Regarding running Android LXC guest in Ubuntu touch by using Mir

[Thomas Voß]

On Fri, Jan 8, 2016 at 2:55 AM, Daniel van Vugt
<daniel.van.vugt at canonical.com> wrote:
> That all said, there might be a short-cut. If you accept that it may
> only work on Ubuntu phones that were formerly Android phones, then with
> maybe some weeks/months of effort you might be able to build the
> required translation layers for graphics and input etc...
>
>

The approach would not necessarily be limited to phones (or devices
that used to be Android devices).
I illustrated a hypothetical stack (and a bigger picture) in the
diagram attached to this mail. In that diagram:

  * orange boxes refer to services and middleware provided by Ubuntu.
Please note that
     the respective layer in the diagram refers to arbitrary system services.
  * green boxes refer to Android drivers (binary blobs). Please note
that we are able to consume those drivers
    without modifications and as they ship for Android (built against
bionic etc.).
  * the dotted rectangle refers to a potential container hosting the
app and its runtime. It is not strictly required, though, depending on
the
    type of app (pure java would likely not require a container, apps
using the ndk might likely need one to provide necessary elements like
bionic).
    Please note that the "Android container hosting apps" is a thought
experiment only, we haven't done this for Android apps, only for
legacy X11 applications.

With that, we are able to swap out the Android (driver) layer and
replace it with alternatives (e.g. on vanilla linux setups).
The app could follow along together with the emulation layer and "just
keep on working". This is subject to a few unknowns though,
specifically
on the set of services that an app requires. If the app was deeply
integrated with Android, there is a fair chance that one or more
services provided by Android cannot be
mapped to the Ubuntu platform.

A closing remark on the "Android emulation layer": It could
theoretically be Android itself (with the app and the layer running in
an LXC container) in which all
required system services have been patched to talk to the Ubuntu side.
It would require significant effort, but I think it *could* work.

HTH,

  Thomas

> On 08/01/16 09:21, Daniel van Vugt wrote:
>> Support for Android apps is unlikely to happen any time soon.
>>
>> You have to remember that Ubuntu Touch is not Android, but is a full
>> Ubuntu system. The fact that Ubuntu Touch uses an Android kernel is not
>> sufficient to support Android apps.
>>
>> Although parts of Android remain and are visible in the Ubuntu Touch
>> filesystem, they are unlikely to function correctly. Certainly even if
>> you could run an Android app right now then it would not appear on the
>> screen. That would require significant work. And even still, not all
>> Ubuntu Touch devices will be based on Android devices. So you would
>> really need to start from the ground up and build a self-contained
>> Android emulator.
>>
>>
>> On 07/01/16 17:53, 유재용 wrote:
>>> Come to think of it, by replacing surfaceflinger, would it be possible to
>>>
>>> run Android apps as it appears as a native application in Ubuntu touch?
>>>
>>> In other words, if you run Angry bird in Ubuntu touch, it will launch
>>>
>>> Angry bird in Android and the screen output is composed to the application
>>>
>>> inside Ubuntu touch?
>>>
>>> Thanks,
>>>
>>> Jaeyong
>>>
>>> ------- *Original Message* -------
>>>
>>> *Sender* : Andreas Pokorny<andreas.pokorny at canonical.com>
>>>
>>> *Date* : 2016-01-05 03:07 (GMT+09:00)
>>>
>>> *Title* : Re: Re: Regarding running Android LXC guest in Ubuntu touch by
>>> using Mir
>>>
>>>
>>> On Mon, Jan 4, 2016 at 6:50 AM, 유재용 <jaeyong.yoo at samsung.com
>>> <mailto:jaeyong.yoo at samsung.com>> wrote:
>>>
>>>      Oh, I see. That is more interesting.
>>>
>>>      Just for the curiosity, could you tell me more about the early days
>>>      before Mir?
>>>
>>>      I'm wondering why you choose to remove surface flinger.
>>>
>>>      Is there some constraint if you keep using Surface flinger and
>>>      another graphics server?
>>>
>>>      Thanks,
>>>
>>>      Jaeyong
>>>
>>>
>>> Hi Jaeyong,
>>> Replacng surfaceflinger with a mir based system compositor - next to the
>>> obvious idea of making the android driver based stack similar to the
>>> mesa/kms based stack - allows us to separate the user session from the
>>> output and input devices allocation. With that it will be easy to have a
>>> seamless switch between user sessions or have them running in parallel
>>> and move devices between sessions .. and more.
>>>
>>> regards
>>> Andreas
>>>
>>
>
> --
> Mir-devel mailing list
> Mir-devel at lists.ubuntu.com
> Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/mir-devel