Android on a Raspberry Pi

[hippy]

https://www.raspberrypi.org/blog/magpi- ... spberry-pi

"Hey folks, Rob here with good news about the latest edition of The MagPi! Issue 71, out right now, is all about running Android on Raspberry Pi with the help of emteria.OS and Android Things."

So does that mean the Android forum can be brought in from the cold ?

[Imperf3kt]

Android things is not Android. There is a forum for it though, so there is no need to lump it together with Android.

[thagrol]

Android things is not Android. There is a forum for it though, so there is no need to lump it together with Android.

Guess you haven't read the article. It's full fat android: https://emteria.com/

[hippy]

Android things is not Android. There is a forum for it though, so there is no need to lump it together with Android.

Android Things is not that different to Android though. The main difference is it starts your application at power-up rather than an Application Launcher. From a programming perspective Android Things is exactly as Android though some things are not supported.

Not sure what you mean by "There is a forum for it though" as I cannot find one, and the most recent post asking about Android Things I saw got moved to the locked Android forum.

It's not under Operating System Distributions / Others nor under Programming / Other programming languages where Windows IoT is for some reason hiding out.

If Windows IoT for a Pi is worthy of its own place I cannot see how Android Things for a Pi is any less deserving. Especially when its use is being encouraged by the Foundation through MagPi.

[Imperf3kt]

This would be the Raspberry Pi forums, is what I meant. It doesn't have a subforum for Android Things.

[fruitoftheloom]

https://help.emteria.com/kb/devices-rpi-3

Free to Evaluate and reboot every 8 hours, Personal Edition is €15

It is not Google's Android Operating System.

It is Emteria's own version of Android Things with a Desktop Environment as far as I can ascertain, solely for Raspberry Pi 3B (presumably also the RPi 2B v1.2 and CM3)

Google Play Store is not "officially" supported, it relies on APK packages from 3rd Party Package Managers.

https://support.google.com/googleplay/a ... 5974?hl=en


Bottom line it is not justified to call it Android and all rejoice.......it should be called emteriaOS


There is a version of Android which costs $9 and "supposedly" supports Google Play Store:

viewtopic.php?f=63&t=217004

[bensimmo]

It would be nice to open the 'Android' section back up, at the least to gather links and info on the various sections.

Google AIY, if you are thinking of that, has a forum section (and this is the official help forum/section linked to from them for the AIY).

[fruitoftheloom]

It would be nice to open the 'Android' section back up, at the least to gather links and info on the various sections.

Google AIY, if you are thinking of that, has a forum section (and this is the official help forum/section linked to from them for the AIY).

There is a Community Android Forum

https://groups.google.com/forum/m/#!forum/android-rpi

[jamesh]

This is an entirely third party OS, nothing to do with us. If you are going to use it I would suggest their forums as the best support area.

It's also a cost option if you want the full version.

My opinions of Android and its usefulness to the Pi are unchanged.

[hippy]

This is an entirely third party OS, nothing to do with us.

Apart from using the Pi the same can be said for Windows Iot, Google AIY, most of the programming languages, media centre and gaming ventures.

What is most odd is that it is only Android and Android Things which gets treated this way. If Android and Android Things have "no educational value" it is hard to see how Windows IoT does either. It is difficult to understand why MagPi would 'waste their time' on Android and Android Things.

Half of all Pi sales are to commercial users which of themselves have no direct educational value. They do however generate income for Raspberry Pi Trading and funds for the Foundation. It seems to me there is considerable interest in Android and Android Things and I don't understand why the Foundation would not want the revenue generated from that.

I would argue there is actual educational value in Android. In programming terms it is just a platform and framework like any other and learning to program for Android is little different to learning to program for something else. A useful program running under Android is just as useful as the same running under something else. The platform and framework used is just incidental to programming in most cases.

