but what choice we have, to not make all this work in a waste
<nove>
so jemk and me decided to push the driver forward
<premoboss>
nove, more driver cames, better it is
<nove>
and came with the decision, in that i will handle the v4l2 things, and jemk the codec things
<nove>
but my motivation almost nonexistent, and what the little that is done up until now
<nove>
is only a dummy (not a real driver) used principality for learning, that the must it does is initially the hardware enough to read the version register
<nove>
so there are still alot to do, and will take time until it arrives to a state that can be usable
<premoboss>
nove, do you have idea about how many time u need?
chil has quit [Read error: Connection reset by peer]
<hramrach>
but with the reverse-engineering mostly done you do not need any extraordinary skill to write the driver, just time. so well in reach of a Chinese company hiring a developer for their project
<nove>
at the current pace, months, alot of months
astr has quit [Ping timeout: 256 seconds]
<hramrach>
there are few people using encoding. it's nice to have feature but low priority for most people
<hramrach>
and for decoding to make sense to hack at you should have mainline video working which is approaching but not quite there yet
konradoo77 has joined #linux-sunxi
deasy has quit [Quit: Nom d'un quark, c'est Edmonton !]
astr has joined #linux-sunxi
<hramrach>
bbrezill1: with the sunxi-nand-next the bbt sticks
<hramrach>
I will try to test the remaining stuff
<hramrach>
do you want some of the patches sent to ML or something?
<hramrach>
arete74: you probably want to use initrd or dtb command in extlinux.conf
<hramrach>
append should work, though
<hramrach>
or just use boot.cmd if extlinux.conf is not generic enough
<arete74>
hramrach: i have need extlinux.conf for menu select kernel
NiteHawk has quit [Remote host closed the connection]
<hramrach>
I thionk you should be able to load script.bin as ramdisk so long as you can find a way to specify that ramdisk is loaded where kernel expects script.bin
<arete74>
hramrach: you have an working example?
<arete74>
howto use dtb comamnd?
bonbons has joined #linux-sunxi
Faisal has joined #linux-sunxi
<hramrach>
I don't use mainline u-boot so far
<hramrach>
but in the boot.cmd you specify where to load script.bin
<hramrach>
if you can use any command in extlinux.conf to load a blob of data at that location it should load you a script.bin just fine
leviathanch2 has joined #linux-sunxi
<hramrach>
arete74: ok, there seems to be no way
<hramrach>
so just load script.bin as initramfs and if that does not work try hacking the kernel to look for script.bin where initramfs should be
<hramrach>
also it seems syslinux commands are in UPPERCASE
<arete74>
hramrach: ok try con UPPERCASE,thanks
mcscruff_ has joined #linux-sunxi
konradoo77 has quit [Ping timeout: 245 seconds]
konradoo77 has joined #linux-sunxi
nove has quit [Quit: nove]
<bbrezill1>
hramrach: the only thing you can send is the NAND chip definition in nand_ids.c, other stuff depend on things that are not mainlined yet
seppeel has joined #linux-sunxi
mcscruff_ has quit [Ping timeout: 240 seconds]
mozzwald_ is now known as mozzwald
mcan has quit [Changing host]
mcan has joined #linux-sunxi
y0g1 has joined #linux-sunxi
<y0g1>
hello
tomboy65 has quit [Ping timeout: 250 seconds]
y0g1 has quit [Quit: brb]
FreezingCold has joined #linux-sunxi
<ssvb>
libv: how would it be best to handle migration to the mainline u-boot in the linux-sunxi wiki?
vishnu_ has joined #linux-sunxi
paulk-collins_ has joined #linux-sunxi
tomboy65 has joined #linux-sunxi
seppeel has quit [Quit: Leaving]
viccuad has quit [Read error: Connection reset by peer]
domidumont has joined #linux-sunxi
<libv>
ssvb: shouldn't hansg be responsible for that?
<libv>
which should be true for porting all supported devices in the sunxi uboot to mainline
Faisal has quit [Remote host closed the connection]
<hramrach>
ssvb: somebody who succeeded using legacy kernel with mainline u-boot and syslinux should write a wiki page about it ;-)
<hramrach>
if anybody did
* hramrach
considers hooking up a screen to something to see the mainline u-boot pretty graphics
<libv>
ssvb: in any case, a mainline kernel and mainline uboot section has been part of the device page example for a long while.
<libv>
once all currently supported devices have been ported over, the manual build howto could be altered as well as the targets in the device pages
<hramrach>
bbrezill1: that ID part also depends on the previous extra ID in your tree
<libv>
ssvb: but i guess that that would require that hansg gets over his refusal to work the wiki
<hramrach>
another developer refusing to write about what he does /o\
<libv>
hramrach: apparently, for such a fundamental change, just stating "i find it good enough, now everyone else should go do the work" in an email, seems to be enough
<hramrach>
it's fine if he writes a blog about it if he does not like wiki but at least *something* you can refer people to when they ask WTF is going on
<hramrach>
problem with that is that the email has no nice location
<hramrach>
you have to fish it somewhere from archive
imcsk8 has joined #linux-sunxi
<hramrach>
and when there is update it's not necessarily reflected on the old email thread
konradoo77 has quit [Ping timeout: 265 seconds]
<hramrach>
bbrezill1: regarding the invalid bad block markers
<hramrach>
I just tested both cubieboards and they have like half dozen detectable bad blocks each
leviathanch2 has quit [Ping timeout: 244 seconds]
<hramrach>
it does not seem erasing whole flash is atomic operation so you could delay writing out bad block table after first erase
<hramrach>
and once the bad block markers are gone there is nothing you can do. they are not written back. it is assumed that markers are placed at the factory and never change afair
<hramrach>
so even if you do a full erase without support for bad block table no markers are placed and on next boot with bad block table support all blocks are assumed good
wickwire has quit [Read error: Connection reset by peer]
<hramrach>
I moved the test so that other errors (like address errors if flash format is incorrectly detected or whatever) are still found but that's about as good as this gets
<hramrach>
I think it would be worth making it a DT option so that if somebody designs a piece of hardware and has it shipped with empty flash they can tyrn the option off for their device
montjoie[home] has quit [Ping timeout: 245 seconds]
diego71_ has joined #linux-sunxi
diego71 has quit [Ping timeout: 244 seconds]
<hramrach>
I guess I can re-test by erasing the BBT now with the working driver but reading the code there are two types of bad block BBT_BLOCK_FACTORY_BAD and BBT_BLOCK_WORN
<hramrach>
the earlier having no sense on AW devices that a Chinese vendor touched
montjoie[home] has joined #linux-sunxi
leviathanch2 has joined #linux-sunxi
<hramrach>
yes, added a message for skipped BBM and none is printed
NiteHawk has joined #linux-sunxi
Andy-D has quit [Ping timeout: 245 seconds]
JohnDoe_71Rus has joined #linux-sunxi
<ssvb>
libv: would it be ok to format the "Current status" section in a wiki device page as a table?
<ssvb>
To show the status of "mainline / sunxi-3.4 / allwinner sdk" support
<ssvb>
and also "supported / wip / unsupported" features
<ssvb>
not global conversion of the whole wiki, but maybe just a few pages as an experiment
<libv>
i am not sure that formalizing that section really is the way to go, as it might not be possible for all devices
tomboy64 has joined #linux-sunxi
Netlynx has quit [Quit: Leaving]
tomboy65 has quit [Ping timeout: 250 seconds]
<libv>
ssvb: since mainline kernel/uboot was pretty limited so far, just bullet points worked fine for the few devices
<libv>
but that section has always been pretty freeform, so feel free to experiment