[8.90.1] CoreELEC for S905/S912 devices

    Quote from Poida

    I fail to see why, LE can't continue along the path they've chosen and let CE do the same, without trying to discredit / insult them :/

    No one in Team LibreELEC is trying to discredit or insult CoreELEC or any other community builders, they are doing a fantastic job at supporting Amlogic devices! There is mainly just a split view on what to focus development efforts on.


    A bullet point list from a former amlogic LE dev (helped ensure Odroid C2 worked with the common LE 3.14 amlogic kernel and made it possible to wake/power-on your C2 device with help of CEC), now self proclaimed LE Rockchip maintainer:

    • Amlogic userbase is ~10-15% of total users (most use different
      community builds) yet generate massive amount of support and help posts in the forum
    • Current amlogic platform support in Kodi is scheduled for removal after Kodi Leia v18 is released
    • Majority of LE devs want to focus on mainline linux (and single-board computer) support across the different arm platforms (Amlogic, Allwinner, NXP, Rockchip)
    • Kodi gbm/drmprime platform is the platform that amlogic (and other embedded platforms) will have to use in future versions of Kodi, if all goes well even Intel and RPi can be switched to use this Kodi platform
    • Mainline linux is currently not fully supporting multi-channel HDMI audio, 10-bit color depth, HDR or a full YCbCr HDMI pipeline
    • Amlogic, Allwinner and Rockchip mainly use same HDMI Tx IP and any progress made on above missing features should benefit all platforms

    This is in no way meant to discourage/disrespect use or development of current 3.14 based builds, it is only a list of facts and things to look forward to in future builds.

    Quote from Kwiboo

    This is in no way meant to discourage/disrespect use or development of current 3.14 based builds,

    That is refreshing to hear, especially the absence of snipes such as

    Quote from CvH

    The current LE AML devs working at a mainline Linux (aka latest Kernel) and an proper standard implementation at Kodi for S905 that don't require a ton of special fixes.

    and

    Quote from CvH

    some nerd hacks it together over the weekend

    as seen in prior posts, suggesting otherwise. To persist along those lines only really serves to diminish LE credibility as remarks of this kind portrays them as being rather unprofessional


    Quote from Kwiboo

    There is mainly just a split view on what to focus development efforts on.

    My understanding, as an end user, is precisely that and that both CE and LE are equally deserving of respect for the services they continue to provide; one continuing to focus on today's devices with the other focusing more on the shape of things to come - both are required and appreciated by me. ;)

    Quote from CvH

    You know that AML is basically working under mainline linux - today ! Ofc some important bits are missing regarding Kodi but that is wip (therefore the current situation).

    Btw Kodi has no intention to support and maintain an not standard out of tree codec that don't follow the basic defaults. They already killed the Android non standard codecs and with K19 they kill the last linux codecs that are not following the defaults. So sticking with the legacy driver/codec is not an option for the future.

    This statement is only a partial telling of the story. AML are only releasing a mainline kernel because it is an obligation under the Android Oreo release - otherwise it would never happen. On top of this, due to licensing agreements with ARM, the mainline that they do release will be built around a closed source graphics stack which will require substantial hacking to work within LE. Already most of the 4.x LE builds only have software decoding for this very reason. The ongoing work for a open source MESON driver continues as it has done for the last 4 years - at each stage falling further behind the release schedule of actual AML products.
    On top of this AML logic has already thrown there substantial s905 user base under the bus as they will not be getting mainline kernel and so they will be left stranded by LE/Kodi (without more of that nasty real world hacking).


    The upshot of this seems to be we can have a hacked around 3.14 CE now, or a hacked around 4.x LE sometime in the future.

    AMLogic is a shitty company who don't give a damn about their end user base and they are never going to become the paragons of Linux support we would all want them to be - why should they when they can sell Android boxes by the million by doing the absolute minimum for their customers, thats the reality we are stuck with.


    Shoog

    Quote from Poida

    My understanding, as an end user, is precisely that and that both CE and LE are equally deserving of respect for the services they continue to provide; one continuing to focus on today's devices with the other focusing more on the shape of things to come - both are required and appreciated by me. ;)

    +1 :thumbup:

    Quote from CvH

    So sticking with the legacy driver/codec is not an option for the future.