Programming is one of the things the Foundation says it is keen to encourage, but not it seems if the word Android appears in relation to that. I really cannot see how a program written to run under Windows IoT or Raspbian has less educational value when the same program is written to run under Android or Android Things, why getting a games console emulator running or a Media player installed is more educational than actually writing Android programs.

Billions of people have Android tablets and phones. Just encouraging some of them to write code for what they have would be entirely within the Foundation's remit, would fit its core mission goal, and all the better if they use a Pi for that or it leads them towards using a Pi and expanding their programming and computing interests in the future.

At one point the Foundation seemed in agreement, were enthusiastic in anticipating Android coming to a Pi, but something changed. I doubt we will ever truly know what or why. "No educational value" is the officially stated reason, or pretext and excuse depending on one's viewpoint.

[jamesh]

At one point the Foundation seemed in agreement, were enthusiastic in anticipating Android coming to a Pi, but something changed. I doubt we will ever truly know what or why.

The amount of work required to make it function correctly was huge. The VC4 has been used on Android phones (Samsung), and the team at Broadcom was about 30 people, working for years to get that to a working state ( I was one of them). The RPF have simply never had the team size required to do it, even though a lot was already working. Broadcom sold millions of chips on the back of the Android work, RPT wouldn't make back the development money in new sales specifically down to Android.

[bensimmo]

I guess Amazon benefited from that Android work as well, they sold quite a lot of FireTV sticks and kicked off a whole series of them.

[fruitoftheloom]

I guess Amazon benefited from that Android work as well, they sold quite a lot of FireTV sticks and kicked off a whole series of them.

..not to forget Roku which used the BCM2835 & 7208 in a few products !!

[hippy]

The amount of work required to make it function correctly was huge.

And that's fair enough but it doesn't really explain why near everything Android related is almost immediately cast out to a locked forum. Barely entertaining discussion of Android and Android Things is a long, long way from "we don't have the resources to do development ourselves".

Android and Android Things are considered good enough 'as is' for them to be covered by MagPi, and good enough for people to use if not everyone.

[jamesh]

The amount of work required to make it function correctly was huge.

And that's fair enough but it doesn't really explain why near everything Android related is almost immediately cast out to a locked forum. Barely entertaining discussion of Android and Android Things is a long, long way from "we don't have the resources to do development ourselves".

Android and Android Things are considered good enough 'as is' for them to be covered by MagPi, and good enough for people to use if not everyone.

Because we were getting a lot of Android support questions - for an OS we don't support. But I suppose that argument could be made for WinIoT etc, but there again, we have hardly any support requests for those other OS's even now. I suppose we could just reopen the sub forum and just ignore it, from a RPF point of view.

[jamesh]

The amount of work required to make it function correctly was huge.

And that's fair enough but it doesn't really explain why near everything Android related is almost immediately cast out to a locked forum. Barely entertaining discussion of Android and Android Things is a long, long way from "we don't have the resources to do development ourselves".

Android and Android Things are considered good enough 'as is' for them to be covered by MagPi, and good enough for people to use if not everyone.

Because we were getting a lot of Android support questions - for an OS we don't support. But I suppose that argument could be made for WinIoT etc, but there again, we have hardly any support requests for those other OS's even now. I suppose we could just reopen the sub forum and just ignore it, from a RPF point of view.

And that's what we have done - the Android forum is reopened. However, do not expect Raspberry Pi engineering support!

[hippy]

I suppose we could just reopen the sub forum and just ignore it, from a RPF point of view ...

And that's what we have done - the Android forum is reopened. However, do not expect Raspberry Pi engineering support!

Many thanks and that's good enough for me.

[fruitoftheloom]

I suppose we could just reopen the sub forum and just ignore it, from a RPF point of view ...

And that's what we have done - the Android forum is reopened. However, do not expect Raspberry Pi engineering support!

Many thanks and that's good enough for me.

+1


..for completeness it would appear that the various ""Android"" offerings are based either:


A0SP

https://source.android.com


Android Things

https://developer.android.com/things/

[bensimmo]

I'm not sure anyone expects help from the engineers in questions, even RetroPi, Kodi etc are community helped and then told to go have a look at their respective websites if nobody knows.

