On 8/25/20 4:46 PM, Nick Desaulniers wrote:
This would solve a common pattern in the kernel where folks are using
`extern inline` with `gnu_inline` semantics or worse (empty `asm("");`
statements) in certain places where it would be much more preferable
to have this attribute. Thank you very much Martin for writing it.
is direct equivalent of Clang's no_stack_protector.
Unlike Clang, I chose to name it no_stack_protect because we already
have stack_protect attribute (used with -fstack-protector-explicit).
That's pretty easy for us to work around the differences in the
kernel, but one final plea for the users; it would simplify users'
codebases not to have to shim this for differences between compilers.
If I had a dollar for every time I had to implement something in LLVM
where a different identifier or flag would be more consistent with
another part of the codebase...I'm sure there's many examples of this
on LLVM's side too, but I would prefer to stop the proliferation of
subtle differences like this that harm toolchain portability when
possible and when we can proactively address.
All right, I'm open to unify the flag name to what LLVM has ;)
Martin
