conrade-ctc wrote:

> @mizvekov @cofibrant thanks for the revert!
> 
> @conrade-ctc @vgvassilev I just took a look locally -- `__emutls_get_address` 
> is implemented In the compiler-rt builtins library, which clang-repl doesn't 
> link by default. I think that this fix in this patch is correct, but to make 
> the testcase (and thread locals in general) work on Darwin your options are: 
> (1) add the compiler-rt builtins library to the Platform JITDylib (so that 
> __emutls_get_addr can be found that way), (2) add an absolute symbol for 
> `__emutls_get_addr` in libClangInterpreter (this will force 
> `__emutls_get_addr` to be linked into the clang library), or (3) use native 
> TLS on Darwin (you'd need to also link the ORC runtime).

Ya, makes sense. The absolute symbol pathway is the smallest scope/targeted 
approach I think. I'll put up a new PR with an update along those lines and tag 
you so you can review to make sure it makes sense to you. Thanks for the 
pointer @lhames 

https://github.com/llvm/llvm-project/pull/207161
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to