Linux/Android ARM-hosted RISC OS?

Primarily it would need to work with a mouse and keyboard, any additional tablet mode support would be a bonus.

People were talking about Android in which case touch is required, keyboard is additional and mouse would be an unlikely bonus.

So you’d need a custom kernel to do it this way? Which would then remove the millions of devices advantage.

Not a custom kernel. Something completely different.

I don’t see that.

Take the big picture. It’s much easier to make RISC OS running under KVM. No need to make a whole new AMP kernel (incompatible with Linux, so not Linux), no need to port KVM to Android (it would need a specific Linux kernel). You run it on an ARM server, and once virtualized, you can use it on your connected device with a Spice client.

Primarily it would need to work with a mouse and keyboard, any additional tablet mode support would be a bonus.

If the idea is to use it with keyboard and mouse, why not simply use a Pi? is it too pricey?

If the idea is to use it with keyboard and mouse, why not simply use a Pi? is it too pricey?

No idea, I wasn’t the one with the idea of running RO on top of Android.
Android does have a number of off the shelf sets of hardware though so all the fiddly construction surrounding a Pi based tablet does go away.

Curiosity value from my viewpoint.

Android does have a number of off the shelf sets of hardware

Indeed, and since each contract renewal gets me a new telephone at a bargain price (my S5 Mini cost me €100 with an immediate rebate of €50 and a further €50 off my bills), I seem to have a collection of old phones. One I use as a tablet. One I don’t use, and the other – a Sony Ericsson Xperia Mini actually manages to fit in a slide-out keyboard, so with the right driver it could make an interesting RISC OS machine. There’s a blue function key on the keyboard. That plus cursors could move the mouse (in addition to touch), with a a long tap of it being Menu? It could have been a nice machine for the Beeb emulator, but that doesn’t work well with the keyboard. http://www.heyrick.co.uk/random/pics/xperiabeeb.jpeg

I also did Win3.11 for a laugh under DOSbox. Same issue, but actually worked better than I thought it would.

Problem is, while there are plentiful amounts of pre-built Android hardware around (the number of cheap’n’cheerful tablets coming out of China is ridiculous), the documentation at a level to get an operating system running is……less plentiful.

If the idea is to use it with keyboard and mouse, why not simply use a Pi? is it too pricey?

It would be a portable device. It would have wifi and mobile data via the host. It would still have access to all the Android software.

It could provide a virtual keyboard and mouse. Transparent and in front of the RISC OS screen, with a trackpad that passes on regular mouse movements to RISC OS.

However the revelation that such a system would have to run on a VM, rather than being a subsystem that interfaces directly to Android reduces the attractiveness.

It would be a portable device. It would have wifi and mobile data via the host. It would still have access to all the Android software.

I have an Android tablet that I’ve kind of made to “run” RISC OS (read on before you get too excited!)

It could provide a virtual keyboard and mouse. Transparent and in front of the RISC OS screen, with a trackpad that passes on regular mouse movements to RISC OS.

It does this!

I use Jeffrey’s VNC Server (and Steve Potts VNC Front End, just for ease of use) on a Raspberry Pi, which is connected directly to (and powered from the USB port on) my router. I set up a free DDNS account on my router, opened and directed a port suitable for VNC, and I just connect to it that way. On Android, the screen acts as a giant touchpad and a semi transparent keyboard can be called down at any time. Using other computers, the existing keyboard and mouse devices work as normal. It should be noted that the Android RealVNC app allows “middle clicks” as well as left and right clicks, using a different number of fingers to click.

Obviously, sound and direct downloading onto the tablet/device aren’t available.

Probably not what you were hoping for, but it works like RISC OS on Android in a way!

However the revelation that such a system would have to run on a VM, rather than being a subsystem that interfaces directly to Android reduces the attractiveness.

Why? VM runs at almost (99%) native speed today.
But you need an hypervisor, so a specific Android kernel.