    I think all would agree with that. But future is future and now is now. Now all those legacy stuff works. So why throw them out and wait for the future. Is the same as buying new TV? Let's wait for a better model ...... and then you will newer buy anything :D

    When future come all will jump to better environment. Or someone thinks that old legacy stuff will be used forever :)

    Gone...

    Quote from Shoog

    On top of this AML logic has already thrown there substantial s905 user base under the bus as they will not be getting mainline kernel and so they will be left stranded by LE/Kodi (without more of that nasty real world hacking).

    So we throw these s905's in the bin ?

    Where did you get "this nasty real world hacking" term/phrase from ?

    So all the work done by kszaq on Amlogic devices was all nasty real world hacking ? and yet no one put any heat on him to move on.

    This is getting all too seriously ridiculous.

    This is meant to be a forum for hobbyist developers and users.

    It's now turned into some Apple Corporate proprietary software development branch.:cursing:

    Quote from Shoog

    The main complaint been levelled at the CE build is that it is full of hacks and this cannot be allowed to pollute the main branch of LE through pull requests. This is because the decision has been made to drop 3.14 support so any hack which makes Kodi work with that kernel are unacceptable.

    Thats the underlying nub of what seems to be happening.

    Already existing legacy builds don't really matter to this debate because they are out of the main branch, so you can keep your s905 on Krypton and thats fine with LE. None of the old development work is lost - it just cannot be rolled into future Leia builds.


    The exact same split happened within the Kodi development project over general Android support due to the self same Kernel split. It was nasty for a short time but things quickly settled back down.


    Shoog

    Edited once, last by Shoog ().

    The decision to go for v4l m2m api in Kodi v19 is part of this jihad against everything that is not in the wide embrace of Linus&co.

    Apart that there are still discussion of how and when mainline those api (and knowing the story of v4l in linux I won't be so sure that it is going to happen really soon) then you need that all the VPUs backend got ported to this new API

    How many VPUs you know that are in some extend ready to fill this?

    Without mentioning that then you still need a OpenGLdriver to render the kodi GUI and again, how many SoC and GPU combination you know that are supported in mainline?

    This is not kernel, this is a codec jihad. There are already some standard backend (VDAPU for instance) that is possible to hook on top of vendor specific codec, making a standard layer usable for the player.

    It would have been a working, stable, middle term solution when the rest of the community waits and hope for a full open source stack for the GPU (probaly will take 2 years) and v4l m2m codec implementation (3, 4 years?)

    Because of this we are talking: years not months

    That is why a constructive discussion among LE would have been of benefit for everyone, trying to overcome Leia and check what can be done for kodi v19 that, as it is the status of the SoC support so far, is going to work in pure linux mode on basically no ARM SoC I know about today

    Quote from Shoog

    AML are only releasing a mainline kernel because it is an obligation under the Android Oreo release - otherwise it would never happen.

    CvH was most probably referring to the work done on mainline linux by the community and not by amlogic corp. The LE amlogic branch along with the Amlogic_GX project already runs on mainline linux 4.16+, the main part missing is a video decoding driver, on the upside there is good momentum on this front across all platforms, exiting times ahead.

    Just some user thoughts :


    Is the libreelec team in contact with Companies like Amlogic and rockchip or with tvbox manufacturer like videostrong ?

    I guess these companies do not have much interest for long term driver evolution for (cheap) tv boxes - having already been
    sold. What is the part of baylibre in the game?


    Which chipset will be in the chinese tv boxes next christmas ?

    What will these companies have to do and implement for selling a lot new boxes in 2019 ?

    Doesn´t look like Kodi development takes care as long as "chinese pirate box sellers kill kodi"

    They will have newest Android Version. But which chipset support will be dropped until then ?

    I wonder if companies like wetek did know that S905H api Support beeing dropped with Android 6.


    I am not optimistic that I still can use my AML stuff in 2 years with newest kodi Version.


    And Linux ? Is the mainline kernel development something can be done without support from the companies named above ?


    What I mean is : on the first hand its all about making money and selling new tv boxes.


    So - maybe I´m completely wrong - I think a split development would be nice for all the users having already a mecool, wetek, odroid or linglangsomething connected to the tv screen. One team for "here and now" and the rest for "long term evolution" sounds good for me.

