Building images without compiling kernel myself ?

[Aleks]

Edit: Uh i guess it should be in the "Development" section instead ... but I can't find how to move the topic :s

 

Hellow,

 

thank you very much for all the work achieved by the Armbian project.

 

I am using Armbian to build images (using https://github.com/armbian/build ) that I provide to users, with custom packagers installed via userpatches and it works pretty well ! However, the whole "kernel compiling" thing is quite mysterious to me and I don't really understand it.

 

My questions are :

- I am confused by the whole "default", "next" and "dev" terminology ? I am guessing this is some sort of stable / testing / unstable, but not sure ... Is using "next" okay ? Is "next" less stable than "default" ?

- Most of the building time is spent compiling the kernel image. I am guessing that this is needed at some point, because boards have specificities and the kernel must be patched to handle each of them correctly (?). But for my use case, I am not really interested in compiling the kernel myself, I would rather use the stuff already available in Armbian's repo (apt.armbian.com) directly. So : is there any way I could disable the build of the kernel and just use "official" builds when building my images ? That in fact goes beyond just the kernel I believed, as I saw some packages like "armbian-firmware" are flagged as "local".

 

Sorry if these questions are already addressed somewhere obvious ... I've been trying to find answers to those elsewhere but couldn't find anything :/

 

Cheers !

Edited January 15, 2019 by Aleks
Bad category

[Werner]

Default represents the legacy kernel (often 3.4.113)

Next is stable mainline kernel (currently 4.19.x)

Dev is development and does not depend on a specific kernel version. It is all WIP, testing, experimental.

(Values may vary depending on SBC/vendor)

 

On SBCs it is not possible to compile a generic kernel that basically runs on all devices out there as ARM does not have any methods (like PCI or ISA bus on x86/amd64 architecture) to probe which devices the kernel is actually running on. Therefore the kernel has to know beforehand what the hardware it will run on. So every device needs at least its unique device tree compiled in.

 

Usually the second time you compile the same image it should go way faster if nothing has changed at the config as the build script is using ccache to buffer earlier built code.