[lld] r254428 - [ELF] - Refactor of tls_index implementation for tls local dynamic model.

[Michael Spencer via llvm-commits]

On Thu, Dec 3, 2015 at 2:01 AM, George Rimar <grimar at accesssoftek.com> wrote:
>>addLocalModuleTlsIndex was the correct name. It is the TLS Index
>>object for the local module.
>>
>>- Michael Spencer
>
> Ok. My point was not about it is incorrect, but to make it a bit close to specs.
> Urlich has next description "to generate a tls index entry for symbol x1 this code uses x1 at tlsld(%rip) which creates a special kind of index which refers to the current module" so
> module is called as "current" there. At the same time he talks about Local Dynamic TLS Model. "local model" is used often in whole spec, but "module" is always "current" I think.
> So for me addLocalModelTlsIndex is a little closer to what spec uses.
>
> George.

addLocalModelTlsIndex is wrong. It is adding a tls_index entry for the
module being linked, not for the "local model". There are also
tls_index entries created which may not refer to the module being
linked. I'm fine with the name addCurrentModuleTlsIndex, but
addLocalModelTlsIndex is wrong, and "local model" is never used in the
spec.

- Michael Spencer