[PATCH] [lld] [ELF] Support for general dynamic TLS relocations on X86_64

[Rafael Auler]

Hi Joerg,

I see, I agree that this does not mix well with static linking, I was very much glibc-focused. In fact, this is a preliminary patch that addresses only the first TLS general dynamic use case, I would work in the static cases later. To address both dynamic and static cases, I think that Kledzik's suggestion of putting weak symbols to work will fit in perfectly: if the linker doesn't find any objects defining tl_get_addr at build time, it will leave to the dynamic linker at runtime. What do you think?

http://reviews.llvm.org/D5505