The Project To Compile The Linux Kernel With LLVM Clang Is Still Alive

[TeoLinuX]

Pretty much this.

Granted, no matter how useful, compiler specific behavior doesn't belong in the kernel. So I side with LLVM in this instance; the kernel should really be free of any platform limiting behavior, period.

I'm with you, but many people give a rationale for the GCC specifics:
https://stackoverflow.com/questions/...cific-features

http://www.ibm.com/developerworks/library/l-gcc-hacks/

[sarfarazahmad]

I'm not holding my breath, but I hope so. Then all that would be left to shut Stallman fans up about that "GNU/Linux" name (too impractical and inaccurate to ever happen for a laundry list of reasons) is seeing BusyBox+musl mature into something closed-source apps get along with more consistently.

yep you got that right. Docker is moving to Alpine Linux. Suggest you read up on that too if you havent already

[smitty3268]

Could you be more specific about those "20 places" where Linux kernel is unavoidably gcc-dependant ?

Nothing is "unavoidable". The devs just believe that their current code is better than it would be if they took out the GCC-specific features, so they are understandably reluctant to go "backwards". Especially when it affects old crufty code that has been around for a long time and is complicated to debug and rewrite.

The specific issue i was thinking about in particular happened to be "variable length arrays in structs", but you can see a whole host of what they're working around here: http://llvm.linuxfoundation.org/index.php/Bugs

It's all just random stuff spread around everywhere.