为什么在编译Linux内核和uBoot时使用arm-linux-gnueabi-gcc而不是arm-none-eabi-gcc？

Question

I have some experience compiling bare metal code for ARM cortex-m devices as well as the Linux kernal, uBoot, and applications for the Beaglebone Black (BBB) (more featured ARM with MMU, for those living under a rock). It makes since to me that the cortex-m code should be compiled using arm-none-eabi-gcc (as there is no OS) and the application code for the BBB should be compiled with arm-linux-gnueabi-gcc (as there is an OS, for which system calls can be made and program loaders and shared objects that can be utilized).

What I don't understand is why uBoot and the kernel also should be compiled with arm-linux-gnueabi-gcc. In my mind, uBoot at least, is a bare metal program with no fancy OS to account for. This has been bugging me for sometime, but I can't find an answer. Is there anyone out there that can enlighten me?

Answer 1

U-Boot was designed to mirror the Linux design philosophy as much as possible. It uses the same configuration system, general directory structure, etc. It shares some APIs with Linux - see the include/linux directory. As the comments above mention, the ABI compatibility doesn't matter at this point, but using a Linux compiler isn't philosophically inappropriate for U-Boot.