It will allow people to advise where to look for help now.

[feelslikeautumn]

The amount of work required to make it function correctly was huge. The VC4 has been used on Android phones (Samsung), and the team at Broadcom was about 30 people, working for years to get that to a working state ( I was one of them). The RPF have simply never had the team size required to do it, even though a lot was already working. Broadcom sold millions of chips on the back of the Android work, RPT wouldn't make back the development money in new sales specifically down to Android.

Surely these latest android builds are just using Eric Anholt's vc4 driver? I very much doubt there is a huge amount of work to get android working on the pi 3.

I don't know a lot about android, but would it be possible to take an image made for a phone, add a raspberry pi kernel and boot it up?

Compiling from source seems to require an awful lot of resources (memory/hard drive space) which makes it impossible for a lot of people (myself included). I would be very interested in an official raspberry pi version of android. A lot of IoT uses seem to have a smartphone frontend, so I think there is value in providing an environment to develop and test these applications.

[jamesh]

The amount of work required to make it function correctly was huge. The VC4 has been used on Android phones (Samsung), and the team at Broadcom was about 30 people, working for years to get that to a working state ( I was one of them). The RPF have simply never had the team size required to do it, even though a lot was already working. Broadcom sold millions of chips on the back of the Android work, RPT wouldn't make back the development money in new sales specifically down to Android.

Surely these latest android builds are just using Eric Anholt's vc4 driver? I very much doubt there is a huge amount of work to get android working on the pi 3.

I don't know a lot about android, but would it be possible to take an image made for a phone, add a raspberry pi kernel and boot it up?

Compiling from source seems to require an awful lot of resources (memory/hard drive space) which makes it impossible for a lot of people (myself included). I would be very interested in an official raspberry pi version of android. A lot of IoT uses seem to have a smartphone frontend, so I think there is value in providing an environment to develop and test these applications.

Sorry, cannot help being harsh, but you clearly know nothing about the complexity of making Android work properly. It's a horrible and difficult job.

[feelslikeautumn]

Yes you are right I know nothing about android. But I've already said that.

I know that none of these android builds have had 30 people working on them for years.

I know you continually exaggerate things to push Raspbian.

[jamesh]

Yes you are right I know nothing about android. But I've already said that.

I know that none of these android builds have had 30 people working on them for years.

I know you continually exaggerate things to push Raspbian.

Let me explain. When I worked at Broadcom, I worked on Android (and the changes required on VC4 to make it work). I spent 10 weeks in 2013/14 in S.Korea working on a tiny part of the Android system for Samsung (camera, multimedia). I was one of a rotating apps team, I think there were 10 of us rotating from the UK, joining a team of Brcm Koreans of another 6 or 7. Now include the Broadcom team who sat behind their computers back in the Cambridge office (another 10 at least) AND the Sunnyvale office (not sure, at least 10 I think). We worked on this crap for 2 years.

I've also worked for another company on Android, that was also horribly painful.

Getting Android to work PROPERLY is a big big job. If a smaller team manages to get a subset working that's great, but have they got the camera working? Full graphics acceleration? Multimedia acceleration? And remember, they are working on top of work already done. Eric's driver has been in development for three years now - still not finished.

And what's even more fun, every new Android release means a lot of that work has to be done again, because they changed some API, or the way memory is allocated, or other such PITA.

Android is horrible internally - unbelievable complex. There is a very good reason why so many manufacturers fail to update their phone or tablets ranges to newer android version.

So yes, I will push Raspbian over Android because it's better for education, easier to use, easier to develop on, faster, more compact, and doesn't suffer the whims of the Google Android development team.


* These Android views are my own, but I could point you in the direction of lots of other people who have worked on it with very similar opinions.

[feelslikeautumn]

Getting Android to work PROPERLY is a big big job. If a smaller team manages to get a subset working that's great, but have they got the camera working?