    The reality is that if you want to run a box off the latest Kodi/LE you would be stupid to choose any ARM based solution. Intel products cost a bit more but offer a level of stability/dependability and future support just not possible with the shifting ecosystem of ARM. I made a mistake this time round (believing dubious marketing speak) but will never make the same mistake again. Intel every time.


    Shoog

    Other user opinion:

    It's not nice to disrespect someone else work because it's based on outdated kernel when there is no alternative.

    I understand we are currently in some transition period before working mainline linux support is available.

    It's the future no doubt and I want to be optimistic that one time sooner or later current generation of AML chips will get proper mainline support.

    But until then we are stuck with old kernel based builds because there is no other way.


    I appreciate hard work of devs on both fronts and would like to ask both groups - please, respect each other and don't waste your time fighting.

    Work of both groups is very important, we need you to continue improving our KODI experience NOW and in the FUTURE.

    Thanks 8)

    Tanix TX7 @ LE 8.90.3, Mecool KI Pro @ LE 8.90.3, Beelink MiniMX III @ LE 8.90.3

    Quote from Shoog

    The reality is that if you want to run a box off the latest Kodi/LE you would be stupid to choose any ARM based solution.

    Not if your ARM based solution costs 1/4 of the Intel product price and will work perfectly well for the next couple of years. Then you would be stupid to pay for the Intel product as, sure as night follows day, in a couple of years (or whenever it is that there is no alternative) you will absolutely get a far more capable product for your buck than you do today, and would have lost very little on your interim product.

    Quote from Poida

    Not if your ARM based solution costs 1/4 of the Intel product price and will work perfectly well for the next couple of years. Then you would be stupid to pay for the Intel product as, sure as night follows day, in a couple of years (or whenever it is that there is no alternative) you will absolutely get a far more capable product for your buck than you do today, and would have lost very little on your interim product.

    There are some very capable Intel based product out there for a little over $130.00. I spent about $90 dollars on my VIM2 with vTV board - and when all the other additions were reckoned the final bill would be about the same.
    I have a old S812 box which is stuck on Jarvis after just over a year and when I factor in all the wasted time I spent on it , it looks like very poor value for money.

    Shoog

    Quote from Shoog

    The reality is that if you want to run a box off the latest Kodi/LE you would be stupid to choose any ARM based solution. Intel products cost a bit more but offer a level of stability/dependability and future support just not possible with the shifting ecosystem of ARM. I made a mistake this time round (believing dubious marketing speak) but will never make the same mistake again. Intel every time.


    Shoog

    I bought a couple of s905 boxes for about £30 each in 2016. They've played everything I've thrown at them, cost peanuts and use bugger-all power. If they don't work with anything after Leia I'd happily buy an equivalent - if there is one.

    What would I have to spend on intel hardware to get equivalent performance?

    Quote from bylderup

    I am not optimistic that I still can use my AML stuff in 2 years with newest kodi Version.

    There is no guarantee that you'll be able to use any particular device in 2 years with the then newest kodi version /shrug


    Shoog I have never paid more that around $35AU for the china stuff (maybe the beelink but still less than $50AU), and really haven't wasted any time at all on them, thanks to OE -> LE -> CE (If it don't work I revert uand update again when it does) - Simples! :thumbup:

    Edited once, last by Poida: spelling ().

    Quote from Shoog

    There are some very capable Intel based product out there for a little over $130.00

    Mate thats $180 AUD in my country. I have AML 3 boxes in the house running as TVH Clients as well.

    So a user like me would have to fork out $540.00 AUD for intel to replace 3 AML boxes which cost a total $135.00 AUD

    Your either having problems with adding up or there is another new type of stupid i'm not aware of.

    Quote from Shoog

    The reality is that if you want to run a box off the latest Kodi/LE you would be stupid to choose any ARM based solution. Intel products cost a bit more but offer a level of stability/dependability and future support just not possible with the shifting ecosystem of ARM. I made a mistake this time round (believing dubious marketing speak) but will never make the same mistake again. Intel every time.


    Shoog

    Interesting view.

    I actually went back the other way recently because the Intel solution I had wasn’t powerful enough for x265 and I didn’t want upgrade and invest in a small PC just to use Kodi, or use an RPI or something similar as a media centre that is more for tech minded people.


    Minix have nice little boxes that are reliable and powerful enough and should last a few years more at least.


    And this is also the very reason as a user of the earlier versions CE, I randomly got involved with the project because of how welcoming and open they were to new people,