I use Jeffrey’s VNC Server (and Steve Potts VNC Front End, just for ease of use) on a Raspberry Pi, which is connected directly to (and powered from the USB port on) my router

That’s an indirect request for Jeffrey :)

I work on many professional software. To be able to connect the Pi on a PC USB port, to see it as an Ethernet device, then to use it with a VNC client and a Samba share would be fantastic. Connect, and use from a PC with very little configuration and just one cable. That would permit me to make my software for RISC OS and not for both RISC OS and Windows.

That’s probably too much job, but if it’s possible, a can finance it (a bit).

It’s not exactly what you’re asking for but a while ago I did manage to get a build of ArcEm running on an Android phone. I can’t release it at the moment because I borrowed code from another emulator for the Android UI. It ran RISC OS 3 reasonably well but wasn’t really usable due to keyboard and mouse issues (keys not mapping correctly, mouse difficult to position for example). I also never got HostFS or floppy emulation working. One day I might revisit this project but I wouldn’t hold your breath.

Why? VM runs at almost (99%) native speed today.

What I was thinking of was a version of RISC OS that would interface with Android directly, however as far as I can tell it would need Android to be able to allow complete control of one core, which apparently isn’t possible (on a standard version).

Were this possible, the advantage would be simplicity. ie you install one App and you have a RISC OS environment. (With indirect support for WiFi, mobile internet, bluetooth devices, etc.)

However with a VM you need a working VM, with maintainers, and then have to have maintainers for the client version of RISC OS. The end result would be likely to require far more technical knowledge for the end user. I don’t think this scenario would provide the rewards (i.e more users) that the (sadly) impossible one would have.

What I was thinking of was a version of RISC OS that would interface with Android directly

I understand your point, but if Linux works on one core, RISC OS will not running on the other like this. There is no magic.

Case 1: RISC OS provides its own drivers. Then you need a version of RISC OS that could run on the same hardware WITHOUT Android or Linux and the two kernels must be changed to understand AMP mode. Just forget it: this is not possible.
Case 2: RISC OS relies on another kernel. Then you need drivers to talk to the core kernel. Windows only, as only the NT kernel provides possibilities to define subsystems. Marginaly faster than virtualization, and complex to maintain.
Case 3: RISC OS relies on an hypervisor. Then you need to support the virtual hardware provided by the hypervisor. RISC OS will be a virtual machine, with 90% of the native speed, 98-99% if you provide paravirtualisation drivers (then RISC OS will interface more directly with the Linux/Android kernel).

Last solution: emulation. Very slow, but no need for anything else.

ie you install one App and you have a RISC OS environment.

It’s the same with virtualization.

I don’t understand why you do not want this solution. That’s the one Linux provides to make other OS running. It’s simple, easy to use, available under Linux and soon available under Android. And since Qemu (used by the KVM hypervisor) provides a RPI emulation, RISC OS almost boots on it ‘as is’. So where is the problem? We have here a fast path to make RISC OS running on any ARM hardware, and even on other processors.

A friend of mine manages to compile a version of Qemu that can make RISC OS running. It was with some patches. The best would be a specific version of RISC OS, so no need for patches any more on the Linux/Android side. Direct support.

I don’t understand why you do not want this solution.

I wouldn’t say that exactly. What I wouldn’t want is this solution taking priority over other things.

A viable system would require suitable QEMU to be well maintained plus a RISC OS port being maintained.

The latter should only be done if the former is pretty certain.

It is. Qemu is the base of KVM, the main hypervisor of every Linux solution.

Be compatible with it, and RISC OS will work unchanged on any forthcoming ARM server in the world, on a growing number of ARM desktop solutions, and sooner or later, on some Android devices too. And of course in emulation mode on every Windows, OS X, Linux and BSD systems available.

Some Qemu ports are available on Android for the x86 part. So it could be done for the ARM mode too. But beware, Qemu is just an emulator, so much slower than KVM+Qemu. Intel managed to get a version of Limbo (based on Qemu) that can use KVM on a specific Android kernel, with almost native speed for the guest system.

Industry needs hypervisors. So let’s them doing this hard part of the work :)