Yes. And the official touchscreen.

Full graphics acceleration? Multimedia acceleration?

No idea, but Linux doesn't have full multimedia acceleration either.

And remember, they are working on top of work already done.

Which was my point.

Eric's driver has been in development for three years now - still not finished.

I'm sure he welcomes your support and encouragement. I'm truly baffled by the negative response that the vc4 driver always receives on this forum.

And what's even more fun, every new Android release means a lot of that work has to be done again, because they changed some API, or the way memory is allocated, or other such PITA.

Android can now use the regular Linux graphics stack.

So yes, I will push Raspbian over Android because it's better for education, easier to use, easier to develop on, faster, more compact, and doesn't suffer the whims of the Google Android development team.

You are ignoring the benefits of Android, the main one being it is something that people actually want to run with its huge range of apps. This obsession with Raspbian is going to end up harming the pi. In particular the development of the Raspbian Desktop is just wasting time that could be spent on say finishing the vc4 driver.

[jamesh]

Getting Android to work PROPERLY is a big big job. If a smaller team manages to get a subset working that's great, but have they got the camera working?

Yes. And the official touchscreen.

Full graphics acceleration? Multimedia acceleration?

No idea, but Linux doesn't have full multimedia acceleration either.

And remember, they are working on top of work already done.

Which was my point.

Eric's driver has been in development for three years now - still not finished.

I'm sure he welcomes your support and encouragement. I'm truly baffled by the negative response that the vc4 driver always receives on this forum.

And what's even more fun, every new Android release means a lot of that work has to be done again, because they changed some API, or the way memory is allocated, or other such PITA.

Android can now use the regular Linux graphics stack.

So yes, I will push Raspbian over Android because it's better for education, easier to use, easier to develop on, faster, more compact, and doesn't suffer the whims of the Google Android development team.

You are ignoring the benefits of Android, the main one being it is something that people actually want to run with its huge range of apps. This obsession with Raspbian is going to end up harming the pi. In particular the development of the Raspbian Desktop is just wasting time that could be spent on say finishing the vc4 driver.

Wow.

1. Not dissing Eric's driver. He was sitting next to me in the office the other day, nice bloke, very clever. But it has limitations that we still haven't got round.
2. Developing Raspbian desktop requires an entirely different skills set to Mesa drivers. But we do have a guy helping Eric who does have the required skills set, difficult work though.
3. Yes, there would be benefits to having full Android. However the required work, and the time taken off other projects make those benefits not cost effective at this time. It would probably take our entire SW engineering staff over a year to get Android working properly. Time that is currently spent developing new products, improving current products, and fixing bugs in existing product. * See below
4. What is 'the regular Linux graphics stack'?
5. The Pi is a general purpose (mostly) desktop computer, Raspbian is a general purpose OS that suits it perfectly. Android isn't really suitable for desktop use, and is certainly not general purpose.

Come back in a year or two. If the lack of Android has dented our sales we can discuss it then.

And just so people understand, AFAIK, there are no plans whatsoever to develop an in house Android. However, if other people make the effort, that's great. However, the fact the Pi has been out over 5 years, and there still isn't a really good implementation should speak volumes.

* Cost benefit analysis. Lets say it takes a team of 10 people 1 year to make ANdroid work properly on the Pi, Lets say each engineer costs $150k a year. Let's say the profit per Pi is $5 (I have not idea what the profit is, using that as an example figure). So, cost to develop is $1.0M. We would need to sell an extra 300,000 Pi to pay for that development. Not a huge number but significant. BUT, we have now lost a YEAR of development time on other products, for example, the Pi4 is now a year later than planned. We sell about 5M devices a year, let's say we lose about half of that as the Pi3 becomes 'obsolete', so that 2.5M device sales lost, so the Android port would need to pull in 2.5M extra sales. It just does not add up.

And just to say, this sort of conversation is the reason we closed the Android forum in the first place.