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

[Rafael Auler]

Hi Shankar,

Thanks for putting in time to review this. I agree that we should not add this atom in dynamic libraries, I did this to replace the previous code that would always add a definition of tls_get_addr regardless the output type, but removing this atom would still work well. However, I am not sure if I fully understood your last comment -- I cannot see why we should remove this atom from ExecutableWriter, since, to complete linking an executable, all symbols must be resolved and we cannot finish it with undefined references. 

Best regards,
Rafael Auler

http://reviews.llvm.org/D5505